Thoughts on Success with Data Analytics Transformation
Building BI & data analytics platforms can be challenging, especially in large corporations with data locked away in disparate legacy systems. My intention in this article is not to provide a complete roadmap, just a reminder of some basics from my own experience, which I believe are critical to successful data analytics transformation.
The hype-driven race to the (yet-to-be defined) finish line for advanced analytics or data-driven "AI" can sometimes cause companies to ignore the requirement for building vital foundations. Establishing programme rigour, good data management and the right mix of roles will increase the likelihood of success.
First and foremost, create a well-defined future-state Vision which is “hype-free”, achievable and realistic. This Vision statement paints a picture of a better future and promises to deliver benefits to a broad range of stakeholders (and hence implies their own role in transformation success). Involve key senior stakeholders in the creation of the Vision statement.
A significant “programme definition” phase could be required to fully define the future state (blueprint) and hence to understand the nature of the projects required to achieve it, and the resulting programme benefits. This phase could highlight the need for data foundation projects, with no visible analytics outcomes (e.g. modelling, data scrubbing, integration), which need to complete before the full “data leverage” and faster benefit-generating projects can start. Maybe this means a longer-term commitment to realising the Vision than the CEO/Sponsor had originally assumed.
Identify the corporate KPIs to be influenced in realising the Vision (e.g. sales revenue), and their relationship to underlying business processes and data. If this requires a business process mapping exercise, go ahead and plan for it. Focus early data governance on this priority data and establish the data dictionary.
At this stage, it may be beneficial to tactically aggregate data sets using your existing BI tools, to highlight data quality anomalies. Use your data analysts/scientists to look for root causes. This is a good time for your data governance team to begin establishing data quality rules. Consider industrial strength data quality tools later, once the governance culture is embedded. Establish the effort for data integration and cleansing on the priority, KPI-related data only at this stage.
It is important to identify these building-block requirements early and incorporate their time, cost, effort as inputs into the Programme Business Case. Formal executive agreement regarding the nature and scale of the programme will then follow.
Data Governance culture should be established with regard to the overall Vision (the Vision should probably mention governance as a requirement). It is very important for those involved in governance roles (everyone who touches data) to understand the business benefits of getting priority data well-managed.
Avoid coding complex data transformations. “Lean Before Digitize” as my ex-GE colleagues will remember. It’s possible that a change in business process, and the way data is captured through it, could remove the need for hard-coded, hard-to-maintain data scrubbing routines implemented in an ETL tool.
In considering how the projects are sequenced within the broader programme, they should be grouped to deliver meaningful step changes in either foundation building-blocks or Vision-related benefits. At each delivery, the Business Case should be re-assessed to ensure that the overall programme is still on track to realise the end-state Vision.
On the People Side
Other than roles such as data scientists, the programme manager and the entire technical team needed to design, build and integrate data and configure BI tools, best-practice dictates that the following roles are key:
1. An engaged Sponsor who is accountable (but not necessarily responsible) for almost everything transformation programme-related. Including the realisation of benefits. In a data analytics programme, should this be the Chief Data Officer? Any CDOs out there have a view?
2. A complete set of applicable stakeholders from across the business obviously, but critically those from corporate legal and compliance to ensure appropriate use of stored data.
3. An analytics-savvy Business Change Manager, focussed on benefit realisation, training and driving the adoption of any new tools & processes. This person has a good understanding of the affected business areas and has been instrumental in defining initial Vision-related use-cases. It is possible that this responsibility will need to live on after the core programme completes, with agreement regarding the ongoing process to monitor delayed benefits. A process is also required to identify any emerging projects or programmes that could add further analytics value to the company… creating a BI CoE perhaps.
There are many other factors that should be in place to ensure success. This article hopefully addresses just some of the structure required, based on programme and data change management best-practices.
I heartily welcome your views, comments and discussion.
Mark Cottier
