Overview Segment Rankings Detailed Analysis Methodology Self-Assessment

Methodology

How we score each market's Stack Tax

We score each market on four measures of software bloat and rank it against its peers. Next to the score we estimate what the people who run the software cost. Where the evidence is thin, we say so instead of guessing.

How it works

Benchmarks and cross-checks

The index looks at markets, not companies.

We score groups of companies, by size and by industry. Each measure comes from a named benchmark published independently of the sponsor: Vertice's SaaS spend benchmark, the Capterra and TrustRadius review panels, and Vendr's contract prices. The people cost comes from the BLS Occupational Employment and Wage Statistics survey. The benchmark sets the level, separate data sets check the direction, and we never average figures that are not comparable. Each profile on the Detailed Analysis page lists the sources behind its charts.

The indicators

Four scored measures

Named sources, with overlaps flagged.

App Sprawl: How many apps a group runs for its headcount. Nobody publishes an app count by industry, so for industries we count how many software categories their reviewers use.
Unused Seats: The share of paid seats that go unused or nearly unused.
Spend per Employee: Yearly SaaS spend per employee, against the benchmark for each group.
Price Increases: How much faster a group's software prices rise than the market as a whole. Tools with AI features have been raising prices fastest, so this also counts how exposed a group is to them.

Price Increases weighs the least because its evidence is the thinnest. Unused Seats and Spend per Employee read the same spend benchmark, so to put a dollar figure on waste we apply the unused share and the price premium to that spend separately, then add the two. The overview charts that waste for each company size. The people cost carries no weight at all.

Measure weights

App Sprawl: Middle
Unused Seats: Highest
Spend per Employee: Middle
Price Increases: Lowest

Relative weighting. The exact figures are set and versioned with each published edition.

Scoring

Ranked against peers, then compounded

A weak score on one measure cannot be hidden by a strong score on another.

01
Cap the extremes
Extreme raw values are capped before anything is compared, so one unusual reading cannot carry a group.
02
Rank against peers
Each group is ranked against the other groups it can fairly be compared with, never against the whole field.
03
Position becomes the score
A group’s position among those peers is its score for that measure. Because peer pools are small, those scores land on a few round steps.
04
Compound, asymmetrically
The measure scores combine so that a weak measure drags the total down more than a strong one lifts it. A group has to score well across the board to rank high.

The round steps come from ranking within small peer pools. They are not a rounding of the underlying data.

What we measure

Size bands and industries

We slice the market two ways, and only as finely as the evidence allows.

Company-size bands

Micro: 0–49
Small: 50–499
Mid: 500–2,499
Upper Mid: 2,500–9,999
Enterprise: 10,000+

What each grain carries

Measure	Size bands	Industries
App Sprawl	measured	measured
Unused Seats	measured	not published
Spend per Employee	measured	not published
Price Increases	measured	measured

Industries are ranked on the two measures with dependable industry-level evidence today.

Ranked, or a range

Ranked: Every measure clears the evidence we require. The group carries a single overall score and a place in the table.
Provisional: One measure or more rests on thin evidence. We hold the single score back and publish the range the score could sit in, with the reason, rather than estimating past it.

What each figure needs

How often a category shows up: Capterra reviews from 2023 to mid-2026. Each bar is a category’s share of the group’s reviews divided by its share of the whole pool, so 1x matches the pool. A product listed under several categories counts toward each. Categories with fewer than 200 reviews from the group are left out.
Overlap of competing products: Reviews from January 2023 to June 2026. A category needs at least 50 named reviewers, and only the top eight by share are shown. Reviews under one name count as one person. This counts reviews, not purchases, and an overlap does not mean the products were in use at the same time.
What vendors bill for: Negotiated prices Vendr publishes from real contracts. Each price is the median of the points we hold for that unit, each point is the midpoint of a published range, and the count is Vendr’s number of contracts. Unit names are the vendors’ own and left as they are, so “Users” and “User” stay separate, and any plain-English gloss is ours. Some units price per person and some a whole contract, so prices never compare across units or between groups. A unit needs at least five price points, and the eight with the most contracts are shown.
Variety of software categories: Each year we count the distinct software categories in 1,000 review entries drawn at random from the group. A review counts once for each category its product is listed under, so 1,000 entries cover only a couple of hundred reviews.
Sprawl complaints: Reviews that use any of ten fixed phrases about tool overload, such as “too many tools”, “unused” and “duplicate”, each counted once. Each group is compared with the whole TrustRadius pool, which includes a “large” bucket with no employee range that no single band covers. These are different reviews from the Capterra pool behind the other figures.
Spend share: Shares come from Vertice’s 2025 benchmark, which covers all its customers with no industry split, so every page shows the same shares and only the category list changes. A category appears only if the group’s reviews feature it often and Vertice uses the same name for it, so most are left out.
Most popular applications: Industry pages show that industry’s top three vendors, and the enterprise page the top five of a separate ten-vendor ranking. Both come from the benchmark’s enterprise view, which starts at 5,000 employees, below the 10,000 where our enterprise band starts. The benchmark does not say what makes a vendor popular, and its industry names are matched to ours.

The fifth axis

The second bill: the people who run the stack

Paying for the tools is the first bill. Paying the people who run them is the second.

These are the people who keep a group's seats, renewals and integrations in order. Their wages and headcounts are measured, but the share of their time spent on the stack is our assumption. So we publish a low, central and high figure, and the central one is never a measured point.

The people cost does not count toward the score. The overview adds it to total software spend for a combined figure. Software waste is already inside that spend, so it is never counted twice, and the total is not a savings estimate.

Who does the work

Seven occupation groups from the US wage survey: systems administrators, user and network support, security analysts, systems analysts, IT managers and a share of general operations specialists. Software developers are excluded.

How much of their time

The low, central and high estimates attribute 30%, 40% and 50% of those roles’ time to running SaaS. General operations roles use a smaller share: 2%, 5% and 10%. These are assumptions, published alongside the wage inputs.

What a person costs

Wages are multiplied by 1.43 to cover benefits and payroll taxes, then scaled by company size. Employment and wages come from BLS OEWS, and the 1.43 comes from BLS Employer Costs for Employee Compensation.

People bill per employee = the wage bill of each role × its attributed share of time, summed across roles ÷ all employees × 1.43 × company-size factor.

The company-size factor

Each band’s wage estimate is scaled by this factor. 1.0 is the all-company average. Small companies sit highest and micro companies lowest.

Taken from about 1.3 million software reviews that list a job title. Reviewers are not a staffing census, so the curve is an assumption we use to scale the wage estimate.

Every estimate is a range

The share of time attributed to the stack is the assumption that moves the figure, so each band is published low, central and high rather than as one number. All industries, in US dollars per employee per year. The filled dot is the central estimate.

Micro (0–49)

Low $473 · Central $639 · High $813

Small (50–499)

Low $1,516 · Central $2,048 · High $2,602

Mid (500–2,499)

Low $1,438 · Central $1,942 · High $2,468

Upper Mid (2,500–9,999)

Low $1,187 · Central $1,604 · High $2,038

Enterprise (10,000+)

Low $1,131 · Central $1,528 · High $1,941

Every people figure is modeled. Micro companies also have a provisional composite ranking, separate from this estimate.

Occupation inputs and publication thresholds

BLS OEWS 2025 national occupation inputs. Industry estimates use each industry’s own employment and wages. The software benchmarks have no industry split, so industry comparisons pair those wages with the company-size software benchmark.

Seven operator occupations, national annual mean wage and employment, and modeled time shares
SOC	Occupation	Annual mean wage	US employment	Time: low / central / high
15-1244	Network and computer systems administrators	$99,130	314,340	30 / 40 / 50%
15-1232	Computer user support specialists	$61,860	717,190	30 / 40 / 50%
15-1231	Computer network support specialists	$76,220	146,190	30 / 40 / 50%
15-1212	Information security analysts	$129,180	190,650	30 / 40 / 50%
15-1211	Computer systems analysts	$105,850	519,530	30 / 40 / 50%
11-3021	Computer and information systems managers	$175,140	670,570	30 / 40 / 50%
13-1199	Business operations specialists, all other	$83,050	1,087,090	2 / 5 / 10%

All three conditions must hold to publish an estimate: a complete set of operator occupation cells, an employment denominator of at least 10,000, and at least 1,000 reviews with a job title in the company-size band. Segments that fail are held back with the reason disclosed.

Is the people bill waste?

Not all of it. It is the cost of running the stack, and some of that work has to happen. A simpler stack frees people for better work, but the index does not estimate job cuts or claim the whole bill can be recovered.

Caveats

What the index does not claim

The limits of our sources, and of what we conclude from them.

Ranks move with new data. Band ranks rest on named reports that publish on a schedule, so they move when a new edition comes out. Industry ranks can move in between: their App Sprawl reads a rolling pool of reviews, so a re-run on fresher reviews can shift them.

The revenue check is a sanity check, not proof. We compare the band ranking with revenue per employee. Matching that benchmark's revenue bands to our headcount bands takes an assumption about revenue per employee, so the check leans on what it is meant to test. In this edition it covers three bands, and across them a bigger Stack Tax goes with lower revenue per employee. Three bands only hint at a direction.

The index describes markets, never a company. No company is scored in it. The self-assessment is separate: it scores only the answers you give it, against your band.

The app count is a floor. It comes from an identity provider, which only sees the apps people sign in to through it. Shadow IT is missing, so it is a floor, not a full count.

Self-assessment

Understand where your stack stands.

Tap your industry to start. In about 90 seconds you get your Stack Tax score, how much of your software spend is waste, and how you compare with firms your size. We'll ask who you are before we show it.

What industry is your company in?

Retail & E-commerce Healthcare Software & Tech Financial Services Other

Take the self-assessment

View the full rankings

Commissioned by

Research on the Stack Tax index was commissioned by Supernal. Red Tape Index owns the research and determines the data, weights and rankings independently. The findings are openly available.