Building customer analytics capability for organizations is a scary proposition these days. What your organization likely craves for is something like this – insights about customers that one has not even thought of analyzing, let alone do something about it. Unfortunately, you start with a combination of massive legacy data mess and a huge mumbo jumbo of technology building blocks that require significant investments in the environment where these blocks are changing at the rate that has never experienced before.
No other part of most organizations’ solution stack is as complex as the analytics stack. Most companies build it from ground up – make a choice of database technologies, ingestion tools, visualization tools and whole other bits and pieces. It takes tremendous amount energy to make all these work together and then to actually organize and govern data to make sense out of it. In short – months before anything of interest can come out. Additionally, during that time, technology changes and evolves several times. And organization keeps waiting…
A strong case therefore could be made that there is an unfulfilled need for a broad platform around customer analytics that “just works”.
Data Clutter and Workflow Fragmentation – Big Hinderance to Customer Analytics
Customer analytics is a challenging problem to solve to begin with for many reasons. Most companies have multiple systems and tools that manage multiple customer touchpoints. Each of them generate a set of digital breadcrumbs that are essential to stitch together to form a single view of customer.
The problem is that, in typical organization, not only the sources are fragmented, but also analytic databases are fragmented. Additionally, stitching together customer identity is a big challenge and conventional master data management initiatives often face losing battle. (more on this here).
Workflow fragmentation – propensity or need to move data into another database for deeper analysis – creates another set of challenge. It not only introduces delays but wastes a lot of productivity. I recently surveyed a group of data scientists, who mentioned that accessing data, flexing compute and productizing models takes almost 80% of their time with the remaining 20% on actual analysis of data and building models. Shouldn’t this be the other way round?
ERPs and Problem They Solved
In good old days, pretty much every enterprise system used to look like the one described above. You wanted a procurement system, you setup a server, installed a database like Oracle on it, wrote code and made it work. You managed integration between procurement and accounts payable – typically another system. And then between accounts payable and GL. And so on.
Then came ERPs or Enterprise Resource Planning systems. They solved a huge problem – they basically took away the complexity of selecting technology stack and made basic process work together and integrate out of the box. They took away the complexity of managing technology and instead changed the focus on managing business processes. Once you configured business processes – the rest just worked. In subsequent years, cloud based systems took this concept to the next level.
Alas, we still do not have a comprehensive platform that would just work when it come to analytics. (even though some claim to be ones).
Essential Characteristics of ERP for Customer Analytics
- Value proposition is based on how easily and rapidly data can translate into actionable business insights and NOT based on coolness of technology components to make it happen.
- In fact, the technology stack from compute to databases to visualization should not be a concern for users of the platform. The platform should be capable of switching out technologies if newer technologies give bigger bang for the buck.
- The platform takes care of data access policies including compliance to regulatory policies of various jurisdictions, SOX compliance and compliance to organization’s data access policies.
- Holistic approach to solving customer identity problem should be foundational and out of the box. The platform should be able to automatically stitch every new piece of customer identity to an existing customer identity ecosystem. Graph network modeling with probabilistic data stitch does this way better than traditional approaches. Additionally, data profile tools as well as metadata management tools should be out of the box and integrated with the identity ecosystem.
- Data analysis workflows, visualization and presentation should be seamless.
Many of the players in data ecosystem are focused on solving a piece of puzzle in best possible way. This has neither served their customers well nor has it served them well in long run as demonstrated by churn and obsolescence. I believe that there is a strong case in the market for offering that solves the data and analytics problem in its entirety just like ERPs.