Predictive Analytics and AI Drive Up Customer Lifetime Value

This Autumn RedEye is releasing the results of an 18-month initiative, testing the outcome Predictive Analytics can have on an organisation’s multi-channel campaigns. We have focused our efforts on seeing whether using Predictive Analytics combined with AI driven marketing automation can help improve the customer experience around the key stages of the customer lifecycle – prospect’s first purchase, second purchase, multi-purchase, VIP and churn. Our strategy is to improve marketing performance at each of these stages by using Predictive Analytics to understand where each customer is on their own journey. When the brand understands the customer’s next likely action, they can specifically target those individuals with more effective comms, ultimately, driving up total customer lifetime value.

Results at each stage of this lifecycle have been excellent. For instance, one brand saw an increase of 83.5% in second purchase rate. This, and other case studies, can be found here. Anyone who attended my presentation at either Technology for Marketing or Festival of Marketing recently, would have seen me present the outcome of the longer analysis to see if we could improve Customer Lifetime Value. For those of you who could not attend, you will have to wait for the release of the new case studies to the website in the next couple of weeks.

The obligatory Q&A session followed my presentations at both these events. But to be honest, I always find these questions instructive and rather good fun. Too often, and I’m not alone in this, I get carried away with what I want to say, and questions illustrate key elements that I’ve missed! So, these were the six questions that were asked (although I must admit I thought there were more) with a few more thoughts than I had time to give on the day.

1. How has GDPR affected your data gathering? How did you fight an increase (if any) in unsubscribed customers?
   

   Whilst it has felt like forever, the period since May 25th is still, in the grand scheme of things, relatively short! Our impression is that, in general (can you see me caveating this response very heavily!) the long-term impact on sign ups and consent is relatively little. However, for some organisations their ‘re-permissioning’ experiences have been fairly disastrous. For instance, a database of active contacts of 500,000 reduced to 6,500 (if you are in this group then you are not alone). It’s not the objective of this blog to cast aspersions on the quality of advice given to some organisations, all I can really say is that without the correct permissions, processing data for comms or even for Predictive Analytics is not possible. There are minimum amounts of data required to make Predictive Analytics work, so for many organisations with smaller databases Predictive Analytics may not work and the issues surrounding GDPR only serve to increase that group.
   

2. Do you have an example of using Predictive Analytics for recruitment initiatives – getting new customers rather than increasing the value of current customers?
   

   RedEye has not worked with any organisations to develop models around acquisition. However, our whole strategy is built around recent prospect/customer behaviour as the key driver for predicting their next likely action. Marketers can better understand how an individual prospect or customer is behaving in relation to their brand. By tracking as many interactions, across as wide a number of channels as possible, this can then be compared with the typical behaviour of customers who have completed certain journeys. And this is applicable to many different market sectors.
   

3. What were the actions that came out of the predictive model to reduce churn. How were they implemented?
   

   25 minutes is a very short amount of time to pack in a lot of things. One that I often leave off the list is a detailed description of the treatments employed at each of the stages. But there is a very specific reason for this… the platform RedEye has developed provides the data to the marketer, and it is up to the marketer to then leverage this information. They know their brand and customers better than anyone else. A review of the treatments used by Travis Perkins would be a completely different presentation. Every brand will develop specific treatments and the insight of what Travis Perkins did is therefore of less relevance when we’re looking at how the system was plugged together to provide the outcome. I often say ‘if you knew a specific customer was likely to never buy from you again - what would you want to say to them?’. Every marketer would have a specific answer to this, I am sure!
   

4. How did you link website behaviours to an individual? Was it logged in users only?
   

   At FoM I briefly shot off an answer, which was that we utilise a tag management solution, which was a bit blasé. The RedEye solution has always been built around a personalisation capability centred on the value of an individual’s browsing behaviour, which is also at the core of our approach to Predictive Analytics as described above. We then link this to channel engagement information, transactional data and any other type of data a client has that has a personal identifier of any kind. It is this data that is at the core of the CDP function and therefore the bedrock of Predictive Analytics. With regards to the issue of ‘logged in’, no, the customer or prospect does not need to be logged in, they just have to have given their consent.
   

5. Did any of your clients face major hurdles in pulling together all the data from siloed and legacy data pots? If so, how was this overcome?
   

   I would say that the vast majority of organisations that RedEye work with have internal hurdles with regards to data silos. Some clients who want to input more data find they are restricted by internal systems, and there is very little that RedEye can do to overcome these bottlenecks. But assuming that the data is available somewhere in an organisation, the CDP is there to help marketers resolve these issues. We try to make this work more effectively in two ways. Firstly, we create easier ways to format data into the system, using simple connectors to input (and export) data. And secondly, we offer support staff to help this happen for clients who are resource strapped.
   

6. Which is the best CDP you would recommend for publishers?
   

   If I remember this question from the day it was asked by Nish! Well Nish, as an executive of RedEye I would say get in touch with us! But being a bit more professional, and having asked my colleagues on the Customer Data Platform Institute I would recommend BlueConic and Lytics who I’m informed have good experience working with publishers.

If anyone else has any other questions I would be delighted to do my best to answer them, get in contact with me here.