Methodology guide · 03
You enter a deal. Five things happen before a number comes back, and one of them is a refusal to answer. This page walks through all of them, including the arithmetic.
Vendors do not all price on a comparable unit, so we rank on whichever measure is actually meaningful for that vendor.
| Metric | Used when | You are ranked on |
|---|---|---|
| Net fee per unit | The vendor sells on a unit you can count and compare: per user, per worker, per fulfiller, per core, per FUE. | Your net spend divided by your units. Lower is better. |
| Discount off list | There is no comparable unit, because the deal is a committed spend or a bundle. Cloud commitments, enterprise agreements, capacity deals. | How far below the list value you landed. Higher is better. |
Amounts in other currencies are converted before anything is compared, so a euro deal is never ranked against dollar deals at face value.
This is the step that decides whether a benchmark is honest. You are not compared against the whole market for that vendor. You are compared against the slice of it that looks like you.
The refusal. If fewer than ten comparable deals sit in your cohort, we do not produce a percentile. Not a wide one, not a hedged one, none. A benchmark built on four deals is a rumour with a decimal point, and shipping it would be the single easiest way to lose your trust in every other number we give you.
Your value is placed inside the cohort's distribution. The output is a percentile: the share of comparable buyers you are beating.
The percentile comes with a precision range derived from how many comparable deals sit behind it. A deep cohort gives a tight range. A shallow one gives a wide range, and you see the width rather than a false point estimate.
Two buyers can pay the same price and have struck very different deals. A three year commitment with a capped renewal uplift and a live competitive alternative is not the same purchase as a one year deal signed sole source in March.
So each vendor's benchmark carries the two structural levers that genuinely move price for that vendor. You set them to match your deal, and your standing is adjusted before the verdict is given.
A percentile is a diagnosis. The number you take into the room is the ask.
What the ask is not. It is not a price the vendor has agreed to, and it is not a promise. It is the position the evidence supports you arguing from. What you actually land depends on your leverage, your timing and how the conversation goes, which is the part no dataset can do for you.
Every band the platform shows carries one of three labels, on the face of the output. You never have to guess which kind of number you are looking at.
A real cohort of comparable deals sits behind it. We print how many, and the date the data is current to.
No per deal cohort, but the band was calibrated by an analyst on a known date. We say exactly that, and we never attach a made up count to it.
Neither a cohort nor a calibration date. We call it a directional estimate and you should treat it as a starting hypothesis.
The rule behind the rule. Attaching a sample size to a band that does not have one is treated internally as a defect, not a presentation choice. If we cannot show you the working, we tell you that instead of dressing an estimate up as a measurement.
Three honest limits, so you can discount our numbers appropriately rather than discovering the limits yourself.
| Limit | What it means for you |
|---|---|
| Very large deals | Above roughly $5M of annual value, deal counts thin out for every vendor and for many vendors the curve is extrapolated from tier structure rather than dense observation. Treat those benchmarks as a strong prior, not a measurement. |
| Vendors that publish nothing | Where a vendor quotes everything privately and publishes no reference point, the curve rests entirely on what our sources and our own negotiation work have seen. That is real evidence, but it is narrower, and the confidence label will say so. |
| Very recent pricing changes | When a vendor changes its commercial model, there is a window before the new behaviour is well observed. We re-cut the curve as soon as we see the change in live negotiations, and mark the date it was calibrated so you can judge for yourself. |