I’m a creature of habit. Seriously, I am an assassin’s dream. When I find something that works for me, whether it be a daily routine, a food choice, an exercise program, or a catchy phrase, I tend to beat it to death. For example, awhile back I somehow adopted the habit of calling any set of three things that are better together a “trifecta.” Internally, I applied this term to the Oracle Retail Insights Cloud Service Suite (Home + Insights + Science cloud services) more times than I can count, and it manifested externally a bit too. But I’m proud to say that I’m moving on – to risk overusing other new-found favorite terms.
A recent Forbes article centered on 2019 AI (artificial intelligence) predictions across major industries, including retail. Some worthy use cases were mentioned on:
While not mentioned in the article, I’d elaborate on:
Anyway, a reference in the article to glass-box AI was the gateway to my latest catch phrases. The concept is that AI, and its first cousin ML (machine learning) should be applied and surfaced wherever, whenever and to whomever it is best leveraged. For example, AI should be embedded in retail applications to organically support workflows where practical. And unlike black-box AI which requires data scientists to unearth and apply insights, glass-box AI enables citizen scientists, and progresses toward a common goal of data democratization. Two more terms that I might have to drive into the ground.
For retail AI to be glass-box, it should also underpin a retailer’s KPIs where practical.
In a previous post on the New KPIs of Retail - Part 1, I described how the essential KPIs for retail are evolving, largely bolstered by several key trends:
It’s that last one that I’m pushing on here.
Most traditional retail KPI derivations are straightforward, for example Gross Margin Return On Investment:
But many are more complex, like Customer Lifetime Value (CLV).
It has been widely reported that Starbucks’ average CLV is about $25,000. Think about that for a minute - $25,000 per customer, on average! So, the strategic ball that they need to keep an eye on is not a $5 cup of coffee (excellent coffee, I might add), but a $25,000 CLV.
Here’s an example of a fairly simple CLV derivation:
Of course, this formula relies entirely on the customer’s past behaviors. But CLV can, and I believe should, be taken much further than that by applying AI/ML to predict future behaviors.
For example, to what extent is the customer influential enough, say, in social media, or within their household, their town, etc. to encourage other customers to buy? How about including a customer’s latest demographics and psychographics as factors? For a hard lines retailer, the predicted CLV of a 30-year-old might change with a marriage, a home purchase, or an expanded household. Or what if that Starbucks customer relocates much closer to a flagship Starbucks location? Further, with trends toward more retail personalization and targeted offers, history won’t likely be as strong a predictor of the future as it used to be.
And what if, by employing the Oracle Retail Science Platform, we can not only exploit the flexibility of a more advanced CLV calculus through the configurability afforded by our Innovation Workbench, but also parlay CLV into more advanced customer segmentation, targeted offers, and better promotion halo and cannibalization prediction across more intelligent customer segmentations? The value snowballs by mashing-up the Science Platform’s customer sentiment scores from social media, and more – I could go on and on.
So, back to the glass-box, whether your CLV formula looks something like this:
or more like this:
…it should be embedded into operational and planning application workflows as appropriate, injected into relevant and pervasive descriptive, predictive and prescriptive analytics, and to further democratize the data for citizen scientists (see what I did there?) surfaced to the Oracle Retail Home portals of relevant roles, like the Loyalty Manager and the Pricing Analyst.