Insight at the Point of Action

About 7 years ago I started using a phrase to sum up what I saw as the fundamental shift of data. “Insight at the point of Action”, I used this to describe the shift that was underway, away from reports towards data within decisions. This is why I think that the primary consumer of data will be applications by 2025. It’s also how at Capgemini we’ve been advising clients for several years on how to become data powered enterprises.

The thing about this phrase is that it makes three key points

1. Insight means not just data display, but the intelligence as well
2. It is contextual to the action
3. It is done with the action (i.e real-time)

Well the folks over at McKinsey appear to agree as well they’ve recently released an advisory on how to become a data-driven enterprise, the first three points are exactly about this

Data embedded in every decision, interaction, and process

Data is processed and delivered in real time

Flexible data stores enable integrated, ready-to-use data

The rest of the points are things I agree with too, including of course the importance of collaborative data ecosystems, which McKinsey state will be the norm by 2025.

The term “insight at the point of action”, for me, sums up the journey we are on, and why the Data Inversion is such a large shift. To move from a world where data was a passive participant to where it actively drives the business requires us to fundamentally think differently.

As an example consider the following:

A customer is looking to buy a “widget””, so they do a search, then they find one they want, then they hit “buy”. Consider the following variations.

Base Case:

1. The search list is ordered based on ‘best match’ from a text perspective
2. The customer clicks on the first item, this is Widget Widget Widget, from Widget Widget LLC
3. They add to basket
4. They click buy
5. It’s out of stock
6. You tell them its out of stock, and around the cycle they go

Old school we’d get a report at the end of the day saying which items had been added but failed to complete due to stock levels. This is well beyond the scope of any of the actions.

Variation 1: Insight to compensate

So the first variation is to use the Digital Looking Glass on the buy, here we use a recommendation engine to suggest alternatives so its not just a fail, but its already helping them find something. Our point of action here is the failure point.

Variation 2: Insight to inform the action

Here we show the stock-level, and a forecast demand, so they know that it is out of stock and not hit add, we can combine this with the insight to compensate to prompt an alternative action, preventing the failed buy.

Variation 3: Insight to guide the action

The final variation is to alter the search to bring into that search an ordering that factors in stock levels, propensity to buy and customer satisfaction. In other words we use insight to guide the customer towards an action we believe will more quickly result in them getting what they want.

Context isn’t static

The other point intrinsic within Insight at the point of action is that the context of the action is key. To take a human version for a second, everyone in tech knows that if you are asking a software sales guy for a discount the time to do it is just before the quarter end, and ideally just before the year end. We know that their context is different in that last week of the quarter than it is in the first two and a half months of the quarter.

The same is true in almost every decision, the drivers and indicators of “good” can change. The relevance of insight needs to change as that context changes.

Actions aren’t just by people

This variation of context and need for insight isn’t limited to human decisioning, its arguably more important for machine based decisioning as context is something that machines need to be explicitly told, they can’t just infer based on precedence, emotion or the work culture around them. You can’t fire an AI and expect other AIs to change their behaviour. Contextual insight needs to be more precise and more complete for machine based decisioning than for human decisions. Insight at the point of action can, rightly, leave some context down to the individual, but should never do that for a machine.