Marketing Segmentation & Predictive Analytics

Enterprises at any level need to target their consumers, clients, users with campaigns , measure the result of these campaigns and hopefully improve sales/contacts after each iteration. Practically this job requires a very broad spectrum of experiences starting from web/art/video design (how we format our message), going to language specific skills (how we write our message to raise interest) and of course some black magic art called segmentation (write the right message to the right potential or actual costumer). How marketing people build segments? Well using the attributes of the customers (age, sex, zip code, children etc…) they can create segments (for example all young women living in NY without children) and create campaigns for them (discounted shirts? Why not! ) . Now what is the role of predictive analytics in this area? Well in theory it should help marketing people really a lot: 

  • Discovering clusters (segments?) into the customer base
  • Identifying the key features or influencers that lead to buy an item or do some action
  • Showing what are the items  bought in combo and propose them as new packages to offer
  • Using social data identify leaders and followers into the customer base
  • ….surely  another hundred of insights like the ones mentioned.

Normally a marketing targeting tool leverages a classical database with tables, fields , records and actually the segmentation result it’s “just” a sql query that with some “where” conditions identifies the impacted customers. This tool it’s usually part of a suite, in other words from a vendor you buy a package with it and the package is made usually also of a CMS to build your web site pages, a eCommerce platform to sell your stuff on the net, an analytics/predictive analytics package to do the tasks described before. 

Looking at this , from a pure marketing department perspective, to maximize the value of your investments you should buy the package and enjoy the entire integrated platform without worries. It’s a bit like when you buy an iPhone and you start to enjoy it really when you couple it with an Apple TV, a MacBook and lately an apple watch. 

But while you can be the only one at home that takes this decision in medium/large enterprises you have several good reasons to do exactly the opposite:

1) predictive analytics runs on the largest possible combinations of data not only on the pure marketing/clicks/orders world.

2) it requires tools (big data clusters, in memory engines, complex algorithms, etc..) that are much more sophisticated  than the “normal” analytics capabilities provided by mkt packages . Usually it leverages best in class open source (R,Python,Scala libraries etc..) or specific vendor software (SAS, IBM,etc.. and lately several startups like Ayasdi etc..).

3) people working on it are miles away from marketing people from cab abilities perspective and from a target perspective (mkt usually want analytics to prove its guts feeling, analyst looks at data with curiosity trying to figure out patterns) 

4) From an analytics stand point  usually we want to buy software that we can reuse for all the business cases (supply chain, logistics, operations, etc…) and not only for the marketing business cases.

The usual “attempt to solve” this conflict it’s separation of diuties : analysts discover insights, they translate these insights into features of the customers and this info “should” fly to the segmentation tool and the good mkt guys should leverage these to do super effective campaigns. The result ? A huge amount of effort wasted without any ROI: 

1) mkt people do not own these features so they do not trust them or they simply ignore them

2) when you start the customization journey ( new interfaces that trasport data back & forth from analytics world to mkt planet) you will face bugs, data quality issues, data synchronization issues etc…

3) analyst try to cover multiple business cases as said, so only a fraction of his job it’s actually targeting marketing needs

When you are a startup however all the complications of this approach are mitigated by the fact that your team is small and mkt/analytics teams are in practice 2-3 people in the same room.

So we can apply the startup model everywhere?

The key factor here it’s people and objectives. We can go with a complete E2E marketing & predictive analytics integrated and customized solution but we have to add to it a dedicated team that works on it : creative marketing and data driven decisions can coexist and actually help each other if they are the result of a team of people laser focused on the same business objectives.

For large enterprises building this team is the key to provide  these capabilities as service to the various internal clients. Clients can be for example other marketing departments spread around the world in different countries working on different labels. These clients will benefit of this because they can focus really on their business at regional level and obtain not only an “agency like” service but they will benefit from the entire chain of experiences and results that data flowing at global level can bring.

    Here I am…

    start

    Hi to everyone, according to blog guides in this first post I should describe my self and why users should read my blog. In reality I just needed a place where to write freely about my experiences with different products/technologies.

    You can find here security related articles, specific Microsoft technology posts , some experiences with some Big Data products/platforms and also some other collateral stuff.

    Sometimes I loved that stuff , some other I hated with all my self, but in the end in both cases the important thing it’s the experience coming from it.

    Several times even if I was working on something that was non particularly interesting I discovered precious info that it became really useful later on other projects.

    Of course it also happens the opposite , working on something nice and you discover things that will became later very useful .

    Just an example that comes to my mind :

    I was reading for personal interest Stefan Esser findings on IOS jailbreak and from that I figured out a bit some strategies that Apple uses to handle kernel memory and this became fundamental later when I started to help a team that was struggling to design a decent protection on their app from memory dumps.

    My primary objective with this it’s not to showcase something but receive primarily a feedback on some activities that I’m doing .

    Good or bad feedback does not matter, it’s just interesting to hear multiple voices and ideas.