Methodology
Most tools in this space ask you to trust a score. This page is the opposite. It sets out what Accolgo measures, the published research each choice rests on, the places where we made a judgement call instead, and the claims we deliberately do not make.
What we measure
A 2025 meta-analysis in the Journal of Management pulled together decades of research on how new starters settle, and identified four things that consistently predict whether someone stays: whether they feel accepted socially, whether the role is clear, whether they are mastering the work, and whether they feel they fit. Every question Accolgo asks maps onto one of those four.
That matters because we did not invent the framework. If you disagree with what we measure, you are disagreeing with a body of published research rather than with our opinion, and you can go and read it.
Whether they feel part of things and can speak up. One of the two strongest signals we track.
Whether the job feels like a fit and their work matters. The other of the two strongest.
Whether the job matches what they were told at interview.
Whether they have the tools, access and support. The weakest of the four as a predictor of leaving, and we say so below.
Not all four are equal
The same meta-analysis reports how strongly each of the four relates to somebody intending to leave. They are not the same, and treating them as if they were would waste a manager’s attention. So belonging and feeling valued raise a flag sooner than the other two do.
Feeling set up is the weakest predictor of leaving, roughly half the strength of the strongest. The logical move would be to quieten it. We have not, on purpose: a set-up problem is the cheapest thing on this list to actually fix, and the cost of missing someone is far higher than the cost of an unnecessary conversation. That is a judgement, not a finding, and we would rather say so than dress it up.
Two things most systems miss
Senior people understate how they are doing. That is a well-evidenced bias in workplace self-reporting, and it matters most where you can least afford it. So a senior hire’s answers are read on a tighter setting: a smaller dip is enough to raise it, because the same words mean something different coming from them.
How much tighter is our own judgement call. The research supports the direction confidently and gives no number for the size, so we chose one and label it as a choice.
The slow fade is the one that costs you. Someone who drops sharply usually has a reason, and usually tells somebody. Someone who eases down a little every week for two months tells nobody, never scores badly, and never has a bad week that anything would notice. Accolgo compares each person against how they were when they started, not just against last week, which is the only way that person shows up at all.
What we do not claim
We looked hard at whether we should. Sectors genuinely do differ: we checked it in three independent public datasets, from the United States, Hungary and across Europe, and they agreed closely on the size of the gap between hospitality and finance.
But it is a small effect next to what actually moves these numbers. Industry explains under two per cent of the variation in how satisfied people are at work. Which employer someone works for matters about three times more than which industry they are in. And comparing a person against their own earlier answers, which is what we do, cancels out a group difference anyway.
So we record your sector and use it to show you how your numbers compare with published figures for organisations like yours. We do not pretend it changes how we read an individual, because it does not, and a claim like that would fall apart the first time somebody checked it.
The limitation you should ask us about
Every threshold in Accolgo is a reasoned starting point, calibrated against published research on what makes new starters stay or go. None of them has yet been tested against what actually happened to the people in this system, because there is not yet enough of that data to test it against.
We think that is the single most valuable thing this product can do next, and it is why the thresholds are written to be corrected rather than defended. If a competitor tells you their model is proven, that is worth asking them the same question about.
Sources