J: What does that actually look like? Are you running AB tests and getting data from a sampling of folks and then using that to dictate larger campaigns or something along those lines?
Sameer: Yeah. We look at historical information for people who have not been in the market but who have responded to a particular offer or a particular proposition. We then find out patterns within that data and we translate that into people who might again be in the market. It’s all based on historical data which is more recent, so you can predict from that. From a practical application, very much like Amazon style, if you bought something and people like you have also bought something and the browsing behaviour and all the other data, then you can predict with much more accuracy what their actual need is and when they will be in the market for that particular product.
J: And does that apply to a specific channel or across channels?
Sameer: No, the information is there to be used across any channel. So pretty much you can use it for one particular channel or you can use it for omnichannel marketing because what data science does, and it’s an important distinction, is it identifies a customer need and that customer need can then be fulfilled through any channel or through a multitude of channels.
J: Do you have any specific examples you can give of where you’ve seen this deployed successfully?
Sameer: Yeah, lots of examples. We ourselves have done loads of case studies for an insurance client for example, where we help them bring new customers into their portfolio, identify a market gap which wasn’t there, so predicted the need for a new product within a new market, and then, as a result, they completely developed a new brand to cater to that market. And it helped them gain market share. I think 5% of their revenue now, within the last 18 months, has been driven by that particular product.
J: So, even in advance of trying to acquire the customer, just doing the analysis to determine where there may be some white space in the market. That’s a great application. How about from a savings perspective? I think there’s a little bit of trepidation to invest in data science because obviously, the skill sets are relatively expensive, and the discipline takes a while to build, but is there a return outside of getting lift (which is I think the thing everyone’s targeting)? Are there additional forms of return?
Sameer: Absolutely. I mean, it’s all about, as we said, the whole definition of data science is about business acumen and the return of investment. The way we justify the investment to the client is around how much return they’ll get from a commercial basis. So again, case studies like a pure acquisition model or a response model developed to predict who will respond to a particular campaign, we have saved clients money in the range of £1 million per quarter purely by doing that kind of modelling. So there are lots of costs saving, but I think the way we need to see data science going forward is as an investment centre rather than a cost centre.
J: Yeah. Makes perfect sense. As we move past the customer acquisition phase, and the journey goes into the customer actually using the product or the service that they purchase, assuming it has some digital component to it, how can data science be used to actually improve that product or service experience in real-time?
Sameer: I mean data science, as I said, identifies customer needs at every stage. So it can be used across multiple industries across multiple channels. We can use data science, for example, to predict crop yield for an agriculture client, or for a financial services client, predict the investor portfolio, what features they would want to use and how the investment will blossom over time. For insurance clients, we can predict which other insurance customers are more likely to buy from and how to increase the lifetime value through upselling and cross-selling.
And the real-time element, which we touched on rightly, can be deployed digitally across the web content management platform. When customers are on site, let’s say, for a travel client, which is another industry that is a big user of data science, you go on, you know somebody’s got high lifetime value but they’re looking to churn, you do something instantly because that’s the point of engagement that we want to capture. So, in real-time, your offers and your content changes for that particular customer.
J: Anywhere that you’ve got data that can help either model out for a customer how something could evolve over time, or predict what that customer might do and then change the experience to influence the right behaviour, that’s a place that we could use data science?
Sameer: Absolutely, yeah. I mean, as I said from the start, data science is all about identifying customer need. And then once you’ve identified that, the execution platforms and the technology bit comes in, and that is where you can do whichever execution you want. For example, in real-time you can communicate with them, you can give them another product feature, you can give them another offer, so it can be used quite universally across a multitude of platforms and across a multitude of media.
J: You kind of alluded to this, but as we move into the retention phase of the journey or the engagement phase, where have you seen value in terms of applying this?
Sameer: Lifetime value is a big thing. We have done lots of studies and work on lifetime value models where you identify people who are more likely to buy again or people who are more likely to buy another product from the same provider, but it’s all about preempting that need and putting that idea in the consumers head and that other product is the need and that is what they want. We have seen lots of examples and done lots of cases with clients where lifetime value models have given them eight times the value of what the first purchase has given them. And we try to change business mindsets in terms of making lifetime value the KPI for the business, rather than assessing campaign performance and the ROI based on one particular campaign or one particular media. It’s more about how the consumer during their lifetime performs and which triggers are helpful to make them the valuable consumer that they are now.
J: I would imagine that the data you’re collecting on that ladder phase of the journey can help influence your acquisition strategy in the beginning.
Sameer: Yeah, absolutely. It’s about the whole journey and understanding of the consumer throughout the whole journey. Once you’ve done that, and you’ve understood your customer and their journey, then you can apply it to the acquisition phase and the prospect phase where you say, “I know who this consumer is.” There are works that we have done in the past where we have calculated prospect lifetime value, so they haven’t even become a customer, but because you’ve got that knowledge that they might become a customer and once they become a customer how they will behave. So exactly to your point, you can use that at an acquisition level as well.
J: It’s fascinating because what we see often is that these different phases of the customer journey are often owned by different parts of the organization. And so, as an end customer, your experience is somewhat disjointed and part of that is a creative problem, but I think part of it’s a data problem and not passing that data and knowledge throughout the customer journey. Can data science be looked at as a red thread that can make these experiences more meaningful for customers?
Sameer: Yeah, absolutely. I mean, that is what data science is all about. The reason that data analysts in the past haven’t done the whole customer experience is because data had existed in a silo. So remember, when I told you about the definition of data science and technology is a key part, what technology has now allowed is for you to stitch all the data together rather than data sitting in silos of acquisition, retention, lifetime value, etc. Once you stitch all the data together, then data science practices allow you to analyze that customer experience as a whole rather than different parts of the customer journey. And then the business acumen helps you in the whole journey as to how you apply the business knowledge.
J: Makes sense. There’s a lot of talks out in the industry today about machine learning and artificial intelligence. And obviously, data science is a component of that, but can you talk a little bit about how those things relate to one another?
Sameer: Yeah, of course. Data science is more of an umbrella term. It used to be used as a catch-all phrase to define every bit of what you can do with the data including data engineering, data quality, data processing, insight generation, and then the deployment of data, so it encapsulates everything. Before we come on to machine learning, I think there’s the old world of analysis around the descriptive and the diagnostic element which was ‘Tell me what has happened and tell me what is working and what isn’t working?’
J: On the maturity curve, those are the first curves, phases of getting insight from data?
Sameer: Yeah. So that was the typical maturity curve, probably 10 years ago, when the data analyst looked at those things, developed reports, developed campaign reports, and post-campaign analysis to do that. When we slightly started migrating towards data science in the maturity model was when we started developing predictive models and more automated predictive models, which is where the real machine learning starts to take shape. Then we moved towards the more preemptive stuff, where we use data to transform businesses. Often good examples are Uber and Netflix, where they use data to develop a taxi business or they use data for content generation and all that.
When we come on to machine learning, I would always anecdotally describe it as how my son learns — the algorithm in him through my genes as to how he learns — he learns from his nursery teacher, he learns from his surroundings, from his peers from everyday world that he sees. That is an automated form of learning. More ingestion of data goes into his brain and he’s learning and picking up words and everything as he goes. And that is essentially what machine learning is in a nutshell, in a very simplified example.
And when we come on to artificial intelligence, machine learning is just a small part of artificial intelligence. Artificial Intelligence has got a bigger remit. So, it can be applied to save lives, it can be applied in agriculture for crop yield, it can be applied in the financial services for investor portfolio, and it can be applied for travel companies for personalization. Machine learning is a part of artificial intelligence, which is a part of the catch-all phrase that is going to be used which is data science.