Predictive analytics has been around for a while, as has machine learning, but it's only now with the profusion of cloud-based software in marketing that this form of data analysis has started to take off. 

AgilOne is a US company, launched 2012, now branching into the UK, that provides predictive analytics software. I spoke to CMO Dominique Levin to find out more about this technology.

Is it powerful enough to make one-to-one marketing a possibility and not a fallacy?

The landscape

Data is of course the foundation of marketing. Data cleansing might be a fairly dry subject, but it’s important.

"Send duplicate catalogues to a customer, get their gender wrong, or try to sell them lawnmowers when they live in an apartment”, Dominique pointed out, and you’ll soon compromise your marketing ROI.

With clean data, accurate segmentation of your customer base is possible. To achieve sophisticated segmentation of a large database is, however, resource heavy. You’ll need data warehousing, a data scientist (not exactly ten-a-penny at the moment) and plenty of time set aside.

Rules based segmentation can get longwinded as it entails digging deeply into traditional measures of RFM (recency, frequency, monetary value) as well as product mix and reverse-engineering marketing spend.

That might be something, though, that’s set to change with the coming of machine learning and predictive analytics.

Of course, machine learning doesn’t make the use of rules-based data science redundant, but the power of computing in the cloud means predictive analytics software often trumps what can be achieved in-house.

Dominique referred to this technology as a ‘democratizing power’ for marketers. But what exactly is machine learning and predictive analytics?


What is machine learning?


Machine learning, a branch of artificial intelligence, concerns the construction and study of systems that can learn from data. 

A core objective of a learner is to generalize from its experience.[5][6] Generalization in this context is the ability of a learning machine to perform accurately on new, unseen examples/tasks after having experienced a learning data set.

How is it used?

AgilOne’s solution has a number of different algorithms for analysing data:

  • A clustering algorithm
    • surfaces segments of an audience exhibiting certain behaviours. This modelling also helps to group company’s products and brands.
  • Propensity modelling
    • gauges a customer’s likelihood to buy (lead scoring), whether they have money left to spend with you, how likely they are to convert or unsubscribe.
  • A recommendation engine, like Amazon’s
    • surfaces other products or services a customer might be interested in.
  • Marketing spend analysis
    • determines the most successful acquisition and retention methods by looking at how existing customers were caught and kept. 

One of the main uses of these algorithms is to plug the results into an email service provider (or more broadly into a customer experience management package)

As with many SAAS solutions, the possibilities for tying predictive analytics into other solutions are manifold, whether in ecommerce, B2B CRM etc.

Third party data can also be included in analysis, from loyalty schemes to local weather. Ultimately, the power of the single customer view is only as powerful as the analysis one can then perform with it.

The data can also be used with social media, allowing a company to better target social audiences that perfectly match an ‘ideal’ customer base (for whatever product, offer or service is in question).

Future uses?

Dominique rightly pointed at that what excites marketers is tailoring a message to an audience.

The logical culmination of this is tailoring content to individual people in the market. In practice, this may never occur, but certainly fine segmentation is a goal.

Machine learning may eventually allow even for automated content creation and curation at an individual level. The goal is ensuring there’s nothing in your brand communications that isn’t compelling.

What about the uncanny valley?

At the moment, the software works with registered users and doesn’t use anonymous visitor data. There’s a lot of talk in personalisation about the ‘uncanny valley’ and how much is too much?

Many believe that behaviour-based personalisation (other than transactional) is a step too far, because many customers are already shopping with intent.

Especially if the user arrives through search, they already have an idea of what they want to achieve, and pushing them somewhere may interrupt this journey.

It’s also the case that the uncanny valley is felt more by customers who aren’t registered, because they may wonder just how the company has information about them that they feel they haven’t explicitly provided.

And what are the implications?

Well, one might infer that this software does so much of the legwork that it could turf some marketers out of a job (that’s the line I tried with Dominique).

But truth be told, if these solutions mean that marketing at a personal level becomes more sophisticated, then for now, the strategy, content creation and customer service involved will only increase.

Increasingly sophisticated marketing means its success and importance will increase, which is surely a great thing for all marketers.

Ben Davis

Published 28 April, 2014 by Ben Davis @ Econsultancy

Ben Davis is Editor at Econsultancy. He lives in Manchester, England. You can contact him at, follow at @herrhuld or connect via LinkedIn.

1243 more posts from this author

You might be interested in

Comments (3)


Timothy Spooner

Interesting. I agree wholeheartedly that it will create more work for marketers to create and manage the process - however, one question is whether existing marketers can learn to work within this new world, as they haven't shown any real signs of being able to get to grips with tailored customer management strategies, or whether it will require a new breed of marketers to be trained and developed.

over 4 years ago


Simon Burton

An excellent article Ben. Clearly the book on doing it right is still being written. To truly deliver 1:1 personalisation in real-time, requires gathering all of the interaction data in the most simple way possible, not just that which you tagged for. It then requires you to build a dynamic session and customer state on the fly, integrate with a real-time decision (RTD) system and then having an ability to take the decision from the RTD and affect an appropriate action in real-time i.e. sub 150ms. Those actions could be injecting targeted, personalised content into a website either directly or via the CMS or triggering an appropriate call from the contact centre or CRM/email system or any combination of these. No small task I am sure you will agree and one that requires the more technically savvy Enterprises to get to grips. It is however, one that many Banks and Retailers are now embarking on, as true 1:1 personalisation is for sure the new battleground.
We at Celebrus have been pioneering this exact capability with our OEM partners Teradata and SAS and our resellers for some time now and as someone said to us recently in a major Banking client, "the requirement to react to a users interactions "in session" is now an absolute given". The tailoring of content to an individual is a reality today, although a step too far for many. The uncanny valley you mentioned above i.e. anonymous data, is one of the keys here. If you were to base decisions solely on a users transactions or the fact that you know them, this would discount literally millions of individuals visiting your website. If however, you could build a users profile over time and talk to them in each and every stage of their lifecycle, surely this was always the promise of the web! The true beauty of the web is that you don't need to know the user to know what interests them, but the challenge is to use that information in real-time to then talk to the customer in a relevant, targeted and appropriate way.

over 4 years ago


Matt Lovell, Head of Customer Data, Insight & Analytics at Eurostar International Ltd.

Nice article Ben. I think the difficulty for a lot of advertisers is that while the concept of being able to provide real time, dynamic, one to one personalisation is really appealing, the difficulty lies in the cost and resource requirement to achieve it as it centres around capturing and collecting clean and accurate data.

As a result, while I'm sure some companies will start to lap this up, I also suspect that the majority will stick to continuing with their more basic but potentially safer RFM models.

over 4 years ago

Save or Cancel

Enjoying this article?

Get more just like this, delivered to your inbox.

Keep up to date with the latest analysis, inspiration and learning from the Econsultancy blog with our free Digital Pulse newsletter. You will receive a hand-picked digest of the latest and greatest articles, as well as snippets of new market data, best practice guides and trends research.