A licence waste audit method for small SaaS estates

Question and scope

This study asks how many paid SaaS seats are assigned to people who cannot use them during a declared period. The population is the active subscriptions on the company card, and the unit is one seat on one product for one billing month. The result is a reconciliation, not a judgment about the value of the tool. Freeze the product list and the billing month before looking at usage. A licence that is missing from the product list should be added before the study begins, not after the first report.

Exclude seats that are intentionally reserved for a role, a shared service, or a pending hire with a documented start date. Include trials, annual plans, and products bought through a reseller if they appear on the company bill. Record the source of each subscription so the owner can trace it later. A licence that nobody can explain is itself a finding. The scope note should say which business units and regions are included.

Reconcile seats to people

Export the user list from each product and match it to the identity provider. An account that exists in the product but not in the identity provider is either a local login or a departed user. An account in the identity provider without a product seat may be covered by a group or may need one. Resolve those mismatches before counting anything as waste. The reconciliation is the part that turns a billing export into a defensible finding.

Keep the matching key consistent and avoid matching on display names. Two people can share a name, and one person can have several addresses. Record the match method and the number of unmatched rows. The unmatched list is often more valuable than the final waste figure because it points to an ownership gap. Ask the product owner to confirm each unmatched account rather than guessing from the address.

Define waste categories

Separate dormant seats, duplicate tools, over-provisioned plans, and seats held for a legitimate but undocumented reason. A dormant seat has no meaningful activity in the period, while a duplicate tool performs the same job as another paid product. An over-provisioned plan includes features the team does not use, but that requires a plan comparison rather than a usage count. Each category leads to a different conversation and a different owner.

Do not label a seat as waste just because usage is low. Some roles need a licence for a monthly close, an audit, or an emergency. Ask the manager whether the access is required and record the answer. A seat with a named owner and a stated purpose is a cost decision, not an accounting error. The report should show those confirmed seats separately so they do not distort the waste total.

Sample and verify usage

If the estate is large, sample products and seats rather than exporting every event. Stratify by spend so the sample includes the subscriptions that matter most. For each selected seat, check the last sign-in, the last meaningful action, and whether the person still works at the company. A sign-in alone can be an automated sync, so look for a real user action. The sample should include at least one seat from every product above a spend threshold.

Use the product's own audit log where it exists and the identity provider's sign-in log as a second source. When the two disagree, record both and note which one the owner trusts. Do not infer usage from a billing portal that only shows seat counts. The evidence should answer whether the person could and did use the tool. If the product keeps no usable log, mark the seat as unsupported rather than assuming it is dormant.

Record the exact query or report used to produce each usage figure. A number that cannot be regenerated after the analyst leaves is only an anecdote. If a product changes its audit log retention, note the change before comparing this month with the last. Store a small screenshot or export with the date beside the reconciliation. The next audit should start from the same definitions and a known data source.

Estimate cost honestly

Calculate the cost of each category using the actual invoice, not the list price. Separate one-time charges from recurring ones so the report does not double count. For annual plans, show the remaining term and the cancellation rules. A seat that cannot be removed until renewal is a future saving, not cash available today. The finance owner should be able to trace each number back to a line on the invoice.

Present a range when the contract terms are unclear. State the assumption behind the range and the person who can confirm it. Avoid adding every low-usage seat into a single scary total. A smaller, defensible number is more useful for a renewal conversation than a large estimate that the vendor can challenge. The report should distinguish confirmed savings, likely savings, and savings that require a contract change.

Handle privacy and contracts

Usage data can reveal how someone works, so limit the review to the fields needed for the decision. Do not open documents or read messages to prove that a seat is used. Share results with managers at the level of seats and products, not individual behaviour. Keep the raw exports in the approved system with restricted access. The review should collect the minimum evidence needed to support a cost decision.

Read the contract before recommending a cancellation. Some agreements require notice periods, minimum commitments, or a true-up at renewal. A seat removed mid-term may not reduce the bill at all. Record the contractual constraint beside the recommendation so the owner can act with the full picture. If the contract is ambiguous, ask the vendor for written confirmation before the report is finalised.

Use a second reader

Ask a second person to recheck the matches and the waste categories for a sample of seats. Give them the same exports and the same definitions, but not the first reviewer's conclusions. Compare the results and investigate every difference. This is especially important when the recommendation could affect someone's access. The reviewer should be able to reproduce the seat count without asking the first analyst for help.

Record the agreement rate and the changes made after review. If the second reader finds a systematic error, redo the affected category rather than patching individual rows. The report should say who reviewed it and when. A reconciliation that cannot be reproduced is not ready for a budget meeting. The review note should also list any product where the data was too weak to classify.

Limits and next use

The audit measures assigned seats and observed activity, not the value the tool delivers. A seat can be heavily used for work that no longer matters, and a lightly used seat can be essential. The result also depends on the billing month, the export quality, and the contract terms. Do not present it as a complete software asset management programme. The method is designed to answer a narrow spending question, not to rank employees.

Repeat the reconciliation at the next renewal and compare the categories. Use the first result to fix the worst ownership gap, then measure again. If the same product appears with unexplained seats every quarter, the problem is the buying process, not the seat count. The audit is a starting point for that conversation. A simple approval step before a new seat is purchased may save more than the first cleanup.