Enabling A Modern Analytics Platform

Over recent years, bi-modal analytics has gained interest and, dare I say it, a level of notoriety, thanks to Garnter’s repositioning of its Magic Quadrant in 2016. I’m going to swerve the debate, but if you are not up to speed, then I recommend taking a look here first.

Regardless of your chosen stance on the subject, one thing is certain: the ability to provision analytic capabilities in more agile ways and with greater end user flexibility is now widely accepted as an essential part of any modern analytics architecture.

But are there any secrets or clues that could help you in modernising your analytics platform?

What Is Driving the Bi-Modal Shift?

The demand for greater flexibility from our analytics platforms has its roots in the significant evolutions seen in the businesses environment. Specifically, we are operating in/with:

  • increasingly competitive marketplaces, requiring novel ideas, more tailored customer relationships and faster decisions;
  • turbulent global economies, leading to a drive to reduce (capex) costs, maximise efficiencies and a need to deal with increased regulation;
  • broader and larger, more complex and more externalised data sets, which can be tapped into with much reduced latency;
  • empowered and tech-savvy departmental users, with an increased appetite for analytical decision making, combined with great advances in data discovery and visualisation technologies to satisfy this appetite;

In a nutshell, the rate at which change occurs is continuing to gather pace and so to be an instigator of change (or even just a reactor to it as it happens around you) requires a new approach to analytics and data delivery and execution.


## Time to Head Back to the Drawing Board?

Whilst the case for rapid, user-driven analytics is hard to deny, does it mean that our heritage BI and Analytics platforms are obsolete and ready for the scrap heap?

I don’t think so: The need to be able to monitor operational processes, manage business performance and plan for the future have not suddenly disappeared; The need for accurate, reliable and trusted data which can be accessed securely and at scale is as relevant now as it was before. And this means that, despite what some might have us believe, all the essential aspects of the enterprise BI platforms we have spent years architecting, building and growing cannot be simply wiped away.

[Phew!]

Instead, our modern analytics platforms must embrace both ends of the spectrum equally: highly governed, curated and trustworthy data to support business management and control, coupled with highly available, flexible, loosely governed data to support business innovation. In other words, both modes must coexist and function in a relative balance.

The challenge now becomes a very different one: how can we achieve this in an overarching, unified business architecture which supports departmental autonomy, encourages analytical creativity and innovation, whilst minimising inefficiency and friction? Now that is something we can really get our teeth into!


## What’s IT All About? Some questions:
  • Do you have a myriad of different analytics tools spread across the business which are all being used to fulfil the same ends?
  • Are you constantly being asked to provide data extracts or have you resorted to cloning your production database and provisioning SQL Developer to your departmental analysts?
  • Are you routinely being asked to productionise things that you have absolutely no prior knowledge of?

If you can answer Yes to these questions, then you are probably wrestling with an unmanaged or accidental bi-modal architecture.

At Rittman Mead, we have seen several examples of organisations who want to hand greater autonomy to departmental analysts and subject matter experts, so that they can get down and dirty with the data to come up with novel and innovative business ideas. In most of the cases I have observed, this has been driven at a departmental level and instead of IT embracing the movement and leading the charge, results have often been achieved by circumventing IT. Even in the few examples where IT have engaged in the process, the scope of their involvement has normally been focused on the provision of hardware and software, or increasingly, the rental of some cloud resources. It seems to me that the bi-modal shift is often perceived as a threat to traditional IT, that it is somehow the thin end of a wedge leading to full departmental autonomy and no further need for IT! In reality, this has never been (and will never be) the ambition or motivation of departmental initiatives.

In my view, this slow and faltering response from IT represents a massive missed opportunity. More importantly though, it increases the probability that the two modes of operation will be addressed in isolation and this will only ever lead to siloed systems, siloed processes and ultimately, a siloed mentality. The creation of false barriers between IT and business departments can never be a positive thing.

That’s not to say that there won’t be any positive results arising from un-coordinated initiatives, it’s just that unwittingly, they will cause an imbalance in the overall platform: You might deliver an ultra-slick, flexible, departmentally focused discovery lab, but this will encourage the neglect and stagnation of the enterprise platform. Alternatively, you may have a highly accurate, reliable and performant data architecture with tight governance control which creates road-blocks for departmental use cases.


## Finding the Right Balance

So, are there any smart steps that you can take if you are looking to build out a bi-modal analytics architecture? Well, here are a few ideas that you should consider as factors in a successful evolution:

1. Appreciate Your Enterprise Data Assets

You’ve spent a lot of time and effort developing and maintaining your data warehouse and defining the metadata so that it can be exposed in an easily understandable and user friendly way. The scope of your enterprise data also provides a common base for the combined data requirements for all of your departmental analysts. Don’t let this valuable asset go to waste! Instead provide a mechanism whereby your departmental analysts can access enterprise data quickly, easily, when needed and as close to the point of consumption as possible. Then, with good quality and commonly accepted data in their hands, give your departmental analysts a level of autonomy and the freedom to cut loose.

2. Understand That Governance Is Not a Dirty Word

In many organisations, data governance is synonymous with red tape, bureaucracy and hurdles to access. This should not be the case. Don’t be fooled into thinking that more agile means less control. As data begins to be multi-purposed, moved around the business, combined with disparate external data sources and used to drive creativity in new and innovative ways, it is essential that the provenance of the enterprise data is known and quantifiable. That way, departmental initiatives will start with a level of intrinsic confidence, arising from the knowledge that the base data has been sourced from a well known, consistent and trusted source. Having this bedrock will increase confidence in your analytical outputs and lead to stronger decisions. It will also drive greater efficiencies when it comes to operationalising the results.

3. Create Interdependencies

Don’t be drawn into thinking “our Mode 1 solution is working well, so let’s put all our focus and investment into our Mode 2 initiatives”. Instead, build out your Mode 2 architecture with as much integration into your existing enterprise platform as possible. The more interdependencies you can develop, the more you will be able to reduce data handling inefficiencies and increase benefits of scale down the line. Furthermore, interdependency will eliminate the risk of creating silos and allowing your enterprise architecture to stagnate, as both modes will have a level of reliance on one another. It will also encourage good data management practice, with data-workers talking in a common and consistent language.

4. Make the Transition Simple

Probably the single most important factor in determining the success of your bi-modal architecture is the quality with which you can transition a Mode 2 model into something operational and production-ready in Mode 1. The more effective this process is, the more likely you are to maximise your opportunities (be it new sales revenue, operating cost etc.) and increase your RoI. The biggest barriers to smoothing this transition will arise when departmental outputs need to be reanalysed, respecified and redesigned so that they can be slotted back into the enterprise platform. If both Mode 1 and Mode 2 activity is achieved with the same tools and software vendors, then you will have a head start…but even if disparate tools are used for the differing purposes, then there are always things that you can do that will help. Firstly, make sure that the owners of the enterprise platform have a level of awareness of departmental initiatives, so that there is a ‘no surprises’ culture…who knows, their experience of the enterprise data could even be exploited to add value to departmental initiatives. Secondly, ensure that departmental outputs can always be traced back to the enterprise data model easily (note: this will come naturally if the other 3 suggestions are followed!). And finally, define a route to production that is not overbearing or cumbersome. Whilst all due diligence should be taken to ensure the production environment is risk-free, creating artificial barriers (such as a quarterly or monthly release cycle) will render a lot of the good work done in Mode 2 useless.