License Optimization
The License Optimization module finds paid licences that nobody uses, estimates what they cost, and tracks what removing them saves. It covers two tools:
- Atlassian: paid Atlassian Cloud seats, per product.
- Microsoft 365: licences on disabled, idle or never-used accounts, add-ons nobody opens, seats bought but never assigned, and admin roles nobody uses.
Open it from the sidebar: Modules → License Optimization. Choose the tool at the top of the screen (Atlassian or Microsoft 365); every tab below it, and the settings, belong to that tool.
The module only recommends. Each recommendation is reviewed by a person and carried out in Atlassian or Microsoft 365. A later evaluation confirms that the seat is gone, and only then does the saving count as realised.
Prerequisites
| Tool | Integration | Notes |
|---|---|---|
| Atlassian | Atlassian Users | Per-product activity comes from the integration. A paid user without it is counted, but the unused product role rule cannot judge them; the screen says how many such users there are. |
| Microsoft 365 | Azure Users | Unassigned seats and the usage signals need a scan that reported the directory's subscriptions. The screen lists what could not be read and why. |
For Microsoft 365, what the module can judge depends on what the Azure Users app registration is allowed to read. The evaluation warns when:
- the usage reports could not be read (the app needs
Reports.Read.All): add-ons judged by service use are not raised; - the usage reports hide user names (the Microsoft 365 setting "Display concealed user, group, and site names in all reports" is on): mail, Teams, OneDrive, SharePoint and Office app use cannot be tied to users;
- the sign-in log could not be read (
AuditLog.Read.All, Entra ID P1): app sign-ins are not used; - the directory audit log could not be read (
AuditLog.Read.All): admin activity is judged by admin portal sign-ins alone; - the subscriptions could not be read: bought and unassigned seats are not shown.
Users the directory no longer reports are left out of the evaluation, and the screen says how many.
How the rules work
Both tools follow the same pattern:
- People who should never be recommended are left out first (see Who is never recommended).
- The rules are checked in a fixed order. The first rule that applies to a person wins.
- A whole-user finding takes every paid licence the person holds, so a person with one gets no per-product findings: they would count the same seats twice.
- A rule you switch off is skipped, and the next one applies.
- Nothing is guessed. Where the data needed to judge a rule is missing, the rule is not raised.
Days are counted in whole days. "Idle" means the time since the person's newest activity; someone who never did anything has been idle since they joined (or since the account was created), not forever.
Atlassian rules
Only users with a paid seat are evaluated. Suspended and deactivated accounts are not billed, so they are out of scope.
| Order | Recommendation | Raised when | Default threshold |
|---|---|---|---|
| 1 | Never-used seat (whole user) | The user never used any product, and joined at least the threshold ago. A user whose join date is unknown is raised at once. | 30 days |
| 2 | Idle external user (whole user) | An external user idle on everything for the external threshold. | 60 days |
| 3 | Inactive user (whole user) | A user idle on everything for the inactive threshold. | 90 days |
| 4 | Unused product role (one per product) | The user uses something, but this paid product has been unused for its threshold. A product never used counts once the user joined at least the threshold ago. Needs per-product activity. | 90 days, can be set per product |
The user's newest activity is the latest of their overall last-active time and every product's last activity.
Microsoft 365 rules
The licence rules apply to users holding at least one paid licence:
| Order | Recommendation | Raised when | Default threshold |
|---|---|---|---|
| 1 | Disabled but licensed (whole user) | The account is switched off but still holds paid licences, and has been off for at least the grace days. While it is within them nothing else is raised for it (a leaver's mailbox is often kept a while). If the module has not yet seen when the account was switched off, it is raised only when the grace is 0. | 30 days |
| 2 | Never signed in (whole user) | No sign-in or use of any kind, and the account was created at least the threshold ago (or its creation date is unknown). | 30 days |
| 3 | Idle guest (whole user) | A guest account idle for the guest threshold. | 60 days |
| 4 | Inactive user (whole user) | No activity for the inactive threshold. | 90 days |
| 5 | Unused add-on (one per licence) | An add-on licence whose usage signals show no use for its threshold. | 90 days, can be set per licence |
Two further rules stand apart from the order above:
| Recommendation | Raised when | Default threshold |
|---|---|---|
| Unused admin role | A holder of an admin role did nothing an admin does (no change in the directory audit log, no sign-in to an admin portal app) for the role's threshold. Raised whether or not the person holds a licence. It saves nothing and is about risk; roles marked privileged are flagged as such. | 90 days, can be set per role |
| Unassigned seats | Seats of a billed subscription that nobody is assigned. One finding per subscription and directory, worth the unassigned seats times the price. | On or off only |
Activity. For the idle and inactive rules, activity is the newest of the user's interactive sign-in, any Microsoft 365 service use the usage reports show, and any app sign-in. Non-interactive sign-ins (an app signing in on its own, such as a phone syncing mail) count only when Count non-interactive sign-ins as activity is on. It is off by default, so a leaver whose phone still syncs does not look active. Never signed in always considers every kind of sign-in.
Add-ons. An add-on is judged only by its usage signals:
- the usage report services: Exchange (mail), OneDrive, SharePoint, Teams, Skype for Business, Viva Engage (Yammer), Office apps, Office desktop apps;
- sign-ins to a named app, from the sign-in log.
Each licence has a default set of signals, which you can replace per licence. An add-on is judged only when every one of its signals could be read, and only once the licence has been assigned for at least its threshold. A signal that shows no use at all proves non-use only as far back as its data reaches: 180 days for the usage reports, and back to the start of the collected history for app sign-ins. Suites (such as Business Premium or E3) are never raised as unused add-ons: they are only ever taken away whole, by a whole-user rule. A licence with no usage signals is never raised as an add-on either.
Admin roles. A role is judged unused only once the collected activity history reaches back the whole threshold and the person has held the role that long; until then it is reported as not judged yet. The history grows with every scan.
Who is never recommended
| Exemption | Atlassian | Microsoft 365 |
|---|---|---|
| Allowlist | Emails or Atlassian account ids. Left out of every rule. | Emails, user principal names or object ids. Left out of every rule, admin roles included. |
| Grace period | Users who joined less than this long ago. Default 30 days; 0 turns it off. | Accounts created less than this long ago. Default 30 days; 0 turns it off. Left out of every rule, admin roles included. |
| Admins | On by default: anyone holding a site, organisation or product admin role is left out of every rule. | On by default (Admins' licenses): admin role holders keep their licences, but their admin roles are still judged. |
The allowlist is compared without regard to case, holds at most 1,000 entries of up to 256 characters each, and is one entry per line in the settings.
Settings
Settings opens the settings of the selected tool. A tenant that never saved settings uses the defaults. Saving does not recalculate: the screen says the settings changed since the evaluation it shows, and Recalculate applies them.
Atlassian settings
| Section | Setting | Default |
|---|---|---|
| Recommendations | Never-used seats (on/off, days) | On, 30 |
| Recommendations | Idle external users (on/off, days) | On, 60 |
| Recommendations | Inactive users (on/off, days) | On, 90 |
| Recommendations | Unused product roles (on/off, days) | On, 90 |
| Never recommend | Grace period for new users (days) | 30 |
| Never recommend | Admins | On |
| Never recommend | Allowlist | Empty |
| Products | Price per product (USD per user per month) and Role threshold (days) per product | See below; thresholds empty (use the default) |
The default prices are approximate Standard-plan list prices per user per month in USD, averaged over annual billing tiers. Enter what your invoice says.
| Product | Default price (USD / user / month) |
|---|---|
| Jira | 7.91 |
| Jira Service Management | 20.00 |
| Confluence | 5.42 |
| Customer Service Management | 20.00 |
| Jira Work Management | 0.00 |
| Jira Work Management (new) | 0.00 |
| Jira Product Discovery | 0.00 |
| Compass | 0.00 |
| Opsgenie | 0.00 |
Microsoft 365 settings
| Section | Setting | Default |
|---|---|---|
| Recommendations | Disabled but licensed (on/off, days) | On, 30 (0 means at once) |
| Recommendations | Never signed in (on/off, days) | On, 30 |
| Recommendations | Idle guests (on/off, days) | On, 60 |
| Recommendations | Inactive users (on/off, days) | On, 90 |
| Recommendations | Unused add-ons (on/off, days) | On, 90 |
| Recommendations | Unused admin roles (on/off, days) | On, 90 |
| Recommendations | Unassigned seats | On |
| Activity | Count non-interactive sign-ins as activity | Off |
| Add-ons | Per add-on threshold and usage signals | Default threshold; catalogue signals |
| Admin roles | Per role threshold (roles appear once an evaluation has seen them); Admin portal apps | Default threshold; the admin centres and admin tools (Azure Portal, Microsoft Entra admin center, Microsoft 365 Admin Center and others) |
| Never recommend | Grace period for new accounts (days) | 30 |
| Never recommend | Admins' licenses | On |
| Never recommend | Allowlist | Empty |
| Prices | Price per licence (USD per user per month) | The catalogue price where one is known |
A licence the catalogue has no price for is listed as Price missing and counts at 0 until you enter one.
Ranges
| Value | Range |
|---|---|
| Rule thresholds | 1 to 3,650 days |
| Grace period, Disabled but licensed days | 0 to 3,650 days |
| Prices | 0 to 10,000 USD per month |
| Allowlist | Up to 1,000 entries, each up to 256 characters |
| Microsoft 365 per-licence and per-role lists, admin portal apps | Up to 500 entries each; up to 20 usage signals per licence |
Running an evaluation
Evaluations run on the server, per tool, on request. Click Recalculate (or Run the first evaluation). The evaluation reads every user the integration found and can take a while; the page shows its progress and updates when it completes.
- Only one evaluation per tool and tenant is queued or running at a time.
- The screen shows the newest completed evaluation. A failed evaluation leaves no partial results and the previous one stays on screen.
- An evaluation that stops reporting progress for 15 minutes is marked as failed.
- Velum Edge keeps the findings of the newest evaluation and the totals of the last 30 for the trends.
The screen
| Tab | Contents |
|---|---|
| Overview | Estimated spend, Open savings (recommendations nobody has decided on yet, and snoozes that ran out), Realized savings (marked done and confirmed gone by a later evaluation), Done, still assigned (marked done, but the latest evaluation still finds it) and Ignored. Per-product use (used within the threshold, idle past it, never used, no activity data), paid users by days since their last activity, admin roles (Microsoft 365) and the potential and realised savings over the last evaluations. |
| Recommendations | Every finding with the user, the recommendation, the monthly saving, last activity, origin and review state. Filter by Finding, Origin (Internal or External for Atlassian, Member or Guest for Microsoft 365) and Review, search, and sort. Click a row for the detail. |
| Audit log | Every review decision: when, the decision, its outcome, the user it concerned, the recommendation, the saving, who decided and the note. Filter by decision. |
Every figure is an estimate from the prices in the settings. Estimated spend for Microsoft 365 includes bought seats nobody is assigned, since they are billed too. The potential saving of a whole-user finding is the price of every paid licence the person holds; of a per-product finding, that product's price; of unassigned seats, the seats times the price. Unused admin role findings are worth nothing.
Reviewing recommendations
Select recommendations on the Recommendations tab (or open one) and choose a decision. You can add a note of up to 2,000 characters, and every decision is recorded in the audit log.
| Decision | Effect |
|---|---|
| Mark done… | You removed the seat in the tool. It counts as realised once an evaluation confirms it is gone. |
| Snooze… | Put off for 1 week, 30 days, 90 days, 6 months or 1 year. After that it is open again. Snoozing again moves the date. |
| Ignore… | Deliberately kept. The finding stays visible but no longer counts towards potential savings. |
| Reopen | Back to the open list. |
A decision is stored against the finding (the person, the kind of finding and the product), so it carries over to the same finding in later evaluations. Only findings of the newest evaluation can be decided on; a finding already in the chosen state is skipped, and the console reports what was skipped.
Realised savings
When an evaluation completes, it checks every finding marked done that is not yet realised:
- People's licences: the finding is realised when the evaluation no longer raises it and what it was about is gone: for a whole-user finding none of its licences is still held, for a per-product finding that product is no longer held. An account the integration no longer reports holds nothing. A finding that disappears because the person started using the product again saves nothing and is not realised.
- Unassigned seats: realised when fewer seats are bought now than when the finding was marked done. The saving is prorated to the seats actually given back, up to the number that were unassigned. Assigning the seats to people instead uses them, which saves nothing.
Realised savings add up over time. A realised finding that comes back later (the seat was given again) is a new, open finding.
Limits
| Limit | Value |
|---|---|
| Findings per decision request | 1,000 |
| Note length | 2,000 characters |
| Snooze | 1 to 365 days |
| Rows per results page (API) | 500 |
| Evaluations kept for the trends | 30 |
Permissions
| Task | Permission |
|---|---|
| View the module, settings, recommendations and audit log | Asset Management read |
| Save settings, request an evaluation, decide on recommendations | Asset Management write |
The module must also be switched on for the tenant; see Modules.