Five hundred data points on one person produce 124,750 possible pairs. Nobody is eyeballing that, which is exactly why most assessment reports stop at the bar chart.
Post 3 in a weekly series focused on building MyAcumen from the ground up using AI-First principles. The decisions we made, why we made them, and how we went about it. Aimed wider than my usual sales audience.
We built the charts, and I’d rather say that plainly than pretend we didn’t. Twenty-five years of consulting and assessment work means we know the questions leaders ask about their teams, and the report answers them directly, in charts and written sections ranked so the ones that matter most for that team come first. That layer reports the facts, and it should.
What a chart can’t do is explain itself. A bar showing five people at 52% on growing the customers they already have is a fact. It doesn’t tell you why those five are at 52% while the rest of the team sits at 78%, and why is the only part you can coach.
That’s the harder job. Put fifty people through the assessments and you have 25,000 data points and over 312 million possible pairs. The explanation for any one chart could sit anywhere in that, and nobody builds a chart for every combination.
Here’s what one of those explanations looks like.
Those same five account managers rate themselves a full quartile above the rating their managers gave them on that same competency. All five are top-quartile on recognition as a motivator. All five share the same dominant personality style. Four separate sources, one pattern, the same five names.
The bar chart can be argued with. Small sample, bad week, the numbers don’t tell the whole story. That explanation can’t be argued with the same way, because every angle of the data agrees with it. It also tells you what to do. Coaching has to land on the feedback, the reward framing and the skill at the same time, because fixing one of the three won’t hold.
How that layer was built, including the part that went wrong.
The first version of the engine matched each team against a set of finding shapes we had written in advance, then padded the list to hit a target number. Every team came back with roughly the same five findings and different numbers dropped into them. It read the band rather than the score too, so a 41 and a 69 produced the same sentence. We pulled it out and rebuilt it around discovery instead, which took months and is the single biggest piece of rework in the whole platform.
What it does now.
It doesn’t match the team against a set of finding shapes we wrote earlier. It reads every source the team has and surfaces whatever earns its place, so nothing gets padded to hit a target.
Findings are weighted on three things. Does it change a decision. Would the leader have guessed it anyway. And how many sources agree, with three or four ranked above two, because that’s the part nobody can work out from the sheets they already have.
The fixed maths stays fixed. Outliers are worked out by a set formula before the AI is involved at all, so it reads a real finding rather than going hunting for one.
Every claim quotes the real score, never the band.
And Ask Mya sits over all of it for the questions no report was going to cover, because yours will always be more specific than anything designed in advance.
What that’s worth to a leader is the difference between knowing five people are weak on something and knowing the three things holding those five back, in the order to tackle them. One of those is a talking point. The other is next Monday’s coaching plan.
The charts tell you what. The analysis underneath tells you why. You need both, and only one of them can be built ahead of time.
