Episode 5: Identifying Data Science Use Cases for your Business
Download MP300:00:00 Dr Genevieve Hayes
Hello and welcome to value driven data science, brought to you by Genevieve Hayes Consulting.
00:00:06 Dr Genevieve Hayes
I'm your host, doctor Genevieve Hayes, and today I'm joined by guest Rob Deutsch to talk about identifying data science use cases for your business. Rob, welcome to the show.
00:00:16 Rob Deutsch
No, thank you so much. Listeners might not know this, but we recording this after you've released the first couple of podcasts, which I've had a a chance to listen to.
00:00:26 Rob Deutsch
So I can say I'm a I'm a fan and I'm really honoured to actually be on the show as a guest, so thank you for having me.
00:00:33 Dr Genevieve Hayes
I am very happy to have you here today, and Rob is a man of many talents. In addition to being the Chief Operating officer of Aku Shaper, a company that uses advanced modelling algorithms and software to build better surfboards faster, he's also a data science consultant with parity analytic.
00:00:54 Dr Genevieve Hayes
And previously founded Boxer, which built software for creating better financial models.
00:01:00 Dr Genevieve Hayes
And because of the diversity of his roles, he has had a lot of experience in identifying data science use cases in places that might not have ever crossed your mind, like, for example, surfboard design.
00:01:13 Dr Genevieve Hayes
So how about we start start with that?
00:01:17 Rob Deutsch
Sure, sure. What would you like?
00:01:19 Rob Deutsch
To know about surfboard design and.
00:01:21 Rob Deutsch
How how it links to data analytics?
00:01:24 Dr Genevieve Hayes
And that's interesting. When I was first introduced to you, Rob, so I was introduced to rob by Amanda, who is our guest on episode one.
00:01:33 Dr Genevieve Hayes
And I remember checking your LinkedIn profile before we first met and thinking I must have the wrong Rob Deutsch, because I couldn't figure out how someone could go from data science to being Chief operating officer of a surfboard design company. But then it turned out it was the right Rob Deutsch.
00:01:53 Dr Genevieve Hayes
So I suppose that's the question that is probably on a lot of that a lot of the listeners are probably wondering about.
00:02:01 Dr Genevieve Hayes
How do you go from being a data scientist to making surfboards?
00:02:07 Rob Deutsch
I see are so quite a lot of skills I found from my data science background are more broadly applicable in my career.
00:02:17 Rob Deutsch
So to staff to start with at Aku shaper, what we do is we make software for designing better boards faster. That's the surfboard design and.
00:02:27 Rob Deutsch
The the tool parts that are needed to eventually cut that.
00:02:30 Rob Deutsch
Surfboard out of.
00:02:31 Rob Deutsch
Like a phone. So behind that is a surfboard model, a 3D model that needs to be modified. It's very mathematical.
00:02:37 Rob Deutsch
Or when it comes to generating surf parts, that's sorry toolpaths. I should say. That's a very mathematical process as well, which my background in data analytics, data science transfers very nicely to.
00:02:52 Rob Deutsch
We're also assess company software as a service, like a like so much software is these days, you pay a monthly fee and you get access to the software. And as part of that monthly fee, we're continually improving the software as well.
00:03:06 Rob Deutsch
No, but it also means that we get a lot of usage data about how people are using the software, what they're doing with it, and how they're using it.
00:03:15 Rob Deutsch
So there's tremendous opportunity for us to utilise that, to examine that, to analyse that do I help improve the software and help our customers improve their are their boards.
00:03:27 Rob Deutsch
Building faster, build them better.
00:03:29 Rob Deutsch
Lastly, when it comes to actually running the company background in data science means I'm I like to think I'm pretty finely attuned to to problem solving which are running.
00:03:39 Rob Deutsch
The company generally involves a lot of and in data science you spend a lot of time thinking about the problem you're working on, making sure you're tackling the the right problem.
00:03:49 Rob Deutsch
And making sure you're tackling it in a smart way. You always try and kind of exist the layer above where you actually are and saying is this the right path I'm taking?
00:03:58 Rob Deutsch
Which is a another really useful life skill that's transferred well into into running our crew shaper.
00:04:05 Dr Genevieve Hayes
I think that's interesting what you're saying about understanding problems. When I think of data science problems, I usually sort of think of any data science project is actually 2 interrelated problems.
00:04:19 Dr Genevieve Hayes
So you've got your business problem and you've got your data problem. So the business problem might be how do we get more customers? And your data problem might be how do you build a predictive model.
00:04:32 Dr Genevieve Hayes
And I think it would be nice if people came to you with well defined data science projects that clearly specified each of these problems.
00:04:41 Dr Genevieve Hayes
But I mean, have you ever had anyone actually come to you with a nicely scoped out data science project?
00:04:49 Rob Deutsch
You know in my.
00:04:51 Rob Deutsch
In my career, I don't think that has ever been the case.
00:04:55 Rob Deutsch
Usually in my experience and I'd be interested to hear about your experience, Genevieve.
00:04:59 Rob Deutsch
People are never really.
00:05:00 Rob Deutsch
Sure about the problem they're trying to solve.
00:05:03 Rob Deutsch
And for what it's worth.
00:05:04 Rob Deutsch
I, I don't think that's necessarily.
00:05:05 Rob Deutsch
Unique to data science. I'm I grew up, so to speak, in consulting companies, so doing actuarial consulting, so helping companies.
00:05:15 Rob Deutsch
To manage their financial risk and also the the data analytics consulting, the data science consulting and just as a general rule of thumb in this kind of corporate consulting work.
00:05:24 Rob Deutsch
Old it's kind of assumed that the the client who you're working with doesn't necessarily know what their problem is, and part of the value you bring is helping them to find that problem, to find the biggest problem and a pathway to a solution.
00:05:39 Dr Genevieve Hayes
So do you have a process for identifying and understanding those problems and finding that pathway?
00:05:46 Rob Deutsch
Gee, that's that's kind of the $1,000,000 question, isn't it?
00:05:52 Rob Deutsch
Like I said, I grew up in consulting companies and they invest some some time, effort and resources into kind of teaching you some some good tools you can use to come.
00:06:03 Rob Deutsch
To identify these problems, to solve the problems.
00:06:07 Rob Deutsch
Uhm. I think data science though in particular poses some unique problems, because the opaqueness of the the techniques themselves and what's possible often isn't common knowledge. And so especially when you're in the.
00:06:26 Rob Deutsch
The solution development phase, you can find that the customer I use the word customer very broadly. I'm kind of using customer to refer to someone who has a problem and they they think they want to solver with data science. So it could be an actual customer, it could be your boss, there could be another business department.
00:06:43 Rob Deutsch
But either way, that customer generally doesn't have a good understanding about what is possible, what is impossible with data science, and beyond that, how much time and effort things generally take.
00:06:55 Rob Deutsch
So what I'm personally a fan of, and I actually heard this best stated by a consultant at EY.
00:07:03 Rob Deutsch
Who gave a talk on the process that they use when they run workshops are with their clients data workshops to try and extract value from their data.
00:07:11 Rob Deutsch
And they say that they have a rule in these workshops which I'm I'm sure they get paid a lot of money to hold these workshops with the outboards, with senior management.
00:07:19 Rob Deutsch
But the the rule is that in the discussions about potential opportunities are with data, they're not allowed to talk about data itself. They're not allowed to talk about what data.
00:07:32 Rob Deutsch
How clean the data is, whether it's expensive because they want to purely focus on the problem to be solved first.
00:07:40 Rob Deutsch
On the assumption that once you properly identify a problem that could hypothetically be solved with data, that's when you can start getting into the nuts and bolts, because I think it's very easy to to go down the.
00:07:52 Rob Deutsch
The garden path, so to speak. When you're looking at our problems and database solutions or data science solutions to just say I'll look.
00:08:02 Rob Deutsch
A certain bit of data is going to be very difficult to get, or we just know it's going to be messy.
00:08:07 Rob Deutsch
For example, it could be a piece of data that your company has been out gathering over the past 10 or 20 year, 10 or 20 years to varying levels of our refinement.
00:08:18 Rob Deutsch
But if you think too much about that.
00:08:22 Rob Deutsch
You might end.
00:08:23 Rob Deutsch
Up, stopping, evaluating a problem are not really thinking about just how valuable it would be.
00:08:30 Rob Deutsch
To solve that problem.
00:08:31 Rob Deutsch
And consequently, you might find out that it's worthwhile to take the time to solve your data problem, so to speak, to to clean up the data to artists, or some data to buy some external data come.
00:08:43 Rob Deutsch
Yeah, you really want to be mindful that if you get too deep in the weeds, you might you might cut off a valuable project before you actually get started.
00:08:52 Dr Genevieve Hayes
Yeah, that's actually an interesting point Rob, 'cause I've actually found the opposite in practise. What I've actually found is organisations often spend start by focusing on the data that they do have and focusing on that too much at the expense of the business problem.
00:09:12 Dr Genevieve Hayes
So, for example, an organisation might have one particular data set and they become obsessed with coming up with a data science project.
00:09:23 Dr Genevieve Hayes
That arises from that data set even though, even if it doesn't actually solve any business problems that are of interest to them.
00:09:33 Rob Deutsch
Absolutely. I'm I've seen the exact same thing and that's I think another danger point that you can have if you spend too long thinking about the data itself as opposed to the the business problem you have or specifically your business strategy, you know if you're in senior leadership and and I think it's somewhat linked to that eternal question.
00:09:52 Rob Deutsch
One of our company strategy, business strategy of do you try and think up a bunch of good ideas for your company and then investigate each one to see which one might work?
00:10:03 Rob Deutsch
Best or do you go for a more precise?
00:10:06 Rob Deutsch
Some more precise method where you speak to a customer for example, and you find out what their biggest problem is and you try and find a solution to that.
00:10:17 Rob Deutsch
Uhm, I think if you speak to different business leaders, you'll get different, different answers. There's whole books written on each of those two approaches, the scatter gun approach versus the precision scalpel approach.
00:10:29 Rob Deutsch
And I think the same thing just exists in in data science. And then again, there's no there's no perfect solution, but there are things in mind.
00:10:37 Rob Deutsch
A lot in common traps that you can avoid if I'm if you're careful.
00:10:42 Dr Genevieve Hayes
And what common traps have you found?
00:10:44 Rob Deutsch
So I think the ones we've talked about so far are by far the most common. It's either assuming a piece of data isn't available, so you stop thinking about it.
00:10:54 Rob Deutsch
Then there is just thinking too much about a particular set of data. Uh, 1/3 notable one. Though actually it's a very good question 'cause other notable one is getting.
00:11:04 Rob Deutsch
Too excited about complex data science techniques?
00:11:09 Rob Deutsch
Buzzwords and the like. I am a big proponent that the simplest solution is quite often the best solution. That old adage of be as simple as possible but no simpler.
00:11:21 Rob Deutsch
So I try not to get too excited about it, but I'm quite fond about Excel as a data science to log tableau for visualisation.
00:11:29 Rob Deutsch
Because especially if you're an organisation that's still UN maturing in your data science process, a lot of low hanging fruit can be picked up relatively easily with our with simple tools, simple techniques.
00:11:43 Dr Genevieve Hayes
I was actually talking to the manager of Analytics team at a organisation that has relatively low data maturity and her team has only just got an SQL Server and only just Scott and I think its power BI but up until recently.
00:12:03 Dr Genevieve Hayes
They have been working entirely in Excel.
00:12:06 Dr Genevieve Hayes
And it sounds like they've actually been doing some reasonably good data science. They can still identify trends in data, identify anomalies, and answer any of the questions that her senior managers have to ask.
00:12:22 Rob Deutsch
Which is really the most important thing, answering questions. And the simplest, quickest way to get to those answers is is usually the best way.
00:12:31 Dr Genevieve Hayes
And because her team only had excel to work with up until recently, they actually spent a lot of time doing what she calls intelligence. So actually speaking to the owners of this data and understanding things. So they weren't.
00:12:51 Dr Genevieve Hayes
Focusing obsessively on data science techniques, they are seeing them as part of this broader job, which also involved speaking to people.
00:13:02 Rob Deutsch
That's that's really interesting 'cause I I find that when speaking to someone, that's when you discover what the real business problem they're trying to solve is.
00:13:10 Rob Deutsch
Because usually there's there's a hierarchy 'cause all businesses at the end of the day they want to increase their revenue and decrease their expenses. That would increase their profit, but that's not a a problem that data science.
00:13:22 Rob Deutsch
Can solve directly. Usually you have to go a few layers into the onion, so to speak before you find a problem you can solve with data science and and sometimes it's not.
00:13:32 Rob Deutsch
Necessarily what you think?
00:13:33 Rob Deutsch
Like usually it is, sometimes it isn't, and when it's not, that's when you can spend the most amount of time, effort, resources, expense, going down a rabbit hole that provides no no results, which is not what anyone wants.
00:13:46 Rob Deutsch
I'm actually listening to a podcast totally not related to data science on the weekend where the host claimed.
00:13:53 Rob Deutsch
That Metallica is not in the business of selling selling music, because these days, apparently, there's not a lot of money to be made by selling CDs or even by streaming.
00:14:03 Rob Deutsch
I I believe I please don't quote me on this. I'm not an expert. I'm purely repeating. Are the point of views of this podcast host that bands tend to make money by doing live performances?
00:14:15 Rob Deutsch
So it was interesting 'cause I would think well if I was a data science trying to help Metallica for example to come.
00:14:24 Rob Deutsch
To increase their their profit, I'd have to find a way to get more people at their at their concerts.
00:14:30 Rob Deutsch
But there's there's some nuance there 'cause apparently, and again, please don't quote me on this. I'm are repeating some Mars and facts I heard on a podcast, so please verify independently. Apparently Metallica makes more money from selling T-shirts than they do from.
00:14:44 Rob Deutsch
Actually selling tickets to their.
00:14:45 Rob Deutsch
Their gigs. So Metallica isn't actually in the business of selling music or selling tickets, they're in the business of of selling T shirts.
00:14:54 Rob Deutsch
So that would be a really interesting challenge for a data scientist who's been employed to help them increase their profit.
00:15:00 Rob Deutsch
'cause when you get down to the nuts and bolts of their business, it could very well be that really. What?
00:15:05 Rob Deutsch
You need to do is find more people who come to their concerts.
00:15:09 Rob Deutsch
To buy T-shirts, just increasing ticket sales are going to increase their profit if you get a whole bunch of people to come, who.
00:15:14 Rob Deutsch
Aren't going to buy T shirts?
00:15:16 Dr Genevieve Hayes
I've actually heard a similar thing in the fashion industry. So you know how you have all these fashion houses like Chanel and?
00:15:26 Dr Genevieve Hayes
Uh, Alexander McQueen, etc. And how, you know, each year at Paris Fashion Week on New York Fashion Week, you know, they've got all these latest fashions and the models going down the runway and all that. And so you'd think that Chanel, for example, is in the business of selling clothes.
00:15:46 Dr Genevieve Hayes
But apparently where fashion houses makes the majority of their money is through selling perfume.
00:15:55 Dr Genevieve Hayes
Wow. So Chanel, for example, is not in the business of selling clothes. It's in the business of selling perfume.
00:16:04 Rob Deutsch
That is fascinating. I would not have thought that I've I've done some reading into luxury brands recently and it's a it's a fascinating industry because I think, I believe that one of the other challenges they have is the part of the value they bring to their costs.
00:16:20 Rob Deutsch
Summers is the exclusivity. A Chanel bag, a Ferrari part of the value. Part of the reason pay big money for it is because no one else can have it.
00:16:32 Rob Deutsch
They want to be one of the few people that we can have it. So as a brand if they want to grow their revenue and there's a lot of our publicly.
00:16:40 Rob Deutsch
Listed luxury goods company now. So they're very interested in our growing their revenue to our support there are shareholders.
00:16:48 Rob Deutsch
How do you grow the revenue of a company whose value proposition is that not everyone can buy it?
00:16:56 Rob Deutsch
And I sometimes just stop and think about, yeah, what it would be to be a data scientist working some of these luxury companies.
00:17:02 Dr Genevieve Hayes
Well, I think, I assume a lot of the business models are based on.
00:17:06 Dr Genevieve Hayes
Keeping the numbers of items low so then you can mark them up massively because everyone wants to get, I don't know what is it uh, Birkin bag or something that that type of bag that everyone wants?
00:17:19 Rob Deutsch
I'm the wrong person to confirm whether that's the right, the right?
00:17:22 Rob Deutsch
Brand or not?
00:17:24 Dr Genevieve Hayes
I I think. I think there Hermes bags anyway. Yeah, but I was actually reading an article in the newspaper relatively recently. I was talking about the second hand market for luxury.
00:17:35 Dr Genevieve Hayes
Three goods. So apparently there's websites that are like eBay for.
00:17:41 Dr Genevieve Hayes
Use luxury items and some of the big fashion houses support them, and you can actually resell your bag or scarf or whatever back to that fashion house and then they'll.
00:17:57 Dr Genevieve Hayes
I forward them onto this eBay for luxury goods.
00:18:01 Dr Genevieve Hayes
But some of them are adamantly opposed to having a second hand market for their goods because it reduces the scarcity of them and makes it harder for them, presumably to Jack up the prices.
00:18:15 Rob Deutsch
Which I think is all demonstrative of how if you're a data scientist.
00:18:19 Rob Deutsch
Working for an organisation.
00:18:20 Rob Deutsch
Mission it really pays to be kind of intimately familiar with the other problems that organisation faces. And look, maybe it's not the data scientists themselves, it's going to be intimately familiar with the problems, but you you need a multidisciplinary team then someone who is intimately familiar with the with the nature of the business and can help provide directions.
00:18:41 Rob Deutsch
Otherwise you're gonna end up recommending that a a luxury good company are lower. Their prices to sell a lot more are good.
00:18:48 Rob Deutsch
Let's generate a lot more revenue, but that's not going to work out in practise because I'm no ones gonna end up buying it. Your models are gonna fall flat on their faces.
00:18:58 Dr Genevieve Hayes
So it sounds like, I mean, the way you've put it and the way I see it, there are a lot of people involved in defining any data science project that you're dealing with.
00:19:08 Dr Genevieve Hayes
What do you see as being the role of the business in defining or scoping those projects versus the role of the data scientist?
00:19:18 Rob Deutsch
It's it's a tough question to answer because I think there's a lot of variables that go into that decision and no two projects are the same.
00:19:33 Rob Deutsch
One of or a few. There are a few aspects that make data science more complicated in that context when it comes to our defining and scoping the project, and I think being mindful of these are challenges is the first step to getting a smooth definition and scoping phase. The the first is.
00:19:53 Rob Deutsch
The knowledge gap about what's possible with data size.
00:19:57 Rob Deutsch
So there's this great TMC comic. I'm I'm not sure how familiar you are with XKCD, but it's kind of a web comic.
00:20:05 Rob Deutsch
Energy or you're nodding at me. Ah.
00:20:06 Dr Genevieve Hayes
Yeah, yeah, I know.
00:20:08 Rob Deutsch
There's this great one from like 2014.
00:20:11 Rob Deutsch
So not that long ago.
00:20:14 Rob Deutsch
Where at the single cell and a customer is asking a programmer to build an app, but he's like the app.
00:20:21 Rob Deutsch
Should I ask the user there to take a picture and then identify whether the picture was taken a National Park and the programmers like?
00:20:28 Rob Deutsch
Yep, absolutely. We'll use the GPS. We can do that. Then the customer goes and it should work out whether the picture is of a bird, and the other programmer responds OK, I need six years and a team of five PhD students.
00:20:42 Rob Deutsch
And it's that that differentiator between it's very easy for a, a smartphone app to work out where a picture was taken, and it's that much more difficult to work out if there's a.
00:20:51 Rob Deutsch
Picture of a bird.
00:20:53 Rob Deutsch
Now, these days, funnily enough, it's much more realistic to actually have some AI to tell whether it's a picture of a bird.
00:21:00 Rob Deutsch
Back in 2014, that wasn't the case, but the the lesson I think we can take from that little.
00:21:05 Rob Deutsch
Comic and the.
00:21:06 Rob Deutsch
The truth that was pointing out was that to the layman who knows their business very well.
00:21:12 Rob Deutsch
Data science is a bit of a mystery, and it's not necessarily obvious about what's quick and easy and what's time consuming and expensive.
00:21:20 Rob Deutsch
So I think when it comes to defining a project, scoping a project, the role of the data scientist at least, is to explain in clear terms what is possible, what isn't possible.
00:21:33 Rob Deutsch
I I think also both people in that equation, both the other business which people could be very plural, but you know the business and the data scientist.
00:21:42 Rob Deutsch
I'm collaboration as early as possible and I think both sides are responsible for this. Is very beneficial when it comes to to defining and scoping data science projects.
00:21:53 Rob Deutsch
If any one of those.
00:21:54 Rob Deutsch
Two sides of the fence, so.
00:21:55 Rob Deutsch
To speak, I go too far. Without thinking about the other one, you're likely not to end up with the success.
00:22:01 Rob Deutsch
That you want in the prod.
00:22:04 Rob Deutsch
Lastly, I I think a good benchmark is our how much time is being spent talking about the problem versus the the solution.
00:22:12 Rob Deutsch
And when it comes to data science projects, I'm kind of a fan of matter that Einstein quote which are not necessarily react, not necessarily sure he actually said. Also maybe it's misattributed, but you know if someone gives me an hour.
00:22:23 Rob Deutsch
Solve a problem. I'd spend hours 55 minutes thinking about the problem and five minutes thinking about this.
00:22:30 Rob Deutsch
I think there's some truth in that for data science.
00:22:33 Dr Genevieve Hayes
I think that might have been Abraham Lincoln, actually.
00:22:36 Rob Deutsch
That could totally be the case. Totally be the case.
00:22:40 Dr Genevieve Hayes
But everything has been attributed to Einstein.
00:22:43 Rob Deutsch
Makes it much easier to be, or not to be Albert Einstein.
00:22:48 Dr Genevieve Hayes
Yeah, if we say anything important in this podcast, it will be attributed to Albert Einstein.
00:22:53 Rob Deutsch
You can just replace the other name of the guest on this podcast as Albert Einstein, I suppose.
00:22:58 Dr Genevieve Hayes
Every every podcast I have my guest itself about Elder Lines.
00:23:03 Rob Deutsch
Sounds. Sounds. Sounds.
00:23:04 Rob Deutsch
Didn't come out of curiosity, what's what's your experience with the other role of our business and data scientists when it comes to scoping uh, data project?
00:23:14 Dr Genevieve Hayes
It's interesting. I think you know, like you, I had originally had a background in actuarial and I found when I was doing actuarial roles.
00:23:28 Dr Genevieve Hayes
Businesses were in the habit of talking to the actuaries and had a lot of respect for the actuaries, so it was quite common for me to be brought into meetings where I would be asked what's my opinion of this and the businesses.
00:23:48 Dr Genevieve Hayes
Usually had some understanding of what was and what wasn't possible in the actuarial.
00:23:55 Dr Genevieve Hayes
So that those were very much collaborative projects where everyone had a voice in the problem and solution and we tended to get very good results. One thing I've found with data science is.
00:24:12 Dr Genevieve Hayes
I I think this is because data science is still.
00:24:16 Dr Genevieve Hayes
Early in its life.
00:24:21 Dr Genevieve Hayes
Is businesses don't have that same level of respect for data scientists as they have for actuaries and you often get a lot more of a I guess what you call a command and control type relationship existing between senior management and the data science team where senior management.
00:24:41 Dr Genevieve Hayes
Will often say, you know, we need a solution for this without providing the context and without inviting the data scientists to the meetings. So that.
00:24:52 Dr Genevieve Hayes
And that.
00:24:53 Dr Genevieve Hayes
Is not a good outcome and they often are less willing to have those meetings with the data scientists because.
00:25:01 Dr Genevieve Hayes
They haven't been conditioned that the data scientists can add value.
00:25:06 Dr Genevieve Hayes
As much as with the actuaries, I assume that will change over time.
00:25:12 Rob Deutsch
It's a very interesting comparison. And you know, I suppose in the actuarial sphere, you know, we both operated in that sphere and.
00:25:20 Rob Deutsch
A very notable thing about it is there's a lot of regulation, laws and government oversight about insurance companies, for example.
00:25:27 Rob Deutsch
So everyone is on board with getting actuaries involved and factories are essentially on data scientists are, yeah, but you you're right, when you talk about more generalised data science, there is a little bit less.
00:25:41 Rob Deutsch
Of that understanding and I I see the same thing that you seeing being that command and control like Atmos.
00:25:46 Rob Deutsch
Yeah, but similarly in my experience I've seen are and I'm look. I'm sure there's some overlap with your experience as well.
00:25:51 Rob Deutsch
Sorry, just to clarify, but I'm I've seen tremendous success being achieved when boards when senior leadership, when anyone in the organisation recognises the advantage of including a data scientist.
00:26:06 Rob Deutsch
Early in the conversation and also talking about the context of what.
00:26:11 Rob Deutsch
They're trying to achieve.
00:26:13 Rob Deutsch
Because that becomes incredibly important when identifying the problem and also what might be the best.
00:26:18 Rob Deutsch
Solution for that problem.
00:26:21 Dr Genevieve Hayes
I was reading a book over the weekend and they were talking about the importance of technological literacy by senior management and by boards.
00:26:31 Dr Genevieve Hayes
So we often talk about how it's important that a data scientist has domain knowledge in order to be able to provide a good solution to a problem. But they're saying that.
00:26:41 Dr Genevieve Hayes
A board or a senior manager needs to have.
00:26:46 Dr Genevieve Hayes
At least a high level understanding of what is possible with data science.
00:26:51 Dr Genevieve Hayes
What it means to code a solution?
00:26:56 Dr Genevieve Hayes
What the strengths and weaknesses are of the various technologies that are relevant to the organisation etc, so that they can have those conversations with the data scientists and.
00:27:10 Dr Genevieve Hayes
They can make informed decisions in their roles.
00:27:14 Rob Deutsch
Lot of lot of truth behind that. I I think also if we're talking about the role of the data scientists at the moment like where we are in the.
00:27:24 Rob Deutsch
In the development of data science as a discipline in businesses is that the data scientists more so now than perhaps in the future, needs to be capable of helping to hold senior management hand, so to speak, through these discussions through these?
00:27:42 Rob Deutsch
Issues because I I imagine it's almost a little bit like when I when computers are becoming more prevalent in businesses.
00:27:51 Rob Deutsch
It didn't just happen overnight. I'm sure a lot of business leaders thought for quite a long time that are a big piece of paper Ledger with a pen was the way to our keep track of your accounting.
00:28:02 Rob Deutsch
But with enough technological advancement and the right people to kind of guide them through the change, now it just it's crazy prospector.
00:28:11 Rob Deutsch
I keep track of your accounting, your CRM or anything by people. You do it all online, you do it all on a computer and I think we'll see the same the same progress happen with data science.
00:28:22 Dr Genevieve Hayes
Well, it's interesting 'cause I think it was about a year ago I was having a conversation.
00:28:26 Dr Genevieve Hayes
With a man who worked in the insurance industry during his working life, but he's now retired and he saw that period of time where the organisation was introducing emails and he said there were some senior managers.
00:28:42 Dr Genevieve Hayes
Who didn't want to use computers and they actually had their assistance print out all of their emails so if they could read them.
00:28:53 Dr Genevieve Hayes
So that's sort of like, you know, what you're describing having a data scientist holding the hand of senior manager management.
00:29:00 Dr Genevieve Hayes
They might not be able to print out the emails, but there must be something that they can do to help them make the move into the next century.
00:29:11 Rob Deutsch
That's that would be fascinating on G I'd love to wake up in the morning and get to my Home Office and find a stack of emails printed.
00:29:19 Rob Deutsch
Bra for me to read.
00:29:21 Rob Deutsch
I I do also think it's worth noting though, that when it comes to data science, sometimes data scientists are the right solution to a problem.
00:29:29 Rob Deutsch
Sometimes there is a a low tech at our solution to a problem like I I know when I work with our with our crew shaper a business problem might pop up.
00:29:40 Rob Deutsch
There's something I don't know which I suddenly realise I I I have to know, and being a data scientist, my first instinct is, do I have some data available? Can I throw an algorithm at it to get me an answer?
00:29:53 Rob Deutsch
But it's not uncommon for the answer to my problem to be called a customer. Ask them, call 12 customers and do a mini survey.
00:30:01 Rob Deutsch
Don't even use any statistical rigour, just kind of feel the room, so to speak. And I think it's very much a balancing act 'cause it's very easy just to go too far onto the the gut feel side of things when data science.
00:30:12 Rob Deutsch
Would actually provide a a more thorough, more rigorous, more dependable answer.
00:30:18 Rob Deutsch
But nonetheless, sometimes that simple way.
00:30:20 Rob Deutsch
Is the way to go.
00:30:22 Dr Genevieve Hayes
It's interesting because I I actually heard an alternative point of view on that from someone. I was speaking to her recently and he was saying.
00:30:30 Dr Genevieve Hayes
Uhm, the pandemic has actually accelerated the move into data science because a lot of organisations prior to the pandemic could reasonably make the assumption that the past is a good indicator of the future. So because the senior management had lived through the past.
00:30:50 Dr Genevieve Hayes
They could make judgments based on their own experiences and gut instincts.
00:30:56 Dr Genevieve Hayes
However, following the pandemic, the past is no longer a good indicator of the future and no one has, you know, decades long experience.
00:31:07 Dr Genevieve Hayes
So in order for businesses and senior management to make effective decisions, they actually now need to look at the data because their gut.
00:31:16 Dr Genevieve Hayes
Instinct no longer works.
00:31:19 Rob Deutsch
A lot of logic in that that makes a lot of sense, absolutely. I do think there are some circumstances though where you can hope and wish for like data to exist or data to be as reliable as you would you would like it to be, but it's not always the case.
00:31:34 Rob Deutsch
Uhm, for example, there was a time when I was working for a consulting firm in a data science capacity where we were looking at retail outlets for a large consumer brand and we were trying to identify a strategy going forward for this.
00:31:55 Rob Deutsch
Brands, retail outlets, so we should be consolidated which should be split up. So we involve some geographic data, for example, it involved on financial data or it involves some economic data, some our census data.
00:32:08 Rob Deutsch
And what we discovered in that project was that there just wasn't quite as much insight in the data as we would have liked.
00:32:18 Rob Deutsch
The state of the financials were not really in the right granularity to provide the direction we were after.
00:32:25 Rob Deutsch
The same was said of the census data. The geographic data was perfect accuracy, but just didn't have the info that really provided business insight. So it was just an example of where we did have.
00:32:38 Rob Deutsch
Not resort, but I rely a little bit more on on guarding statement on human feel because as much as we hoped and wished it, the data just didn't provide the answers we were looking for.
00:32:50 Dr Genevieve Hayes
It's very interesting.
00:32:51 Dr Genevieve Hayes
Changing the topic a bit. One of the things that.
00:32:56 Dr Genevieve Hayes
I've found throughout my career, and you've probably found throughout your career, is that when you're working with people from a diverse range of backgrounds, they can often approach data problems from completely different angles than you could ever imagine.
00:33:16 Dr Genevieve Hayes
Uhm, is that something that you found?
00:33:20 Rob Deutsch
Oh, absolutely, I might.
00:33:23 Rob Deutsch
So I think I've had a similar experience to you and I feel very fortunate to be.
00:33:26 Rob Deutsch
Able to have.
00:33:27 Rob Deutsch
A worked with such a variety of people from are from different backgrounds.
00:33:33 Rob Deutsch
Uhm, one of the more interesting was on when I when I worked for I delight in Japan, and it was in a newly formed data science team within financial advisory.
00:33:45 Rob Deutsch
And and I worked a lot with people that came from an investigative background. So I think kind of law enforcement and government intelligence type thing and they actually came from a very non technical background, more like psychology are degrees and the like.
00:34:04 Rob Deutsch
And what was fascinating was to see that when they first looked at data, they kind of looked through the data into what story must be behind the data.
00:34:15 Rob Deutsch
Their mind didn't go to uh.
00:34:19 Rob Deutsch
Means averages, statistics to generalised linear models or AI or XML or anything like that, but like the story behind the data and it was just really fantastic to.
00:34:31 Rob Deutsch
Work with them and learn.
00:34:32 Rob Deutsch
A little bit about the advantages of of that approach.
00:34:36 Rob Deutsch
And similarly, because they came from non technical backgrounds, they were upscaling very quickly in SQL, in ARM Python, in Tableau and visualisation and they asked a lot of.
00:34:51 Rob Deutsch
Elementary question.
00:34:53 Rob Deutsch
And it's caught was kind of interesting.
00:34:55 Rob Deutsch
To see that your gut feel.
00:34:57 Rob Deutsch
To respond to those questions is to give the standard response.
00:35:00 Rob Deutsch
Like, that's almost like why we define them as elementary questions, but when you just stop and take a step back, you realise there's a lot of logic in what they're asking and it's actually an opportunity to.
00:35:14 Rob Deutsch
To reconsider about.
00:35:15 Rob Deutsch
Whether you're doing things in the best way.
00:35:18 Rob Deutsch
So some examples of where that used to pop up is where we use the mixed technologies on certain projects.
00:35:23 Rob Deutsch
Are both R and Python generally not necessarily the best thing to do? But at the time we were relying on some are. Some are packages that I had some statistical methods that we really needed for this analysis.
00:35:37 Rob Deutsch
And we were using Python because we're a little bit more adept at that and for doing the initial data munging, just preparing and cleaning it up for the analysis, and that seems perfectly logical. We built a very odd, dependable, robust star pipeline.
00:35:53 Rob Deutsch
But speaking with a non technical person are highlighted some things we haven't quite thought through as much as we could have and it was a very interesting perspective that they shared.
00:36:03 Dr Genevieve Hayes
I found I my one of my first bosses was a lawyer and I found that she asked some very interesting questions when I presented work to her. So she was a very intelligent woman, she just didn't come from a technical background.
00:36:19 Dr Genevieve Hayes
And what was interesting was there would be certain things that I'd say that she'd clearly not have any interest in, because they were just giving her the technical nuts and bolts of the problem and she didn't need to know that.
00:36:33 Dr Genevieve Hayes
But then there'd be certain things that she'd be asking a lot of questions about, and it really helped me to understand.
00:36:40 Dr Genevieve Hayes
You know, what are the important parts of my work to someone who has to take that work and present it to a stakeholder of some sort?
00:36:50 Rob Deutsch
100% there's a lot of benefits to working with non technical people, non data scientists and you can generate a lot of insight from that and our improve your our data science function considerably. Also the opportunity to work with some are entrepreneurs and people who are starting businesses looking to start.
00:37:10 Rob Deutsch
Business. And it's really interesting to see like the amount of optimism they tend to have about data science.
00:37:19 Rob Deutsch
Uh and optimism I don't always share, but I think there's a lot of value in our being optimistic. Are thinking about what might be possible that your gut tells you might not be possible.
00:37:31 Rob Deutsch
Uhm, yeah, is that something?
00:37:35 Dr Genevieve Hayes
It actually resonates with something that my previous guest is talking about, which?
00:37:40 Dr Genevieve Hayes
Is that a lot of startups are actually more technologically advanced than more developed companies, even though the startups have less data and less money, because they're looking at ways in which they can use data and that's the only way they can get an advantage. Whereas I don't know you.
00:38:01 Dr Genevieve Hayes
Big companies that have existed for 100 plus years, they might think that they can get by without using data.
00:38:08 Rob Deutsch
Absolutely, absolutely. And I think another advantage that.
00:38:13 Rob Deutsch
Uh, perhaps one might say a good start up would have is that they tend to be hyper focused on the problem they're trying to solve.
00:38:22 Rob Deutsch
They're very clear about us are they live and breathe it, they think about it every day and they're committed to finding a solution.
00:38:30 Rob Deutsch
So they will explore avenues that are perhaps less well trodden and discover our processors techniques, opportunities that might might escape you otherwise if you are.
00:38:42 Rob Deutsch
Get a bit too too deep in your own.
00:38:44 Rob Deutsch
Way of thinking.
00:38:46 Rob Deutsch
So yeah.
00:38:47 Dr Genevieve Hayes
Yeah, I've, I've. I've loved any conversations that I've had with startup owners.
00:38:53 Dr Genevieve Hayes
And probably because of those reasons that you've mentioned.
00:38:58 Dr Genevieve Hayes
I think the other thing I like about speaking to people who are in startups is.
00:39:05 Dr Genevieve Hayes
Because their organisations tend to be smaller and focused on one particular product, as you said.
00:39:12 Dr Genevieve Hayes
Uhm, they're not bogged down by all this red tape that you get in massive organisations.
00:39:19 Rob Deutsch
Or sometimes even not so massive organisations, yeah.
00:39:24 Rob Deutsch
Absolutely. A lot of advantages to that. And I I recently read on Lean startup, which is a little bit of a A Bible in the other startup.
00:39:33 Rob Deutsch
Community, and it talks a lot about the advantages of testing your most critical hypotheses as early in the process.
00:39:42 Rob Deutsch
As possible. So you've got a great idea. You're going to use data. You're going to use algorithms to solve a problem.
00:39:49 Rob Deutsch
But you have to make certain assumptions before you get started on the project. You're making assumptions about the data being clean enough about the they're actually being a reasonable answer in the data.
00:40:01 Rob Deutsch
Like, you might find that the the answer you get is just too high variance. You can't have much faith, but really that's our that's the way things are going.
00:40:11 Rob Deutsch
But but if you test all these things early, probably likely to get to an acceptable result quicker, which startups tend to excel at.
00:40:20 Dr Genevieve Hayes
What you are saying about assumptions that actually reminds me of something that happened to me many years ago on one of my.
00:40:26 Dr Genevieve Hayes
Jobs and.
00:40:29 Dr Genevieve Hayes
I was doing work for a particular client of the organisation that I worked for and I met with the managers in that client organisation and I was doing a particular type of analysis that required the use of insurance claims data.
00:40:48 Dr Genevieve Hayes
And they promised me they had tonnes and tonnes of data that I'd be able to use.
00:40:55 Dr Genevieve Hayes
And I said, OK, this is great and let's say it. And they, they took me to a storage room in this office and opened the door and this storage room was filled floor to ceiling with boxes of paper files.
00:41:20 Dr Genevieve Hayes
And I looked at, I said, look, I cannot go through all those files and digitise them and I can't do anything unless I've got this in an electronic form.
00:41:30 Dr Genevieve Hayes
And then this manager looked at this other poor man who was one of his underlings and said Oh no, no, that's OK, that's what this guy is.
00:41:39 Dr Genevieve Hayes
Trying to take.
00:41:40 Rob Deutsch
Oh, wow. Wow.
00:41:42 Rob Deutsch
Which just goes to show that sometimes the value you get from your data science project is worth investing that time in the first place.
00:41:51 Rob Deutsch
You know, in that particular case, I don't know whether it.
00:41:53 Rob Deutsch
Was or wasn't.
00:41:54 Rob Deutsch
But are sometimes. Sometimes it definitely is.
00:41:57 Dr Genevieve Hayes
Well, the irony was that the data they eventually gave us was actually stuff that we already had in our system, and it couldn't do anything more than what we've already been doing. So that poor guy who had to go through all the files, he didn't actually add any value.
00:42:03 Rob Deutsch
Oh no.
00:42:15 Rob Deutsch
Oh, I'm sorry.
00:42:16 Rob Deutsch
To hear that, what a thankless task.
00:42:20 Dr Genevieve Hayes
So I mean, you've probably heard that statistic that, what is it, 85% of all data science projects fail.
00:42:28 Dr Genevieve Hayes
What do you do to maximise the probability of success of the data science project?
00:42:35 Rob Deutsch
Uhm Ji, the first thing that comes to mind is stakeholder. Buy in.
00:42:41 Rob Deutsch
Especially 'cause data science projects can so often involve.
00:42:47 Rob Deutsch
People from multiple disciplines, multiple departments in an organisational context.
00:42:52 Rob Deutsch
If you don't have buy in from everyone, you're likely to hit a roadblock at some point. So what that would mean, for example, is buy in from it to make sure you have the.
00:43:03 Rob Deutsch
Other technology you need.
00:43:06 Rob Deutsch
Buy in from the owner of the data and the business to make sure that actually provides.
00:43:10 Rob Deutsch
You the data you need.
00:43:12 Rob Deutsch
Buy in from the project sponsored like the end beneficiary of the work, you want to make sure that they are committed to our to the value that the other work is going to bring them.
00:43:24 Rob Deutsch
You also need to take a moment just to think about, well, does my business have lawyers? Do I need to get there?
00:43:29 Rob Deutsch
Client is there any other concerns that are what we're doing might have some legal implications?
00:43:37 Rob Deutsch
I'm going beyond buying though. I'm remaining focused on the key result you're trying to achieve, and I think there's some subtlety in that as well. Like to go back to that simplified example we were talking about.
00:43:52 Rob Deutsch
Uhm, where a business wants to increase profit either by increasing revenue or decreasing expenses, you might go down the rabbit hole and I'll say, well, really what what this business wants is more repeat buyers.
00:44:06 Rob Deutsch
Maybe it's an E commerce business for example, and they've got a huge database of people who have bought their.
00:44:11 Rob Deutsch
Products in the past.
00:44:12 Rob Deutsch
And they're pretty certain that the way to get more profit is to get more replies. Get them.
00:44:16 Rob Deutsch
To come back.
00:44:19 Rob Deutsch
Uhm, I find data science projects go really well when you can help.
00:44:25 Rob Deutsch
Oh, sorry. When you can distil a certain metric that you're going for because the the the description of increase repeat buyers, there's many different ways to interpret that. There's very many different ways to measure that and each of them.
00:44:40 Rob Deutsch
Might have their subtle.
00:44:41 Rob Deutsch
Please come for example, if you use to looser the definition, you could include people who bought a product 10 years ago and by this point in time they've already forgotten about your company that's gotten about their product, and they're really much more like a new customer. And so your your data science project is going to get started.
00:45:01 Rob Deutsch
Going to start to get a little bit confused between what activities are generating new customers versus which are bringing back old customers. So you might want to define a particular.
00:45:12 Rob Deutsch
Trick are as to what it means to bring back repeating customers, such as anyone who's bought the product in the past 12 months.
00:45:21 Rob Deutsch
I think it.
00:45:21 Rob Deutsch
Helps to come up with a a benchmark as well about what it's been in the past.
00:45:26 Rob Deutsch
So you might say last year and the year before and the year before that between 5 and 10% of customers came back for repeat purchases. We wanna make that 20% and I think being really clear on that metric.
00:45:39 Rob Deutsch
Versus outcome. How are they linked together and which one you're explicitly targeting is a pretty important are.
00:45:47 Rob Deutsch
Factor of maximising the probability of success.
00:45:51 Dr Genevieve Hayes
And I think this is something that the big consultancies have got nailed with their management consulting businesses and it's something that data science can learn from them.
00:46:02 Rob Deutsch
Exactly, exactly. So at the end of the day.
00:46:05 Rob Deutsch
As data scientists where our technical people, I think we tend to have a lot of passion for what we do and it's very easy to get lost in the joy of building amazing models that we can point out and go look at how amazing this.
00:46:17 Rob Deutsch
Model is it?
00:46:18 Rob Deutsch
Uses the latest and greatest techniques. It's got the most refined pipeline I could come up with.
00:46:26 Rob Deutsch
But that doesn't necessarily resonate with business. Business leaders you need to be able to point to a particular metric and with a cause and effect explanation as to how you have improved that metric.
00:46:39 Dr Genevieve Hayes
One of the things I was thinking about recently that I think might be getting in the way of data science.
00:46:46 Dr Genevieve Hayes
Data science projects success is, you know, if I look at the actuarial and statistics projects that I worked on before going into data.
00:46:56 Dr Genevieve Hayes
Sites. A lot of those were projects that were based on very clearly defined use cases. So for example, one of the jobs I had was as the pricing manager.
00:47:08 Dr Genevieve Hayes
At work safe.
00:47:10 Dr Genevieve Hayes
So obviously the main task of that job was calculating.
00:47:16 Dr Genevieve Hayes
Insurance premium prices.
00:47:19 Dr Genevieve Hayes
So pricing is a very clearly defined actuarial use case.
00:47:24 Dr Genevieve Hayes
And later on I took on another role at work, Safe, which was their statistical case estimate manager, and that centred around a particular predictive model of the future costs of insurance claps.
00:47:40 Dr Genevieve Hayes
So those were very clearly defined use cases, and I'd say you'd have that with a lot of actuarial roles.
00:47:49 Dr Genevieve Hayes
Whereas with data science, we don't have these standard use cases in place yet. So people, a lot of data scientists, they're not being hired to perform a particular function around a use case, they're being hired basically as internal consultants.
00:48:09 Dr Genevieve Hayes
And I'm guessing that eventually those use cases will be defined.
00:48:15 Dr Genevieve Hayes
And but do you have any thoughts around that?
00:48:20 Rob Deutsch
That's a really interesting one.
00:48:25 Rob Deutsch
It reminds me of an experience I had our consulting to come.
00:48:31 Rob Deutsch
A startup company dealing in down fast moving retail goods essentially.
00:48:38 Rob Deutsch
And and they.
00:48:42 Rob Deutsch
Where did I mention they were a startup?
00:48:45 Rob Deutsch
It's not a copy and they were doing relatively well, so quite a lot of data on sales being generated and they realised that there was a need art like you were saying before to do some analytics on the data.
00:48:58 Rob Deutsch
So I'm. I was called in at the early stages with the app with the Managing Director at our parity analytic. This was done through our.
00:49:05 Rob Deutsch
Through them to have a chat about what they were trying to achieve and we got a very long list.
00:49:11 Rob Deutsch
Of what was on.
00:49:12 Rob Deutsch
Their mind a very long list.
00:49:15 Rob Deutsch
All the different questions, all the different avenues.
00:49:17 Rob Deutsch
That could be explored.
00:49:19 Rob Deutsch
And the biggest challenge there from a data analytics data science perspective especially.
00:49:26 Rob Deutsch
Given that, really.
00:49:28 Rob Deutsch
They were startup. This wasn't something where it's going to be like a year consulting project full time or anything like that. So the challenge was to.
00:49:36 Rob Deutsch
To hone in on what the most important question was.
00:49:41 Rob Deutsch
Something that could be solved in a handful of wakes.
00:49:46 Rob Deutsch
And once we had that one question, which in this case was on.
00:49:52 Rob Deutsch
Was about whether their customers were gaming the system, so to speak. There was some implied optionality in the way that they sold their products are being the price of these products are changed rapidly.
00:50:04 Rob Deutsch
Customers could place an order, then cancel the order and reorder at a better price of a better price came up later.
00:50:10 Dr Genevieve Hayes
OK.
00:50:11 Rob Deutsch
So the question that seemed to be the most important end up being the most important was our customers taking advantage of this.
00:50:19 Rob Deutsch
Honing in on that being the most important question and coming up with a metric to measure that was by far the most important part of the project.
00:50:29 Rob Deutsch
Linking back to what you're saying, because then there was a definite on.
00:50:33 Rob Deutsch
Use case for the data science. Something that everyone data scientist and management customer could point to as the problem being worked on and everyone was clear on the value that was bringing.
00:50:47 Dr Genevieve Hayes
One cat way I've seen use cases coming about with data science is with the rapid explosion of power BI in organisations.
00:50:58 Dr Genevieve Hayes
So I mean I think every organisation is now getting power BI or one of its equivalence.
00:51:05 Dr Genevieve Hayes
And people are now understanding, Oh yeah, data visualisation dashboards are something that we need and people are just being hired as power BI developers.
00:51:17 Dr Genevieve Hayes
So I guess the way I'm looking at it is you're looking at it from a consultant point of view, but I was sort of thinking of it from our permanent employees.
00:51:25 Dr Genevieve Hayes
Point of view, and I suspect it's to do with you.
00:51:29 Dr Genevieve Hayes
The technology is defining the role in that.
00:51:32 Rob Deutsch
That's a really interesting way of looking at it. And I think there's a lot of overlap between these kind of two different perspectives, 'cause I'd be tempted to say that that permanent employee in the organisation in one way or another is still kind of a consultant, especially if they're being hired to kind of build some data dashboards, because we need data dashboards, we have power BI.
00:51:52 Rob Deutsch
And we have data. The role of that employee especially becomes to find out what problem needs to be solved 'cause that hasn't necessarily been communicated clear.
00:52:04 Rob Deutsch
And and I think that over the next few years like as if we're not already in it, but there's going to be increasing focus on not just having the right tools in a business, the right technology, but the right people and the right skills 'cause it's very easy for anyone to sit down and learn power BI.
00:52:21 Rob Deutsch
It needs to.
00:52:22 Rob Deutsch
Be a little bit more complicated to understand.
00:52:24 Rob Deutsch
More about like a business.
00:52:25 Rob Deutsch
Way of thinking or a.
00:52:26 Rob Deutsch
Problem solving, way of thinking and how to use power BI.
00:52:31 Dr Genevieve Hayes
Well now I've been told that what organisations are looking at is not so much whether someone understands a particular technique or a particular piece of technology, it's how well can they communicate technical concepts to non technical audiences and how well can they.
00:52:52 Dr Genevieve Hayes
Extract information from those non technical audiences in order to develop use cases. So I guess storytelling and curiosity.
00:53:03 Rob Deutsch
Which is another component of that, our probability of success for a data science project as well I think is that communication 'cause I'm.
00:53:11 Rob Deutsch
I I like to have a little bit of a pessimistic view, being that like all projects goes different to what you expect.
00:53:17 Rob Deutsch
You can have all the best intentions when you are, when you start off, but the realistic thing is there's gonna be our speed bumps.
00:53:23 Rob Deutsch
There's gonna be our left turns, right turns or unexpected discoveries and and I think data science just because it tends to be so technical.
00:53:33 Rob Deutsch
That can be the death knell for some data science projects if you can't clearly communicate what's going on, why it's going on, what the options are and always linking it back to that, a question of, well, what are we trying to achieve? What's the objective of this project?
00:53:48 Dr Genevieve Hayes
Yeah. And that's exactly what we said from the right, from the beginning, you need to tie everything back to business problem.
00:53:55 Rob Deutsch
Exactly, exactly. And I'm a strong believer that it's not just.
00:53:59 Rob Deutsch
The the project leaders or business management that needs to know this, but it's really good for everyone throughout the project down to the most junior of analysts or the intern to be clear on what we're trying to achieve.
00:54:10 Rob Deutsch
We've come. It saves heartache and it improves happiness, I would say as well if people see the impact they're planning in an organisation.
00:54:18 Dr Genevieve Hayes
Well, I was a manager of a data science team for a while. Does. That's an actuarial team. And one of the things I always insisted on was anyone who was connected to a particular project always got to attend any meetings about that project, even if they were just the graduate or something.
00:54:38 Dr Genevieve Hayes
Because it's the only way they will never understand what's going on. I don't like those situations where it's just the manager who attends and then the manager passes the information onto the people under him or her.
00:54:51 Rob Deutsch
Which reminds us just how critical a part of data science is becoming in an organisation as well.
00:54:58 Rob Deutsch
Being that you know, I would say that communication within an organisation is one of the most important are predictors of success and that just translates perfectly over to the data science function, the data science team to the data science consultants come communication is is key and you always want to avoid as much as possible second.
00:55:18 Rob Deutsch
And information or third hand information or or fourth hand information on GI think that's excellent. I'm yeah glad to hear you are doing that.
00:55:28 Dr Genevieve Hayes
OK. So we're getting close to time now. So I'm just going to go through my last couple of questions.
00:55:34 Dr Genevieve Hayes
Is there anything on your radar, in the AI, data and analytics space that you think is going to become important in the next, say, three to five years?
00:55:44 Rob Deutsch
The first thing that comes to mind, and I think by far the most important, is the the qualifications and skills of individuals in the industry.
00:55:53 Rob Deutsch
Unlike what we were talking about before, it's very easy to hire someone who who knows Power BI think it's harder to hire someone who has the business mind, the problem solving mind, to actually use that effectively.
00:56:07 Rob Deutsch
I was recently for Roku shape are looking for a software developer and dumb as the nature of these things go in the modern world where it's listed on a website that makes it very easy for people to apply.
00:56:19 Rob Deutsch
I ended up getting a few applications from people who were looking for data science roles, not software developer roles, but consequently I got to kind of.
00:56:28 Rob Deutsch
See a little bit about what the what the state of data science resume is looking like at the moment.
00:56:34 Rob Deutsch
And what I saw was a lot of people who had very quickly retrained from a non technical career into data science, into data analytics and looking at their resume and kind of having the pretty solid data science background that I have, I was a bit nervous about some people abilities to our.
00:56:53 Rob Deutsch
Provide the value the organised that organisations would expect from data analysts.
00:57:00 Rob Deutsch
And I think that given where we are as a business community on our data science and data analytics maturity cycle, which is still relatively low, it's hard for businesses to identify.
00:57:12 Rob Deutsch
Who's a good data scientist? Who's a bad data scientist? And I think we're going to see increasing focus on that in the next three to five years.
00:57:21 Rob Deutsch
Exactly how that might look I don't exactly know, but things like our qualifications are being offered by external institutes.
00:57:31 Rob Deutsch
Perhaps UM.
00:57:34 Rob Deutsch
Greater understanding.
00:57:37 Rob Deutsch
Are being generated by hiring managers to know what they're looking for in data science, CVS, data science consultants, data science projects, and the like.
00:57:48 Rob Deutsch
But yeah, I think the the qualifications of the people who are working in the data science field, we'll see a lot more focus on that in the in the next three to five years.
00:57:57 Dr Genevieve Hayes
And are you saying you reckon will have more technical assessments being involved in data science recruitment?
00:58:04 Rob Deutsch
Not necessarily assessments.
00:58:08 Rob Deutsch
But some proof points, so perhaps it will be assessments, perhaps it will be our qualifications.
00:58:16 Rob Deutsch
Uhm, there's numerous organisations in Australia that will provide.
00:58:19 Rob Deutsch
Some form of of.
00:58:23 Rob Deutsch
Degree, of course, when it comes to data science.
00:58:29 Rob Deutsch
And I think a lot more businesses are just going to be asking the question of how do we know we're hiring the right people?
00:58:34 Rob Deutsch
And I guess beyond that, how do we make sure that we're keeping them up to date with the latest tools and techniques?
00:58:39 Dr Genevieve Hayes
Yeah. And so that's and I think.
00:58:43 Dr Genevieve Hayes
All people who are applying for data science roles want to be kept up to date with the latest tools and techniques. I mean, that's something that people look for when they.
00:58:53 Dr Genevieve Hayes
They're seeking a data science role. They want their boss to be investing in their career growth. So I think if you find a good data scientist they will probably be on board with the employee, any endeavours on the part of the employer to help them maintain and build their skill set.
00:59:14 Rob Deutsch
Exactly. And I think that as a business, you want the person as well who's going to be interested in continuing to learn and develop themselves because that's just naturally going to be a far better, better data scientist.
00:59:27 Dr Genevieve Hayes
Yeah, yeah. It's nothing more frustrating than someone who doesn't want to learn new skills in a statistical or technological area, because as their existing skills become stale, it's hard to assign them to projects.
00:59:44 Rob Deutsch
Very true. And I I think another component of that is as well as the the business context like I'm I'm sure what we're going to see more of to link back to the original question you asked about are.
00:59:54 Rob Deutsch
Three to five year time Horizon we're going to see an increasing focus on responsible use of data and gathering of data. So we're recording this a few weeks after a major Australia, our telecommunications provider, had a data leak.
01:00:09 Rob Deutsch
And I think it's not just going to be our governments or regulators focusing on this, but there's an increasing public awareness of data issues.
01:00:20 Rob Deutsch
Even if I think like five years ago from now, there's a lot of controversy around Facebook and their handling and use of data.
01:00:27 Rob Deutsch
And the public became very aware of it. And I started to notice like of my friends, even non technical people, people who don't think about data at all started to ask questions about what algorithms or Apple using, Facebook using on.
01:00:40 Rob Deutsch
On their data and now with this recent our data release data are.
01:00:47 Dr Genevieve Hayes
Data breach.
01:00:48 Rob Deutsch
Data breach. That's the word I'm looking for. With this recent data breach, I think we're going to have a lot more questions from the public about what's going on here.
01:00:55 Rob Deutsch
What are you doing to safeguard my data? Why do you still have my driver's licence five years after I was a customer, those types of things.
01:01:02 Dr Genevieve Hayes
And do these companies need to collect them in the first place?
01:01:06 Rob Deutsch
Well, that's a very good question.
01:01:10 Dr Genevieve Hayes
Yes, I'm, I'm one of the people who had to get a new drivers licence because of that breach. So I I have a lot of questions for that company.
01:01:18 Rob Deutsch
You and me both, actually.
01:01:21 Rob Deutsch
And I'm interested to see as well like what senior management reactions is going to be to this 'cause I I have to deal with some of these problems in my role at Aku Shaper.
01:01:31 Rob Deutsch
We're a relatively small company, but I we still saw some amount of data and we want to protect our clients confidentiality and be in line with what their expectations are.
01:01:42 Rob Deutsch
Yeah. But like bigger businesses where it's a little bit less are well defined. When you're a small business, you tend to have like one, one data set of customers.
01:01:52 Rob Deutsch
It's very easy for a single person to think about it in 5 minutes and kind of ascertain where you are. The bigger your organisation gets, I think the more difficulty.
01:02:02 Rob Deutsch
You have with.
01:02:03 Rob Deutsch
Keeping track of every way you happen.
01:02:05 Rob Deutsch
To have some data.
01:02:06 Dr Genevieve Hayes
Umm, I would agree with that.
01:02:09 Dr Genevieve Hayes
So what final advice would you give to organisations looking to maximise the value of their data?
01:02:17 Rob Deutsch
I would start with focusing on the problem 1st, and then the data second. Be clear on the problem you're trying to solve, the objective you're trying to get out of that.
01:02:28 Rob Deutsch
Then look at the data and work out how you can marry the two. And to that end, start talking with a data scientist early, even even before you started talking about data.
01:02:39 Rob Deutsch
Even when you're just talking about the problem 'cause a good data scientist will know how to you know, just stick talking about the problem before they start talking about the other data.
01:02:49 Rob Deutsch
And they can help you to our.
01:02:51 Rob Deutsch
To find the right place.
01:02:54 Dr Genevieve Hayes
Or even just talking, having a nice casual chat to the data scientist about their weekend and seeing where that goes.
01:03:01 Rob Deutsch
100 percent, 100%. Anything with I can get a coffee.
01:03:06 Dr Genevieve Hayes
So this has been really great. UM.
01:03:10 Dr Genevieve Hayes
For listeners who want to learn more about you or get in contact, what can they do?
01:03:16 Rob Deutsch
LinkedIn is probably the the best way to contact me. I'm I'm sure you'll throw a a link into the other show notes perhaps the people can click on I feel free to shoot me a message if you want to design a surfboard.
01:03:29 Rob Deutsch
I'm the man to speak to if you have any questions about data science.
01:03:33 Rob Deutsch
Truth them through.
01:03:34 Rob Deutsch
If there's a problem you're trying to tackle.
01:03:35 Rob Deutsch
Let me know.
01:03:37 Rob Deutsch
Yeah, LinkedIn is the way to go.
01:03:38 Rob Deutsch
Right.
01:03:40 Dr Genevieve Hayes
And I'll put a link to your LinkedIn page in the show notes.
01:03:43 Dr Genevieve Hayes
Fantastic. So thanks for joining me today, Rob.
01:03:48 Rob Deutsch
Thank you so much Genevieve. I have thoroughly enjoyed it. I really appreciate it. Thank you for having me on.
01:03:54 Dr Genevieve Hayes
And fantastic.
01:03:56 Dr Genevieve Hayes
And for those in the audience, thank you for listening. I'm doctor Genevieve Hayes. And this has been value driven data science brought to you by Genevieve Hayes Consulting.
Creators and Guests
