Built from public evidence, scored against named benchmarks

The Stack-Tax Index studies market groups (company-size bands and industries) for the SaaS bloat that public evidence reveals. Each measure is tied to a named benchmark source, published independently of the sponsor and cross-checked against separate sets of data. The main sources are Vertice's SaaS spend benchmark, the Capterra and TrustRadius review panels, Vendr's contract price points, and SEC filings. On the Detailed Analysis page, each profile ends with an Evidence section listing the exact sources behind its charts. We never average measurements that are not comparable. The benchmark sets the level, and the cross-checks confirm the direction.

Unused Seats carries the most weight in the overall score. App Sprawl and Spend per Employee carry equal, middling weight. Price Increases carries the least, because its evidence base is the thinnest. The exact weights are set for each published edition and versioned with it. Unused Seats and Spend per Employee both draw on the same spend-platform benchmark. To turn that shared spend reading into a dollar figure, we apply the unused-seat share and the Price Increases premium to it separately. We then add the two results. App Sprawl and Price Increases each rest on their own separate sources.

We cap extreme raw values. We then rank each group against the other groups it can fairly be compared with. Its position among those peers becomes its score for that measure. We combine the measure scores so that a weak measure drags the total down more than a strong one lifts it. This means a group has to score well across the board to rank high. Where a group's evidence falls short on any measure, we publish a range instead of a single score, and we say so on the group itself.

Because peer pools are small, the percentile scores land on just a few round steps. That comes from ranking within a small pool. It is not a rounding of the underlying data.

Size bands run from Micro to Enterprise. We publish their employee-count boundaries alongside the ranking data. We rank industries on an industry score built from two measures, App Sprawl plus Price Increases, because those are the measures with dependable industry-level evidence today. We also work out each individual size-by-industry combination, and publish it with the state of its evidence.

Some report pages show how much more often than average a software category shows up in a group. We take the category's share of that group's review activity and divide it by its share across the full pool of reviews we studied. A value above one means the group reviews it more often than the pool does. A value near one means it matches the pool. This reads review behaviour, not the software actually installed.

The band measures rest on named, formally maintained reports. So a band's rank moves when a new edition publishes against refreshed evidence. Industry ranks work only half this way. Their Price Increases measure reads the same procurement benchmark, but their App Sprawl measure reads a rolling pool of reviews. So an industry rank can move on a re-run against a refreshed window, with no new edition involved. We also check the band ranking against revenue per employee. That benchmark is published by revenue band, so matching it to our headcount bands requires an assumption about revenue per employee. That assumed figure is the very thing the check reads, so the check leans on what it is meant to test. With only a handful of bands, treat it as a sanity check on direction. It is not a claim that the result is more than coincidence, and it is not an independent confirmation. It ran as expected in this edition. Three bands meet two conditions: the revenue benchmark overlaps them, and all four of their measures are observed directly. Across those three bands, a bigger Stack Tax goes with lower revenue per employee. The index describes market groups and never scores an individual company. One more limit: the band-level app count rests on an identity-provider system, which sees only the applications employees sign in to through it. So it is a floor, not a full count. Anything running outside it, including shadow IT, is not in the number. A separate source that automatically tracks software usage does estimate shadow IT, but nothing folds that estimate into this count.