The traditional BI Maturity Curve may cause your organization more harm than good....
The traditional BI Maturity Curve. It is ingrained in the collective minds of the data analytics community and continues to serve as the starting point for many data discussions.
It has served us well as a reference to describe the stages of BI capabilities growth.
However it also implies, and is often interpreted by stakeholders and even leadership, that your organization:
Should be doing predictive analytics (AI/ML/Data Science) as quickly as possible
Is not "mature" if not doing predictive analytics
So, let's look at the common pitfalls that an organization can fall into when pondering the traditional BI Maturity Curve...
Pitfall #1: Automatically assuming that wherever you are on the curve is not where you should be.
There will always be pressure to move up the curve - from leadership to stakeholders to analysts.
Especially for organizations that are on the "diagnostic" point on the curve - moving into predictive analytics is often seen as THE THING to do.
We have seen excellent "diagnostic" analytics drive key business decisions - the most notable being a reversal of a new pricing model by showing the negative impact to customer satisfaction and the resulting impact to sales. This involved a detailed analysis but did not utilize a traditional ML predictive model.
Likewise there are plenty of examples where ML predictive models were developed but took longer than expected to build and deploy, took too long to integrate into business systems, never gained traction, and/or never delivered the anticipated business value.
Just because one can, does not necessarily mean one should.
Avoid this pitfall by taking a step back and revisiting or refining your existing analytics strategy. You do have one, right?
Re-evaluate your analytics strategy - with leadership, stakeholders, technology teams. If it becomes apparent that embarking on the journey to the next level is the right thing for the business, then go for it. But watch out for the next pitfall....
Pitfall #2: The often unrealistic expectations regarding time/cost to develop and the generated business value of of Predictive or Prescriptive Analytics
Ultimately, successful machine learning and predictive models are dependent on:
The existence of, and access to, the necessary data
Multiple iterations of "learning"
The ability for the business to consume the out of the models into existing business systems, processes, dashboards, and decision making
The amount of time and resources to get access to the data, build and test the models, deploy into the business, "learn" once in production, and then to yield measurable benefits is often underestimated.
Compounding the challenge is the fact that the technology and staff require a significant investment - and the pressure for a rapid ROI. Taken all together you have a recipe for a major pitfall - with the potential for unmet deadlines and underperforming models.
Avoid this pitfall by developing a complete and realistic plan and level-setting on expectations.
Pitfall #3: Launching predictive / prescriptive analytics before the business is ready to consume the output and integrate it into the existing analytics and business processes.
Stakeholders who have been working with descriptive and diagnostic analytics are used to seeing reports and dashboards.
The output of a cross-sell / up-sell predictive model might be a score of 0 to 1 for each customer and product combination.
This might be something that can be easily explained by the data science team and understood by the business, but much more difficult to effectively consume and action....
Where will the model output go? How will it be integrated into existing dashboards? Do new types of reports need to be created? What business processes need to be changed to ensure the information is fully utilized?
Avoid this pitfall through proactive stakeholder engagement to answer these questions. Bring the analysts, dashboard developers, data engineers, business systems team, and Ops people together in workshops to prepare them for this new way of doing business.
Even if you have successfully managed to navigate around these BI Maturity Curve pitfalls, there is one more that catches many organizations off-guard....
Pitfall #4: Overlooking the impacts of predictive / prescriptive analytics on the analytics operations - such as governance, security, data quality, data access.
Moving too fast beyond diagnostic analytics is probably the most common pitfall: Spinning up data science teams before the organization has the necessary data management in place.
The pitfall happens when the focus is strictly on the technologies and the end-result capabilities.
Predictive analytics requires vast amounts of data to be accessible to the data scientists. Are there processes to ensure the right people have access to the right data - and that there are safeguards on how that data is being used?
How robust are your data quality efforts? Data quality issues can have major ramifications on the predictive models...
Avoid this pitfall by not delaying the effort to ensure your organization is as much an operational powerhouse as a technical one. Acknowledge and plan for the non-trivial foundational work in processes, workflows, governance, and stakeholder management before unleashing more complex and advanced analytics into the business.
That is a lot to take in - so here is a quick recap:
Alternatives to traditional BI Maturity Curve
Despite these potential pitfalls, the traditional BI Maturity Curve is not in and of itself wrong or bad. It just needs to be used in the context of creating a common language to describe high-level capability concepts. Nothing more.
Moving away from the concept of a curve, Gartner has established an analytics maturity MODEL. There are five levels - with each successive level providing a foundation for the next higher level:
This model moves away from analytics capabilities and instead focuses on the business value, culture, process maturity, executive sponsorship, and other areas.
A wholistic view such as this helps to minimize many of the issues described above. So, as you set off to navigate around these pitfalls, take a step back to look beyond "analytics capability".
Beyond the BI Maturity Curve
In addition to the potentially dangerous pitfalls, there is the simple fact that the traditional BI Maturity Curve doesn't reflect the real-world.
Analytics Maturity is NEVER linear. It isn't a curve at all. It is more like an undulating wave with peaks and setbacks.
It should better reflect reality - and look like something we can all relate to:
The “ups-and-downs” are inevitable. There is no way to completely prevent the "downs".
The challenge is to anticipate and mitigate the “downs” as much as possible. To help with this, we advocate for a robust analytics framework.
Check out our 3-Part article on Transformational Analytics for insights to help smooth out your journey on the path to building an organization that is truly data driven.
Comentarios