Contents
Key takeawaysThe dataNormalizationPeer matchingRankingTermsCurrent and confidentialWhat to do nextFAQA software benchmark is a comparison, and a comparison is only as good as its method. This is how ours works, step by step, including the parts that decide whether a vendor can dismiss the result.
- Executed agreements only. Quotes show what a vendor asked for; signed paper shows what it accepted.
- Normalize to effective unit price over the full term, after uplifts, credits, free months, ramps and bundles.
- Match by size and scope: vendor and product, metric, volume band, term length and deal type.
- Rank among ten comparables to find the floor, the stretch and the gap in dollars.
- Score the terms with the same rigor as the price: at market, weaker or missing.
- Ten well matched agreements beat a thousand loose ones. Match quality decides whether the vendor can argue with the result.
What data should a software benchmark be built on?
Executed agreements: signed order forms, enterprise agreements, amendments and price lists, read line by line. Everything else is a proxy for them, and each proxy has a known bias.
| Data source | Its bias | Use it for |
|---|---|---|
| Executed agreements | None systematic, when matched to your size and scope | The benchmark itself |
| Vendor quotes and proposals | Shows the ask, not the close | Tracking how a vendor opens |
| List prices | The top of the discount ladder | Measuring list price increases |
| Buyer surveys | Self reported, unmatched, silent on terms | Direction only, never a target |
What sits in the Vera AI library
- 10 plus years of executed agreements, order forms and price lists, line by line.
- 1,500 plus vendors: software publishers, cloud and AI providers.
- 500 plus contracts a year across the major publishers keep it current, with a platform that scans 20,000 contracts a month.
- Validation by former vendor sales leaders; the ServiceNow set alone covers 250 plus customers.
The five steps, in order
- 1CollectExecuted agreements, line by line
- 2NormalizeEffective unit price over the term
- 3MatchProduct, metric, size, term, deal type
- 4RankThe draft among ten comparables
- 5Score termsAt market, weaker or missing
How are software deals normalized so they can be compared?
By converting every deal to an effective unit price over its full term. Vendors present the same economics in many shapes: a deep discount with steep uplifts, a shallow one with free months, a bundle priced as one line.
Normalization removes the shape and leaves the economics.
A worked example: two quotes, one winner
Two proposals for 2,000 users on a 36 month term, at a list price of USD 60 per user per month.
| Quote A | Quote B | |
|---|---|---|
| Headline discount | 35 percent | 30 percent |
| Year 1 price per user per month | USD 39.00 | USD 42.00 |
| Uplift | 7 percent a year in years 2 and 3 | Held flat for 36 months |
| Free months | None | Two in year 1 |
| Effective price per user per month | USD 41.79 | USD 39.67 |
| Effective discount against list | 30.3 percent | 33.9 percent |
| Total over 36 months | USD 3.01M | USD 2.86M |
The quote with the higher headline discount costs about USD 153,000 more. Ranked on headline, A wins. Ranked on effective unit price, B wins, and B is the number a benchmark compares.
The adjustments, one by one
- Uplifts: apply each year's increase and average across the term.
- Free months and credits: spread them across the term as a reduction in unit price.
- Ramps: weight each year by the quantities actually licensed in it.
- Bundles and renamed SKUs: map them back to like for like products before comparing.
- Currency and price list: convert to one currency and the same price list basis.
- Support and subscription: separate them where the vendor bills both, so the support base is visible.
How are comparable agreements selected for a benchmark?
By the factors that move price, in the order they move it. Spend tier comes first: discounts scale with size, so a comparison across sizes produces a confident wrong answer.
| Matching factor | Why it matters |
|---|---|
| Vendor and product | Discount policy differs by product line, even inside one contract. |
| Licensing metric | Per user, per core, per employee and consumption pricing do not compare directly. |
| Volume or spend band | The biggest single driver of discount depth. |
| Term length | Longer terms buy deeper discounts and carry more uplift risk. |
| Deal type | New purchases, renewals and expansions are priced on different ladders. |
| Region and currency | Price lists and approval levels differ by region. |
Why sector matters less than size
Buyers often ask for peers in their own industry. Vendors price on size, metric and term far more than on sector, and sector filters shrink the sample. We match on what moves price, and withhold sectors from every comparison to protect the source.
Where the common advice on benchmark data is wrong
The common advice is that a bigger sample always makes a better benchmark. We disagree. A thousand agreements across every size, term and metric produce an average that fits nobody, and a vendor can dismiss it in one sentence. In our work, ten agreements matched on product, metric, volume band, term and deal type produce a result the account team cannot argue with, because it describes buyers exactly like you. Match quality first, sample size second, and every number states what sits behind it.
How is a software benchmark result read?
As a rank. The draft is placed among ten comparable agreements of its size, from the best negotiated to the weakest, and the rank produces three numbers.
Where a draft order ranks among ten agreements of its size
Bar height shows the effective discount each agreement achieved over its term. Illustrative.
- The floor: the comparable, what a customer of your size achieves on an ordinary close.
- The stretch: ranks 1 and 2, the best negotiated agreements in the set.
- The gap: the dollars between your draft and each, at your quantities over your term.
Why a rank beats a percentile
Percentiles hide the sample. "Sixth of ten" tells a negotiator how many agreements sit behind the number and how far there is to go. It is also how a deal desk thinks: how many customers got a better deal than this one.
What happens when there are too few comparables
The band widens before the match loosens. If ten agreements are not available at your exact size, the set steps out one volume band and the report says so, rather than mixing metrics or terms.
How are software contract terms benchmarked?
Clause by clause, against the same comparison set. Each term in the draft is scored against what the comparable holds: at market, weaker, or missing.
| Term | What market best looks like | What weak paper looks like |
|---|---|---|
| Renewal cap | A fixed cap on uplift, stated on net price | Uncapped, or capped on list price |
| Price protection | Unit prices held for the term and for added quantities | Additions priced at the then current list |
| True down | Reduce quantities at anniversary or renewal | Quantities can only go up |
| Unused commit | A share rolls into the next term | Forfeited |
| Credits | Migration and final year credits in writing | Verbal assurances |
| Data and publicity | No training on your data, no reference without consent | Silent, or reference rights by default |
In the AWS commitment in our benchmarking guide, the draft sat ten of fourteen terms behind an executed AWS agreement, and twelve terms improved before signature. The full term set is in benchmarking software contract terms.
How is benchmark data kept current and confidential?
Current through volume, confidential through design. Vendors change price lists, bundles and metrics every year, so a benchmark built on old paper misprices today's deal.
- Freshness: 500 plus new contracts a year from the major publishers, and 20,000 contracts a month scanned by Vera AI.
- Repricing: older agreements are compared on the current price list basis, or dropped when the product no longer exists in that form.
- Confidentiality: sectors and dates are withheld, results are stated as ranks and bands, and no single agreement can be identified from a benchmark.
- Independence: Redress is 100 percent buyer side, with zero vendor affiliations, no reseller agreements and no referral fees.
Source: the Vera AI benchmarking library and the Redress Compliance engagement file, November 2026.
A headline discount is a presentation choice. The effective unit price over the term is the deal.
What to do next
- Collect the full proposal, every line, with quantities, term, uplifts, credits and free periods.
- Convert it to an effective unit price over the term before comparing anything.
- Check the match of any benchmark you are given: product, metric, volume band, term and deal type.
- Read the rank, not the average: where the draft sits among ten comparables, and the floor and the stretch.
- Score the terms against the comparable alongside the price.
- Run it in seconds on Vera AI, or ask for a human benchmark within 24 business hours.
Frequently asked questions
What is a software benchmarking methodology?
The method used to compare a vendor proposal with comparable executed agreements: collect signed agreements, normalize them to an effective unit price, match peers by size and scope, rank the draft, and score the terms.
What is effective unit price?
The price per user, core or unit actually paid over the full term, after discounts, uplifts, credits, free months and ramps are applied. It is the only number on which two differently shaped deals can be compared.
Why are executed agreements better than quotes for benchmarking?
A quote shows what the vendor asked for. An executed agreement shows what it accepted after negotiation, approval and signature, which is the number a buyer is trying to reach.
How many comparable agreements does a benchmark need?
Enough matched agreements to rank the draft with confidence. We rank against ten comparables of the same product, metric, volume band, term and deal type, and widen the volume band, not the match, when fewer are available.
Is industry a good basis for software benchmarking?
Less than buyers expect. Vendors price on size, metric, term and deal type far more than on sector, and sector filters shrink the sample. Sectors are withheld from Redress benchmarks to protect the source.
How do you benchmark a software bundle?
By mapping the bundle back to its component products, comparing each like for like, and then checking the bundle price against what comparable buyers paid for the same components, bundled or separately.