Multichannel Success Podcast series
Episode 6 - Data/Analytics and Insights - Transcript
#### David Worby
Hello, and welcome to our podcast series on how to drive multichannel success.
In this series, we're looking to explore a range of issues which you and your
business are faced with every day. We really hope you find the discussion
useful, and we look forward to your feedback.
In today's episode, we're going to explore one of the key drivers of success in today's multichannel world,
the subject of data, insights, and analysis. I'm David Worby from Prospero, and I'm delighted that joining me today is David Kohn from The Multichannel Expert.
#### David Kohn
Hello
#### David Worby
And Mark Pinkerton.
Let's start this subject by just trying to get our heads around what we mean when we talk about data analytics and insight. What is the universe for data analytics and insights? Let's start with a rather weird question. Mark, what's it used for?
#### Mark Pinkerton
Data and analytics and insight well they produce insight but data and analytics are really there within the e-commerce digital world to understand performance online so that we can maximize the potential on a day-to-day basis. It's there to help people make better decisions and we use data analytics to understand
customers what are they actually doing and therefore how can we derive more value from them. It also provides business intelligence to the rest of the business. communication or sideways communication. But it also covers things like market intelligence, so qualitative information about what customers may or may not
wish to do. And for me, I also think of data as being the bloodstream of intelligence systems, so if you have an automated solution that does anything in e-commerce, it's going to need data from somewhere.
#### David Worby
Okay well David I'm going to come to you around the question about what is
data less from a kind of philosophical perspective but more about what types
of data sets do e-commerce businesses tend to use and where do they come from?
#### David Kohn
As Mark said, one of the key things that you're looking for is to understand
how your website is performing. So, there's a whole heap of what I call
statistical measures which will look at your website. For example, page speed,
number of pages visited, speed with which images are being rendered, etc.
That's really important for understanding your website.
There's also a whole range of behavioral data that you're going to pick up
from the way in which your customers are acting on your website. So... listing
page to the product page, whether they added to basket, whether they
particularly completed the purchase. These are all really important bits of
data. So these are more statistical things but there's no reason you should restrict yourself just to the
the statistical or the hard data. There's plenty of soft data that can help you
inform the decisions that you make. that the statistical data leads you in a
certain direction, it tells you what's happening but it doesn't necessarily
tell you why it's happening and that's where you can enlist qualitative data.
So in a sense data is any piece of information that you can collect in whichever way you collect it but that
helps you take the decision, helps you understand what's going on in order
that you can make the best decision and take the best action.
#### David Worby
Mark, some of us, particularly me, probably grew up in an era where we didn't
really consciously think about data. I know you spent some of your career at
an agency who made huge strides in analysing data and being able to provide a
service to retailers. So tell me a bit about why data is important.
#### Mark Pinkerton
I, well yeah, I mean I spent 11 years collectively working for Logan Tod and
Co, which was one of the very first sort of analytics-based consultancies in
the UK. And really I would say that analytics is important because it takes
you away from the HIPPO-based decisions. HIPPO is highest paid person's
opinion, i.e. what the top dog says. It takes you away from that and it takes
you into the more scientifically rigorous world
of having hypotheses, being able to test those hypotheses and then you try to
determine whether or not your actions are successful using data. So it's
important because you get away from a whole bunch of contrasting opinions that
people will have of things. underpins decision-making.
#### David Worby
Okay, all right. So I think we've articulated what it kind of is and where it
comes from, what we use it for, and maybe a little bit about why it's
important. But David, I want to come to you just to understand the arenas in
which e-commerce businesses tend to gather and utilize data. What is our kind
of core understanding that we need to get to grips with using data?
#### David Kohn
The way that I've always tended to look at key data for e-commerce is to try
and think of the sales funnel. So start with traffic, go into the website, and
then look at the profit you make. That leads to three principal areas. The
first is traffic. What are you getting from your existing customers? What do
you understand about their behaviours? Are they repeat? Are they coming back
enough? Are they spending enough? Have they left you? Are they about to leave
you, etc. And then the aspect of that is new traffic, so customer acquisition
is absolutely critical. customers at the cost efficiency and with the volume
that you want. The second thing is website... you can utilise in order to understand how your customers
are behaving, why are they exiting the site, why are they converting, what are
they looking at, what are they like, etc, etc, etc. And then the third area is
what you might call efficiency and profitability. How well are you converting
the sales that you're making into profit? You might be looking at margin, you
might be looking at availability. You should definitely be looking at supply
chain and delivery costs. But those for me are the three big areas, customers,
conversion and efficiency.
#### David Worby
Super. Mark, did you want to add anything to that list of core understandings?
#### Mark Pinkerton
The thing around customers is important. We really need to look at the
customer data.
We're not just talking web analytics data here, we're talking customer data.
So the lifetime value of the customer, how often do they shop with us, how
many things do they buy, so what can we expect from them. I think that tends
to get ignored because that's often dealt with by the marketing person as
opposed to the econ person. But actually you've got to look at those two
holistically because the two things tie together.
I completely concur in terms of the conversion rate and the optimisation side
of things. There's a difference here between the data and exposing the data so
that people can see what is happening and then the analysis layer in terms of actually saying, OK, I've got two
things happening.
Is A causing B? Is B causing A? Or is something completely different causing A
and B? So that analysis piece is potentially hugely powerful.
#### David Worby
OK, so we talked a little offline about the danger that maybe some businesses
have fallen into of trying to track everything. And the danger there is that
you're now grabbing data and you're owning data and you're manipulating data
that has very little end use. So I'd like to now come on to the concept of how
do you avoid creating work that has no net outcome. And David, we talked a bit
before about having a perspective on the business from a very high point from
which you can then drill down into the relevant bits of detail. Do you want to
talk about that a bit?
#### David Kohn
Yeah, I'm a great fan of the helicopter and the scalpel principle. The
helicopter is the dashboard. It's the top level information that you're going
to use on an ongoing basis to identify whether or not you're in line with your
KPIs or whether or not you're behind your KPIs. What that's important for is
helping you to focus on the areas where you need to dig into it in more detail
and get the scalpel out. Whatever level you're in the organization, you need
to be running these sorts of dashboards. help you focus, because the one thing which is a limited resource
for everybody in e-commerce is time. You simply cannot afford to look at
everything, so you've got to decide what it is you look at in more detail.
When you look at things in more detail, the next level, the scalpel, that's
when you might be digging in order to identify the issue that you're faced
with.
In some cases, the data that you'll be able to dig out will be statistical. In
other cases, you may feel that that doesn't give you enough information,
doesn't give you enough insight, and you may need to get more qualitative
data. Ultimately, judgment is going to be required in all of these cases, but
you need the data to be leading you in order that you can identify what you
focus on.
#### David Worby
So the way you've described that sounds like human capital is a vital
component in the whole data story. You've got to have the data to start with
but actually you need people to kind of do the work to identify what insights
are coming from that data.
#### David Kohn
For sure, there's advantages to be had from automation. There are loads and
loads of tools that can identify anomalies and that can point you in the
direction you want. And you must have a range of automated tracking tools and
regular reports that you consult on a regular basis. But in my experience,
more often than not, when you get into digging down into something, you're
gonna want the data to be available. And in some cases, you may find actually
you need to start recording data in a different way. But for those more in-
depth analyses, you're generally gonna need to be quite creative. You are
gonna need that human capital that you described in order to get yourself to
the point where you properly understand it and where you've got enough data
and enough insight in order to more confidently make decision.
#### David Worby
So I've got my resources and I've got my data I've maybe not necessarily got the data at this particular point in necessarily the right order but Mark how do I go about creating hypotheses from that how do I how do
I understand that I should be running off down this avenue because here I'm
suboptimal as opposed to this avenue here where I'm arguably less problematic
how do I make it how do I make a decision on where to go first, given that I can't go everywhere at once
#### Mark Pinkerton
No, you're always going to be time-limited, that's true. In terms of which
hypotheses to follow or how to create a hypothesis, which are effectively your
two questions, the how to create the hypothesis is essentially
have some knowledge and understanding of the field
of which you're trying to create hypotheses around. And then to come up with
ideas about what may be causing the issue that you've identified, or it may be
an opportunity that you want to try and understand whether or not it is a real
opportunity or not. So for me, those are the two main scenarios. On a day-to-
day basis, the problem resolution tends to dominate within most Ecom teams.
But actually, in terms of developing new functionality or developing new
customer segments or whatever, the opportunity hypothesis is also potentially
very powerful. And essentially, a hypothesis is a set of ideas that you are
then trying to validate using data or to disprove using data. It may be, I'm
worried that this channel is cannibalizing this other channel. Okay, that may
be true. Let's go and see if the data supports that or not. But I also want to
come back to one of the other points around the sort of helicopter and scalpel
analogy that David was using, which is that in order to get the right data,
particularly with systems as they are now with Adobe and with GA4, the way
that GA now configured it, you have to have a pretty clear idea of what you
want to start measuring when you set the analytics up. There's two completely
different skill sets within analytics.
One is having an implementation expert who understands the best way to
actually extract the data that you want, which needs a business person to be involved to say, this is the data that I want. And then you've got the analysis that an analyst will do, which will say, okay, based on this, I can tell you this hypothesis works, or I can automate all these reports for you or whatever other facilitation you
want, knowledge transfer. And it's not one skill set. So it's really hard to
have one person that can oversee everything you need to do within the
analytics world.
#### David Worby
OK, well we'll come to your model in a minute.
#### David Kohn
One of the things I was going to add was a critical
thing is magnitude is how big is this issue or opportunity. It's probably the
first question I ask when somebody raises something. The thing that you always
got to try and identify as quickly as possible is how many people, how many
users does this influence? How many interactions does it influence? You may
think something is a massive issue or it may be something you're doing
terribly but if it affects 0.01% of your user base it's not important. On the
other side of the coin where you only make a small improvement but it's
affecting almost everything. So I think one thing I would really advise always
first thing you find out is how many people, how many interactions, how many
transactions, whatever the metric is, does it affect?
#### David Worby
Yeah, I like the likelihood and probability and impact score.
#### Mark Pinkerton
It's a criticality basis, isn't it?
#### David Worby
And just one other thing I would add in here,
we often hear this from clients we talk to, they get quite obsessed at times
about benchmarking their performance against their peers. And therefore, when
it comes to the funnel that you talked about, David, people have a very clear
image of how their funnel is performing and actually feel sometimes comforted,
but also sometimes challenged by understanding how that compares with the
market and the sector. And they'll derive deficiencies or successes from a
comparison with a benchmark, which is sometimes difficult for clients to do
because they don't necessarily know how their competitor set are performing.
But when we tip up, they often ask us to make a comparison between how
effective and efficient their funnel is versus their cohort group.
#### Mark Pinkerton
But I would challenge that to some extent to say, are you a unique business or
not? And if you're a unique business, then actually having competitor
comparisons is pointless because nobody's going to have the same funnel as
you. But the other thing is, actually, if you're looking to optimise your
business, you have a completely diverse set of data within your own sphere. So
you have good weeks and bad weeks, and you can actually analyse the difference
between your good weeks and your bad weeks to say, OK, these are my best
weeks, why are they my best weeks, what's so great about them, and how can I
replicate it? So there are always ways that you can draw comparisons, even if
they're internal ones.
#### David Worby
Let's move on to how, moving more towards the practical side now of things
people could consider. David, we talked offline a little bit about beginning from the sales funnel and
moving out from that and building out a plan from the customer sales funnel. Do you want to talk a little bit about that?
#### David Kohn
Yes, for me there are three principal dimensions to the data that you really
do require on an ongoing basis. So let's take customers as an example.
So retention is a classic metric for any business for whom customer retention is an issue, and it is for most,
should have some measure of retention. Whether it's customers who have shopped with you in the last 12 months or previously shopped with you, whether it's the frequency with which they spend, whether it's the amount they spend. If we're talking about traffic, again there are a series of standard metrics which you really should have. Obviously you've got the volume, if you've got a target it's the extent to which you're performing well against that target. You've got the cost efficiency, how much are you spending through each channel. So for each of the three areas that I touched on, you can build a simple dashboard at whatever level you are in the organisation that helps you keep track of your key metrics and helps you focus on the things that you
should be spending more of your time on.
#### David Worby
And a dashboard as a way of managing action or as a way of communicating to
others? What's your preference or both?
#### David Kohn
Look, I think in any business there are numerous levels of people. At the top
level, you've got a certain number of statistics which need to be seen, need
to be communicated will help your CEO, for example, or your finance director understand at a
high level how the e-commerce business is going. But if you're the PPC manager, you're going to have a different dashboard.
The point is, it's something that is monitoring pretty well the whole scope of
your activity but it's doing it at a level that it doesn't overwhelm you and
it's doing it in such a way that it points you to the things where you need to
be. right now.
#### Mark Pinkerton
Yeah and typically we've built dashboards for clients in the past where
they've been very heavily biased towards making sure that senior management
understand the performance of the digital commerce world and that's built on
the KPIs that are associated with that and communicating those to senior
management. Often you find at that level that there isn't an enormous amount
of variability from week to week. The customer retention doesn't change that
much from month to month let's face it because it's a long scale, long term
PPI. But on a day to day basis PPC performance can be wildly radical so you
end up with horses for courses.
#### David Kohn
And one thing again, a thought that Mark has just sparked with me is,
dashboards are not permanent. You know, there will be periods in your trading
history where some things become more important, some things it becomes more
important to bring them to the forefront.
#### Mark Pinkerton
You might have a supply chain issue, so availability becomes critical, so you
put that in.
#### David Kohn
Exactly, and I think never become complacent about your dashboard because there's always a danger that something new or different may have happened.
#### David Worby
Okay before we come on to the next section which is really a little bit about
the action you should be considering taking some of the challenges that we've
seen with people beginning to take action in this way. Mark I know you just
want to talk a little bit about the subject of GA4 maybe at this point we
might lose a load of our podcast listeners with a rather dry subject.
#### Mark Pinkerton
I'm trying to genericise it a little bit. You need to think quite carefully
about the analytics platform that you are using to be the source of your data
and you need to make sure that that analytics platform is set up correctly to
extract the most insight and most analytical capability that you are paying
for. So, for example, if you go for a new e-com platform, don't assume that
your SI has actually set up the analytics correctly or if your performance
marketing team are paying for the Google Analytics, don't assume that they will set it up for e-commerce because nine times out of ten they don't.
There's a big sea change going on at the moment whereby Google
Analytics, its core technology base is from the old Universal Analytics to GA4
and GA4 works on a completely different basis from the old Universal Analytics
so you can't compare the two. You certainly can't compare the data between the
two either. However, at a fundamental level there are massive problems with
GA4 so part of the analytical community, I'm seeing a lot of discontent with
GA4 and I think it's important that non-technical audience is made aware of that. So,
there's a very good quote from well-known analyst presenter called Jono
Alderson and he's written that GA4 is
absolutely non-designed garbage, it's fundamentally unfit for purpose for either normal or power users.
I won't go through this. more detrimental than those two sentences but you know there's a
big problem with GA4 right now that Google is not addressing
#### David Kohn
And I think with that in mind, Mark, my sense is if you're using third party
tools, for example, on-site search, your on-site search provider should be
able to provide you with a load of information through their portal, which is
going to be incredibly useful for you in not only helping optimize the search
tool, but also in helping you understand the way your customers are shopping.
I really think if GA4 is as bad as everybody says it is, it means you should
put more pressure on those third party tools to supply you with great
measurement, with great tools, and with great insight, because they are
sitting on a load of your data.
#### Mark Pinkerton
But those are isolated pockets of data in their own little silos and much of
the benefit of analytics is actually by being able to put all of these things
together and work out correlations of things. So people who search are more
likely to convert than people who don't search, for example, is a reasonably
well-known truism in econ. But if you don't have search tracking enabled in GA
and you've only got it coming through your search and merge tool, then you're
going to really struggle to tie the data sets together because they're
operating on a completely different basis. So you need a universal basis of
truth to underpin things and if you're having problems setting up that
universal basis of truth then it's a significant problem. And the last point
on that is GA and with Adobe, you actually have to specify what you want to
record and work it out in advance of actually setting the tool up.
#### David Worby
So David was suggesting that we find a way to use other data sources to tell
us that. I think that's sensible. We'll come back to that in a minute. But if
GA4 and Adobe are as bad as you're making them out to be, what should our
listeners do?
#### Mark Pinkerton
I'm not saying Adobe is bad, I'm saying it operates, and GA4 is now operating
in a similar way but it's not, it's an immature, it's not ready, it's not a
market ready product in its current state, but Adobe is. They have different,
you know, you have to think about things a lot more and set them up in
advance, whereas the old Google Analytics that everybody was, you know, they
were just throwing it on every website and it would by and large record 95% of
what you wanted and then you could tweak it. about things in advance and then
implement it and then you'll start to get some analytics out of it.
#### David Worby
Does that mitigate against agility and the ability to change the plan? Okay,
so it's not fundamentally flawed, it will do what you want, but I think I'm
taking from what you're saying that you've got to think about it a little bit
more, you've got to plan a bit further ahead and you've got to set it up with
a plan in mind. Yes, I don't work in that way. It's less spontaneous.
#### Mark Pinkerton
I don't think it would work in that way, and it's an expensive tool and has
expensive resources to enable it to work. Google Analytics was very widely
understood and had a big audience and lots of people working on it, and we're
now getting to a set where it's become a more, they're deliberately making it
a more specialist tool to support Google's advertising ambitions I suspect.
#### David Worby
Okay so let's just move back to the point that David neatly segued us into
which is that there are probably when you begin to work out what you want to
monitor maybe some of the gaps in your own performance and deficiencies you
might stumble across the fact that you've got gaps in your data a lot of you
probably got a lot of third-party tools running on your site and whilst you
are clearly the data owner for elements of that and we're not going to go
there because that's a little bit too legal for today the reality is you may
have gaps in your data because that data sits sometimes exclusively with that
third party so what advice would we give our listeners David when it comes to
how to navigate around that particular challenge?
#### David Kohn
I think for the larger retailers who are listening to this, then they'll be
thinking very seriously about creating data lakes. They'll be thinking very
seriously about their own data structure, what they can collect, how they can
index it, how they can stitch it together in such a way that it enables the
sort of queries that are gonna drive insight. For the smaller retailers,
realistically and I think you've got to be very pragmatic about things. You've
got to ask yourself constantly when you're trying to make a decision, when
you're trying to understand something is, have I got enough data to
confidently make this decision and take the action or do I need something
else? It may often be qualitative data that is going to give you the greater level of confidence.
The one thing I would say is critical, it's actually however much data you've
got, however much analysis, however much insight, you never know categorically
what is going to happen until you do it. So whether you're large or small, if
you are going to take definitive action across your whole site or if you're
going to take a test and trial approach, you must know what you're going to
measure and you must measure it and you must look at that measurement at a
certain time period after you've done it. That for me is probably the most
important thing I can suggest about this this whole podcast is, as I say, you
never know what's going to happen until you do it. can do things and you can
see exactly what happens as long as you've set your measurements up correctly.
#### David Worby
And I think on previous episodes you've wisely advised people not to bet the
shop on anything and I guess that's partly true here, you've got the ability
to segment and try something that isn't going to damage the whole
infrastructure if it doesn't work.
#### David Kohn
Look, part of being in the real world, part of being in a business, is
sometimes you have to take definitive action. Let's say it's a weekend
promotion when trading isn't going quite as well as it should be.
If it's something where the degree of confidence is, let's say, 50% or 60%,
then absolutely do not bet the shop on it. Absolutely set up a way of testing
it. Maybe it's an A, B test, maybe it's an A, B, C test. The key thing is
whatever you do that is significant, make sure you're measuring it, you know
what you're measuring, and you don't forget about it.
#### David Worby
I think also in that world where you have a
test and learn approach, but you're using a third party tool, we talked
previously about making sure that the contractual agreement you have with that
party allows you access to the information you want in order to help your
decision making process. So whilst we're not sitting here advising on a legal
footing, from an operational perspective, if you're going to want to know how
elements of that functionality are performing, make sure it's in the contract.
So you have the right to either ask for the insights from it or as you
mentioned in enterprise environments, maybe even taking the data from it.
Okay, penultimate subject here is GDPR. Mark, I know that the challenges that
we're facing or our businesses are facing on the go forward are probably quite
different from those that we faced in the past, but specifically in a future
looking cookie-less world, what challenges do you think are being faced now by
retailers and brands?
#### Mark Pinkerton
Well, for retailers cookies have been critically important, in that cookies have been critically
important for customer acquisition, i.e. performance marketing, because it has
enabled their agencies to have an understanding of where their audiences are
across multiple different sites and to manipulate some of that audience to
come to your website, particularly through things like Facebook, but that is
no longer going to be allowed and cookies are blocked by most of the browsers
now, so you are back to, I guess, the analogy of newspaper advertising more in
terms of how you attract people to your website, and that is a fundamental
difference in thinking for customer acquisition, which means that
customer retention probably becomes slightly more important because you can't
just drive more and more new traffic and or the fact that the quality of the
incoming traffic is probably going to be lower quality than it historically
has been, so that sort of magnitude.
#### David Worby
And David, what are your thoughts?
#### David Kohn
Yeah, and I think, just building on the point you made about some of your
third-party suppliers, some of your plugins holding data, I think it probably
makes it more important that you do try and stitch that data together. So
let's say you have a search provider, you have a recommendations provider, you
might have a different provider who's doing your pop-ups. All of these people
are collecting first-party data. Yeah, they're collecting the data about the
behavior of customers on your website, okay? So it probably becomes more
important for you to try and stitch that together so you can build the best,
widest possible picture of those customers and their behavior. And if I hark
back to my time at Heals, where we know that many customers embarked on a
multi-visit journey before they got to their final purchase, something like
that is critical because each time they arrive at the site, you don't
necessarily want to have to go through the same process again. It would be far
better if you could present them with an experience. There's all, you know, we
know that you've been looking at sofas three or four times. Maybe it's now
time for you to have a conversation with a member of the team, or maybe, you
know, perish the thought, maybe it's time we offered you a bit of a discount to try and get you over the
line. So I think that stitching together of third-party data, sorry, I'll say that again because that's not right, but that stitching together of the first-party data that you've got from your third-party tools is probably more
important now than it's ever been.
#### Mark Pinkerton
So even offer somebody a free coffee to come into the store and try the sofa
out if they'vebeen looking at the sofa six times.
#### David Kohn
Always a winner. You'd be amazed how much a single cup of coffee can make
someone spend five grand on a sofa.
#### David Worby
Okay, right. So before we come to the final little part of this subject, I
just wanted to touch on the org design question. We're talking, I guess, in a
way that says the customer acquisition person is optimizing their own
performance. We're talking about the way in which certain segments of the
organization are using data to make better decisions to enrich the outputs
that they create. But there'll be some people listening to this podcast who
maybe have moved a little beyond that. Maybe they even have a data science
team of individuals in white coats, or maybe they're aspiring to it. So Mark,
could you just give us a sense of how the world of data science applies to
this kind of subject and what you're hearing?
#### Mark Pinkerton
So I don't think data science is about men in white coats or women in white
coats necessarily doing things, but it is about people taking a data-led view
of the world. And really for me, the purpose of data science is not only to
understand this highly technical landscape that we've talked about with, you
know, if you've got 20 or 30 different systems working across your digital
domains that you're trying to pull all the data together to get an overview.
For me, that's still an analysis job. But the data science bit is really why,
you know, you're trying to look at...
the future, the past does not predict the future, you are at the same time
trying to on the basis of statistical probability. you can take on, you know, somebody
in the conversion funnel would actually be either to try and nudge them into
the final piece or some other thing that you might wish them to do and so
you're trying to change the way that you operate to increase the probability
of the outcome that you want. Typically it's going to be sale but it might not be. Then, you know, you're also looking at the predictive likelihood of saying if I offer a special promo at this point in the customer journey, what's the likelihood that it will make any difference or I'm looking at all this data together and I'm trying to produce some regression analysis across the whole piece to try and understand the relationships between two completely disparate data sets and really the point of that is that I'm trying to understand cause
and effect from things or in fact a causal effect of A on B or B on A.
#### David Kohn
And I guess, my perspective, I've typically come from smaller organisations
and I've always thought data science was a bit of a big company luxury, I've
always felt it's a little random in that here's a load of data, go away and
tell me something interesting. But I think it's time is upon us and I think
all companies, large or small, should be thinking about what they can derive
from their data. The critical thing is that it isn't random, the critical
thing is that you direct it to issues that you know are current and that will
make a difference, so it's no longer just a large company luxury.
#### David Worby
Yeah, I think that makes sense. I think we talked before about very few, at
least UK-based retailers and brands, are setting or appearing to be set up
data science capabilities. There's one or two, and there's maybe one or two
good examples. But for most, it's about a progression from using data to make
good decisions that become elevated. But I guess our advice to most people
would be to avoid the temptation to try to run. Walking will get you a long
way down the journey.
#### Mark Pinkerton
Yeah, and there are lots of things in analytics that you can add in and you
can make sure that you're covering off, so like the impact of availability on
conversion rate or making sure that you have margin analysis thrown in with
your web analytics so that you have an understanding of the impact on that.
But web analytics and a lot of this data and analytics side is not a precise science, that's one another important thing to think about. We're trying to identify trends and the causes of trends, we're not trying to say this is
the absolute answer. There's always a disparity between the finance recording and the analytics recording on a transaction, you get a different number of transactions because the finance system cuts off at midnight and the web analytics might not cut off until 1am, so there's a difference and you just have to kind of accept the
nature of the beast on some of those sorts of things. But data science will also become increasingly important because it's going to be the lifeblood of your AI systems if you are using AI in any shape or form and AI is becoming much more prevalent within the mass market because the engines are now out there to do it, but you've got to feed them some data, so that's going to come from all these sorts of systems that we've been talking about.
#### David Kohn
And I think my final point on this would be picking up what Mark said is not only is data science
not a precise subject, it's almost impossible that you will reach a decision
where there is a definitive right and wrong answer. All this is about is
trying to improve your probability of making better decisions and then
identifying whether or not you have made a better decision.
#### David Worby
Well, what a way to end. Thank you. That was great.
Thank you to my guests,
David and Mark, and thank you, our listeners, for listening to this episode.
We look forward to speaking to you again on the next one.