My undergraduate thesis has two main outcomes. I’m not going to bore you with the details here - if you’re really interested, you can read my thesis comic here - but I had a couple of thoughts about two different kinds of products/goals/outcomes that I’ll be working on.

Outcome One is a tool that provides a lot of information to businesses. This tool doesn’t directly save anyone any money, but it enables people to make decisions that could.

Outcome Two is a tool that provides recommendations to businesses. Recommendations that, if followed, will start saving money immediately (hopefully).

Which do you reckon would be easier to sell?

I’m thinking it would be the first outcome. And not because it’s more useful, necessarily, but because people making decisions in business like to have the evidence to support their decisions - whether they have to defend their decisions or not, it’s important to have the evidence around. People need to feel like they own their decisions.

As such, I think there’s a difficulty selling Outcome Two-style expert systems to businesses. Unless you frame it in such a way that it accelerates people through a decison making process.

The Decision-Making Process

The classic, rational model of decision-making look something like this:

Decision Making Process

Side Note: Is this model valid? It’s called the rational decision model because it assumes that people are thinking… well, rationally. My hunch is that for engineers, it’s a pretty valid assumption - you’re supposed to at least try to think rationally as an engineer.

The problem with selling a recommendation system is that you’re going straight from the ‘Recognise/define problem’ stage to the ‘Implement’ stage.

What you're asking customers to do when you make a recommendation

And as a consequence, it feels utterly wrong to the engineer. Nobody (thinking rationally) would recommend something without being sure that it was the best alternative.

Side Note: Actually, it doesn’t even really matter what decision making model you use: every model has some kind of step in between the ‘Problem recognition’ and ‘Implementation’ stage.

Framing your Recommendations

If you’re going to sell a recommendation system, you need to hold the hand of the decision maker through steps 2-4. They need to be convinced that the alternative they pick is the best, and be convinced that they’re choosing from a fairly broad solution space.

The ‘hack’ is at step two. Don’t sell a recommendation system - sell a design tool. Make the design tool adept at highlighting the steps that you followed to come to the recommendation. Tell the user a convincing story as to how you got your answer, and allow them to come to the same conclusion.

I’ve heard it said before that a good teacher is one that gives you the tools to learn, but at the end of the day, allows you to learn the lesson for yourself. Sorry if that’s a little too ‘Dead Poet’s Society’, but that’s pretty much exactly what we’re aiming to do here.

Thanks to Phil Ciufo for some of the ideas here.