Quite low activity on the blog, I’m aware of that. However, participating in the latest issue of CFO World (again). This time on the topic of predictive analytics and how to become an data-driven organization. Below is an English abstract of the article.
That the amount of data generated in the world is increasing exponentially should hardly come as a surprise. However, one of the main things that unites the digital disruptors and differentiates them from the rest is the maturity to base decision making on date. The access to data isn’t exclusive to the digitally native companies though, instead most companies have all the data needed to become predictive. Yet, decision making is still made reactively, often based on belief or following long investigations with the focus of minimizing risk in the decision.
Around the corner awaits artificial intelligence, where the data becomes autonomous and able to act on itself. The business opportunities are huge, ranging from personal assistants moving beyond self-service, to automation of numerous tasks on and off work. For some years it has been early on the hype-curve, now finally starting to mature into interesting commercial applications. Companies not capturing the opportunity risk lagging behind, ultimately being unable to compete on a market driven by algorithms.
Becoming data-driven is more a question of culture rather than one of technology. Most companies today have an reactive analytics function where reporting is done bottom-up. The analysis is descriptive in its nature, meaning it aims describe what has happened and compare with a previous period.
The next step on the maturity curve would be to become predictive. By using machine-learning algoritms on available data the company would be able to understand what will happen in the future. Take an example – a retailer knows that winter jackets is their most profitable category, wanting to drive more sales of the category. By combining different data sources the retailer would be able to understand that customers buying blue shirts are more likely to buy winter jackets if the weather prognosis shows a high probability of rain. By those insights the company will be able to better coordinate their stocks and marketing to drive profitable sales.
According to Forbes (2015) 86 % of companies working with predictive analytics report higher ROI on investments.
The next step when a company has become predictive is to institutionalize and automate the insights, thus becoming prescriptive. Thus, moving the mandate to make decisions to the algorithms. This might sound scary and is where the real test of culture and maturity comes in. In the above example this would mean letting the algoritms restock the warehouse and automatically market and expose the jackets in relevant channels and in store to relevant customers during relevant whether forecast periods. When the sun shines, there might be another best-selling product for customers buying shoes…
Amazon.com is an interesting and always relevant example when it comes to working maturely with data. About 35% of the total turnover comes from prescriptive models. But, it doesn’t end with marketing and sales, but as much about operational efficiency. In the gigantic warehouses of the company, the warehouse-shelves are using predictive-AI where the shelves move closer to the entrance depending on their current relevance, thus minimizing the time it takes to fulfill every order and restock.
Some other interesting examples include:
- At Stanford, researchers has been able to show that predictions are better to discover breast-cancer than specialist physicians
- A telco has been able to minimize churn by predicting customers three times as likely to opt-out of the contract when it expires
- An insurance company saves 40 MUSD yearly by predicting premiums only based on information on the insured vehicle
Becoming prescriptive is a journey that starts with understanding what questions your company would benefit from being able to predict. Then, start with small-scale experimenting on setting up the models and algoritms, understanding what data you have and where lies the potential. Seeing the models in practice and testing will drive further maturity, curiosity and cultural change as the organization realizes the potential gains.
The final step is to institutionalize and automate, securing a governance where future decisions are made based on data. The journey isn’t done overnight, but in order to catch the opportunities and not risk becoming a laggard competing with algoritms it needs to be started swiftly.
Next week on November 24th, I’ll also participate in a panel on CFOlive at Grand Hotel in Stockholm, discussing new demands and opportunities for the CFO in a disruptive world. Be sure to attend if you’re a CFO and happen to be in Stockholm.