Peter Ruffley, Zizo Chairman, discusses the importance of companies taking control of their data in a way that focuses on questioning the data that is collected, rather utilising the just ‘store everything’ data model, promoted so heavily by IT functions.
Companies increasingly want – and expect – their huge data resources to support key strategic change. Shifting from revenue driven to profit driven; moving the business upmarket; surely data can help? It can – but it isn’t, because companies are still too hung up on measuring their past performance – the ever-present Key Performance Indicators – rather than asking the big questions that deliver true business value, the Key Performance Questions. The result is that all the effort that is put into fishing in a sea of data, with a plethora of increasingly sophisticated analytics tools, including Artificial Intelligence, fails to deliver any real business insight. The just ‘store everything’ data model, promoted so heavily by IT functions has not provided the promised machine-generated insights.
This is because the whole approach is back to front. Companies are currently not data driven; they are being held to ransom by the technical data handlers. If companies are to deliver value from the extensive (and expensive) investment in data, it is time to stop measuring and start questioning?
It is hard to find a business that hasn’t embraced a ‘data driven’ culture, often as part of digital transformation. Yet while the concept of leveraging data to improve business direction and performance is laudable, that is, sadly, not what these organisations are achieving. They are simply tracking performance. Adding Key Performance Indicators (KPI), extending the depth and reach of measurement metrics, even drilling down for more detail, remains an essentially backwards-looking approach. This culture of monitor and measure is not one that actively uses data to better understand performance and drive change.
What is required is a simple but essential change in mindset; a shift from the tracking of KPIs to the strategic relevance of Key Performance Questions (KPQs). The issue is not, ‘did we meet our targets?’ but ‘how did we meet our targets?’. Not, ‘how many shirts have been sold this week?’, but ‘are we selling shirts to the all the potential shirt customers?’. Essentially, is this the right direction for the business?
Take a parcel delivery company, for example, wanting to use data to support its strategic shift from being revenue driven to being profit driven. The obvious KPI is profitability per parcel – but how does that help strategically? The KPQ is who else can we sell our most profitable service to?
Or the holiday company that has decided to shift up-market towards more expensive and hopefully profitable packages. Measuring every possible KPI to track performance has minimal value and certainly doesn’t identify the customers that haven’t been attracted to the up-market offer, where they holiday and what compelled them to buy. The KPQ’s to be asked are not only who are these non-customers, but did marketing reach out to the right audience in the first place?
The real problem for any company that has created a state-of-the-art cloud-based data lake is that it only contains the data for looking at KPIs and will not even have the right data to support KPQs and have no strategy for getting it. No wonder so many companies default to the track and measure paradigm. It’s all they can do.
This underlines one of the very real issues facing businesses today – the only apparent way to deliver any value from current data models is to add KPIs. Why? Because the data scientists and technology vendors have propagated the myth that computing power can do anything and solve anything, providing it has access to data and unlimited computing power. It can’t, not on its own. Data science needs direction. Data needs preparation. According to recent 451 Research, the biggest barrier to successful machine learning deployments is a lack of skilled resources, followed closely by challenges in accessing and preparing data. And the more data-driven the organisation, the worse the problem. Simply adding more data sources without direction is adding cost, not delivering value.
Taking the KPQ approach turns the entire model on its head and brings much needed direction. Rather than layering tools over a sea of data in a blind and typically futile attempt to realise true business value, KPQs focus the attention. A KPQ identifies the subject matter experts within the business who can give insight into those questions and prompts the essential discussions that reveal the data sources required. Suddenly, rather than looking at 25, 30 even 100 data sources, the KPQ may require analysis of just five or six.
This opens the door to leveraging new technology, to experimenting, building prototypes and using AI to dig deeper into the answers. Indeed, there is no need for the data scientist: the combination of the right question, the right subject matter experts and the right, well prepared data, and then the speed with which the business can unlock insight can be truly transformative. Business experts will immediately see and understand trends; they will have the context and knowledge to identify insights that have business resonance.
In many cases that data will not be within the organisation, it will be third party or generic market data that will need to be blended with internal data resources to deliver insight. And this is where the compute power and the clever technology does have a role to play; where AI can be very quickly used to reveal whether there is any meaningful correlation within the data at all. Data landscaping provides unassailable information regarding the existence – or not – of mathematically identifiable connections between data items. If there is not, a business is either looking at the wrong data, or that data is incomplete. And this is an issue that companies will need to embrace: KPIs measure existing performance, based on internal data sources. KPQs may well demand additional external data and computing resources.
Time for Change
Something has gone very wrong with the concept of data-driven business. Rather than providing insight to support essential business decisions, too many companies are simply sitting back and hoping that the vast quantities of data being collected will – almost magically – provide the elusive gold dust of fresh and valuable insights. It doesn’t work that way – however smart the AI or machine learning. Measuring the business in ever greater detail does not create a data-driven business. Where is the change? Where is the true strategic insight?
Data resources will not deliver value without direction and senior management need to step up and ask the questions. What is the biggest business challenge? Can the data provide that insight? Are the potential gains worth the investment? Unless companies begin to proactively question the data rather than continue to monitor performance nothing will change, and the concept of being truly data-driven will remain a myth.
By Peter Ruffley