That seat logged in yesterday.
A recent authentication may belong to an integration rather than a person. Classify login evidence and check the retained period before treating activity as proof of human use.
Activity data answers a question nobody asked
Every seat-management view reports activity. Almost none of them distinguish a person signing in from an integration refreshing a token, a connected app polling, or a scheduled job authenticating. In an estate with mature integrations that distinction covers a substantial share of seats — and it always covers them in the direction that makes waste invisible.
The seat that never sleeps
A licence attached to an integration shows daily logins forever. It appears among your most active users on every report while no human has touched it since the integration was built.
The rebuttal you have to survive
Present a dormant-seat finding without the machine-login contrast and the answer is immediate: that account shows activity yesterday. One sentence, and the credibility of the whole analysis goes with it.
Add-ons nobody reviews
Permission-set licences are assigned separately from user licences and reviewed by nobody on a normal cycle. They accumulate quietly through role changes and departures.
The platform hands you the data and none of the interpretation
Salesforce exposes the fields required to do this properly. What it does not do is tell you which login was a person, which licence class is genuinely billable, or where the contract floor sits — and each of those omissions turns a good finding into a wrong one.
- 01
Use login type, application context and the integration inventory to classify events. Standard retained login history is limited, so state the window and any additional archives used.
- 02
Several licence classes cost nothing. Recommending the removal of a free licence is the fastest available way to lose the room, and it requires knowing which classes those are per edition.
- 03
Permission-set licence assignment and permission-set membership are different records, and the gap between them is only visible if you compare the two sets deliberately.
- 04
Contract minimums cap what is genuinely recoverable, and they are customer-confirmed facts rather than anything the platform will tell you.
- 05
Reducing seat count can trigger repricing of the remainder under volume-based discounting, so gross seat arithmetic overstates net recovery.
- 06
A user who lives in dashboards and forecasts authors no records while being entirely legitimate — so authorship must work as a veto, never as the accusation.
What the analysis reads
Read-only, through the API, against your own org. No configuration is changed. Evidence exports and retention are agreed as part of the engagement.
- 01User licence assignment across every licence class in the org
- 02Permission-set licence assignments, held separately from user licences
- 03Permission-set membership, compared against add-on licence assignment as a set difference
- 04Login history across the full measurement window, by login type and source
- 05Machine and integration traffic identified and separated from human sessions
- 06Platform last-login field, where present, as corroborating evidence
- 07User active status, cross-referenced against every licence and add-on held
- 08Record authorship across standard objects over a twelve-month window
- 09Feature licence enablement against the activity each one entitles
- 10Purchased licence count against assigned count, per class
- 11Edition and the licence classes it includes at no cost
- 12Contract minimums, customer-confirmed, never inferred
What we establish
Which seats no human uses
Dormant seats identified with the machine-login contrast attached to each one — the count of automated authentications that made the seat look active, stated alongside the absence of human sessions.
Which add-ons entitle nothing
Add-on assignments that appear unnecessary after effective-permission and business-requirement checks. Distinguish technical unassignment from a reduction in purchased licences.
Which seats are over-specified
Full-CRM users with no authored records across standard objects, and feature licences enabled for users doing none of the corresponding work. Advisory findings, labelled advisory, priced separately.
What is genuinely recoverable
Permitted seat reductions limited by the minimum retained quantity, with the blocked remainder shown separately, and estimated repricing clawback subtracted — a figure your account executive cannot dismantle.
Every finding ships with its evidence attached: the login rows, the licence record, the set difference. A CFO discards an entire analysis over one wrong name, so each line has to be defensible on its own.
- Which seats can be removed outright at the next renewal
- Which add-on assignments can be removed after confirming user impact and contract treatment
- Which users are genuine downgrade candidates, and which only look like it
- Which light-user workflows merit a portal assessment, subject to licensing requirements
- What the recoverable figure is after the contract floor and repricing effects
- Which findings are certain and which are advisory — stated, not blended
What you receive
The seat ledger
Every seat with its licence class, cost, human and machine login history, and its classification: keep, downgrade, release, or portal.
Evidence pack
Each finding with the underlying rows attached, including the machine-login contrast panel on every dormancy claim. Exportable, and built to be handed to procurement.
The honest number
Potential recovery after retaining the contractual minimum quantity, net of estimated repricing, and captioned with the renewal date it becomes available.
Evidence on every line
Stage 00 is a 30-minute qualifying call at no cost. If the timing or the estate does not justify an engagement, we say so on that call.