Powering Decisions with Data Analytics

Make the most of strategic planning, operational optimisation & BAU monitoring with the right data analytics implementation


“Courage is willingness to take the risk once you know the odds.

Optimistic overconfidence means you are taking the risk because you don't know the odds. It's a big difference.”

Daniel Kahneman – Nobel Prize in Economics & father of Prospect Theory


We are all incompetent in front of uncertainty


“We're generally overconfident in our opinions and our impressions and judgments.”

Daniel Kahneman

We tend to believe that our best-case scenario will realise itself. Which is why companies (and governments – remember USSR) often put tremendous efforts in preparing detailed mid- to long-term plans. If this planning always brings up at least one positive aspect – an alignment in terms of vision and a healthy sense of direction – the painful bit is that events rarely unfurl the way they are expected to. From disruptive competition to slowly sliding consumer base, decision-making requires constant vigilance, openness to change and conflicting ideas/hypotheses, and… data.

“The planning fallacy is that you make a plan, which is usually a best-case scenario.

Then you assume that the outcome will follow your plan, even when you should know better.”

Daniel Kahneman


Instinct is good – instinct AND data are better


“If you don’t get this elementary, but mildly unnatural, mathematics of elementary probability into your repertoire, then you go through a long life like a one-legged man in an ass-kicking contest.”

Charles T. Munger – Vice Chairman, Berkshire Hathaway

In his now-classic Thinking, Fast and Slow, Daniel Kahneman introduced to the world of behavioural economics the concept of System 1 and System 2. To put it in very few words, while System 1 operates effortlessly and automatically (eating, driving and performing other activities), System 2 operates cautiously and prevents us from jumping into conclusions. While System 1 constructs believable stories from insufficient data, System 2’s responsibility is to doubt it, question it and arrive at the right conclusion.

Most long-term strategic decisions justify spending a significant amount of thinking. Decision-makers can – and should – take the time to use their system 2 to its full capabilities. But doing so requires data. Building complex reasoning on no data, insufficient data, or bad data is very often the recipe to failure.

The data speaks for itself: in the study below, McKinsey showed that organisations with more complete and more advanced data capabilities perform much better than their peers. Talent, hard work, and dedication certainly help but – all things being equal – data separates the crowd in two distinct tiers of performers.

Don’t get me wrong, it does not mean that instinct should not take any part of the decision making process. Actually, the best founders and CEO probably owe their success to a better than average gut feeling (along with other traits of character). But the discussions, solutions, and decisions should be based on facts. Even the best gut feeling in the world cannot generate anything valuable without some decent data points.


It is not how big your data is – it is how you use it


“We're a data-driven organisation.”

Everyone

This is a very recurrent statement. Many businesses generate, process, and store a tremendous amount of data. But most can be challenged in their use of the “data-driven” qualification. Only a few are actually driven by data.

Being data-driven means designing hypotheses to be objectively designed, factually tested, and actually implemented down the line.

Some of those hypotheses will come from playing with datasets in a heuristic manner – even though data science now offers a large set of tools to generate insights in a more systematic way. But each hypothesis has to remain fact-based, testable, and ultimately validated/rejected in an objective way.


The right data is better than more data


According to Lean Analytics by Alistair Croll & Benjamin Yoskovitz, a good metric is:

  • comparative

  • understandable

  • a ratio or a rate

  • easier to act on

  • inherently comparative

  • good for optimising on conflicting factors

  • changes the way you behave – "If you want to change behavior, your metric must be tied to the behavioral change you want."

It does not mean that you shouldn’t collect and store a lot of data. As long as you can afford it (in terms of financial, software/infrastructure, and opportunity costs), you should most likely capture everything you can right from the beginning. Who knows what you will need later down the line for more detailed analysis, a new product line, or even pivoting your strategy? But storing it does not mean spending your precious attention on it, or at least not more than what is exactly required.


Change is challenging, and must come from the top...


I would add “Securing internal leadership for analytics projects” to the two highlighted items in McKinsey’s analysis below:

For the best as for the worst, leading by example is paramount. In the specific case of data analytics, the problem is practical as well. Main company-wide initiatives are born from the company’s vision. Since the company’s vision is in the hand of senior management, at least some of the underlying hypotheses to be tested (and then pushed for implementation) have to be defined at the top, then cascaded down to the rest of the business. The hypothesis-testing exercise can thus happen at all level, be propagated back to the top, used for fact-based decision-making, and be transformed into company-wide initiatives.

Only in specific cases (independent businesses or semi-independent business units, geographies with very different profiles/activities, distinct products part of a product ecosystem or portfolio) can we see some exceptions to these implicit rules without implementation failures. And even then: if there is no vision/initiative/driver pushed down from the organisation’s senior management or board, the value of such a disarticulated conglomerate should be challenged by its shareholders.


… with change leaders across the whole organisation


Even though the vision comes from the top, the ramifications are profound and impact the whole breadth of the organisation. Change has to happen at all levels, and data input needs to be incorporated into each and every business process of the company. It makes sense: your analysis can only be as good as your underlying facts, and the output will be implemented directly into shaping the way business is made. Business, Technology and Analytics are de facto intertwined.

The skills required for a proper execution are broad and exhibit significant overlaps, but again can be grouped into three main domains as a first approximation: Business, Technology, and Analytics.

  • Business leaders lead analytics transformation across the organisation

  • Delivery managers deliver data- and analytics-driven insights and interface with end-users

  • Data architects ensure quality and consistency of present and future data flows

  • Data engineers collect, structure, and analyse data

  • Data scientists develop statistical models and algorithms

  • Visualization analysts visualise data and build reports and dashboards

  • Workflow integrators build interactive decision-support tools and implement solutions

  • Analytics translators ensure analytics solve critical business problems


Key Takeaways

A good diagram is worth a thousand words, but I would still like to insist on the few points below:

  • Not being fact-based is not an option for self-respecting organisations/professionals.

  • Having a lot of data available is... nice. However, consistently focusing on the right metrics at the right time is what creates long-term value.

  • Data helps decision, which enables action. Without implementation, the best metrics remain vanity.

  • Making your organisation data-driven is not a punctual/localised action, nor is it easy. It requires the right talents to achieve full alignment and integration into business processes.


Share this essay:




Further Readings

Thinking, Fast and Slow by Daniel Kahneman

Lean Analytics by Alistair Croll & Benjamin Yoskovitz

Superforecasting: The Art and Science of Prediction by Philip Tetlock



This article was streamlined and originally published by Checkout.com as Better Data, Greater Opportunities