More Decisioning

Blog /More Decisioning

Digital Decisioning — use of AI to reduce or eliminate the human element of decision making — has been applied to retail’s long-standing trade promotion management challenge. We can explore this use case as a way of highlighting both power and limitation.

Consider a consumer products company issuing coupons and promotions. Let’s use the fictional example of Gizmoglide, the world’s greatest shaver, with the smoothest, handsomest face guarantee. When consumers take advantage of promotions, retailers then honor discount offers, and price adjustments flow back to Gizmoglide headquarters. Someone in HQ must settle up with the retailers, ensuring they are fairly compensated for their part in the promotion. This can be complex. Promotions might vary by amounts, percentages, timeframes, specific retailers, geography, etc. Gizmoglide HQ will need to compare invoices from retailers to various contracts and then balance reported sales against point-of-sale data and market measures. The complexity is such that not all invoices are validated. Instead, sometimes they are just paid, and only spot checks are used to discover incorrect claims.

But wait Gizmoglide’s brilliant systems integrator brings AI to their rescue: Machine learning is employed to implement an automated intelligence that senses from all the various data sources the correctness of each invoice, alerts when disputes are necessary, and even negotiates the accurate settlements. And because this learning system is tireless, it never needs a vacation or a snow day, and there is no more spot checking. All invoices are validated. Gizmoglide quickly begins saving millions, achieving return on investment within the first year.

Initially, this seems like an impressive application of AI. Millions saved, right?

Remember, though, machine learning is our new super power. With great power comes great responsibility. (Hey, that’s a good line. Did I just think of that, Uncle Ben, or did I hear that somewhere before?) We must ensure we aim this force at the optimal target.

Many decisions are still being made at Gizmoglide headquarters. Humans are deciding what promotions to run when and where, and generally making the pricing optimization decisions. They are reacting to sales performance, third-party market measurements, retailer pressure, and negotiation around shelf-space, optimal product mix, and placement. They must respond to the pricing and promotion decisions made by competitors selling shavers that do not make us quite as handsome. How many of these decisions can be based more on data and what machines learn from statistical patterns? Maybe we need to aim our AI at a more central part of the pricing game than trade promotions.

Or are we guilty, in an even more alarming scenario, of just using our super power to bring new efficiency to a dated business model? Maybe the millions saved in one form of efficiency is a short-term increment that distracts us from the potential transformation available. For example, should we be using data analytics and machine learning to invent Gizmoglide Shaver Club, which establishes an innovative customer engagement, becomes a fully digital business, reaches beyond the limitations of traditional retail?

Determining the optimal use of AI requires creativity, imagination, and vision. It is perhaps the most critical challenge of this new digital age.

And our Digital Decisioning framing provides a useful guide. Consider the decisions made by all the humans sweeping across a 360-degree perspective, by customers, by partners, by our employees. Which of these decisions have the most influence on our performance? Which of these decisions can be made based on the data sensed from our systems? And where can the machine automate our continuous learning? This is a recommended approach for discovering where to aim our super power.

Combining strong industry insight with this Digital Decisioning method can then ensure we live up to our great responsibility.

Power on, digital deciders.

Post Date: 15/03/2018

Scott Boettcher

About the author

Scott Boettcher developed his first database application in dBase III Plus, back in 1983, when we backed the databases up to 5.25 inch floppy disks. A bit has changed since then. In the time that has passed Scott has led application development centers, global consulting practices, large outsourcing accounts, and Business Intelligence departments for Fortune 100 businesses. He currently leads NTT DATA’s Analytics Practice, an innovative organization that now, for example, enables clients to gain new business insights applying machine learning to petabytes of streaming, cloud based information. Again, a bit has changed.

VIEW ALL POSTS
EXPLORE OUR BLOGS