Cognos licensing, the role mix is the bill
Cognos Analytics runs several license metrics at once: role based Authorized Users, capacity in Processor Value Units, and, increasingly, the Cloud Pak for Data repackaging in Virtual Processor Cores. The cost rarely tracks actual use, and the gap lives in three places: the role mix, the dormant users, and the conversion quote.
Prepared by Redress Compliance · August 6, 2026 · IBM licensing advisory. Based on 20 to 30 analytics engagements advised 2024 to 2026.
Executive summary
Cognos licenses two ways simultaneously: role based Authorized Users, priced per named person by capability, from Analytics Administrator through authoring roles down to view only Analytics User, and capacity metrics, Processor Value Units on the servers.
The role ladder is the cost driver on the user side, because the role you assign decides the price, and role assignment follows convenience unless someone owns it.
The waste is dormant and structural. Across our analytics engagements, dormant or duplicate Authorized Users made up 12 to 30 percent of the named population at first review, licenses renewing annually for people who left, changed roles, or were counted twice across environments.
Named user hygiene is the cheapest recurring recovery in the IBM analytics estate.
Cloud Pak for Data repackages Cognos into a capacity model priced in Virtual Processor Cores, and the conversion is a repricing event: quotes in our reviews were built on peak capacity, running 15 to 35 percent above measured concurrency.
Whether the bundle helps or hurts depends entirely on your real concurrency, which is measurable before signature and rarely measured.
The sub capacity condition follows Cognos everywhere PVUs do: the discount requires the IBM License Metric Tool deployed and reporting correctly, and in roughly 1 estate in 3 the savings were forfeited because ILMT was not reporting.
On the analytics estate as everywhere in the IBM portfolio, the measurement discipline is the discount.
How Cognos is licensed, both meters at once
| Metric | What it counts | Where it leaks |
|---|---|---|
| Authorized User, by role | Named people, priced by assigned capability from administrator to view only | Role inflation: authoring roles assigned to consumers, and dormant names never retired |
| Capacity, PVU | Processing power on the Cognos servers, sub capacity where ILMT earns it | The full capacity fallback when the measurement tooling lapses |
| Cloud Pak for Data, VPC | Virtual Processor Cores in the bundled capacity model | Conversions quoted on peak rather than measured concurrency |
The role ladder is a price ladder. Administrator and authoring roles carry multiples of the view only rate, and role assignment drifts upward: report consumers accumulate authoring roles from projects, templates, and helpful admins.
Nobody experiences a downgrade, so the mix only ever climbs, until a review prices the drift.
The named population, roles and ghosts
The Authorized User bill divides into two recoverable layers.
The first is existence: 12 to 30 percent of named populations in our reviews were dormant or duplicated, leavers never removed, role changers keeping analytics access they stopped using, and the same person counted in production and development environments.
The second is classification: active users holding authoring or administration roles whose usage history shows consumption only.
The reconciliation is the same discipline as every named metric: the license table joined to actual activity, per user, per role, per quarter. Users with no activity leave the count; users whose activity fits a lighter role move down the ladder; and the renewal prices what remains.
Estates that ran it annually held the recovered position; estates that ran it once watched the drift return.
- Percentile standing for your exact deal size and industry, from real closed transactions
- Scenario simulation before the call: test alternative terms and see the financial impact of each
- A negotiation playbook, talking points, and a two page executive brief on day one
The Cloud Pak for Data conversion, measured concurrency or peak fiction
IBM's strategic motion moves Cognos into Cloud Pak for Data, repriced in Virtual Processor Cores inside the bundled capacity model.
The conversion can genuinely help estates with broad, spiky analytics usage, the pooled capacity absorbs variation named users cannot, and it can genuinely hurt estates whose usage is narrow and steady, which were cheaper on the metrics they had.
The Cloud Pak strategy guide works the decision, and the Cloud Pak licensing guide carries the VPC mechanics.
What decides the outcome is one measurement: real concurrency, measured across a representative period, against the capacity the quote assumes.
Conversion quotes in our reviews were built on peak capacity, 15 to 35 percent above measured concurrency, because peak is the number the sizing conversation naturally produces.
The buyer who arrives with the measured curve buys the capacity the estate uses; everyone else buys the capacity the demo needed.
The IBM analytics licensing playbook
The Cognos role reconciliation method, the Cloud Pak conversion math, the concurrency measurement approach, and the negotiation sequence for the analytics estate.
Get the white paper →The ILMT condition, again
Wherever Cognos runs on PVU metrics, the sub capacity discount is conditional on the IBM License Metric Tool deployed, scanning, and reported, the same four leg chain that governs the whole IBM estate, covered in the sub capacity and ILMT guide.
The analytics servers are where the condition most often silently fails: they belong to the BI team, not the infrastructure team, and the agent nobody deployed becomes the full capacity assessment nobody budgeted.
In roughly 1 estate in 3 we reviewed, sub capacity savings on the analytics estate were forfeited to exactly this gap.
And the finding compounds at audit, where it leads the claim for the reasons the audit penalties guide works in full: it converts compliant looking estates to full capacity pricing without proving a single over deployment.
What we saw across analytics engagements, 2024 to 2026
Across roughly 20 to 30 IBM analytics engagements Morten Andersen advised between 2024 and 2026, Cognos cost rarely tracked actual use, and the role mix was the biggest swing factor:
Leavers, role changers, and duplicates renewing annually in the Authorized User population at first review.
Cloud Pak for Data quotes built on peak capacity against what measured concurrency actually supported.
The unifying pattern was ownership: the analytics estate sat between the BI team, which owns the users, and the licensing function, which owns the contracts, and the reconciliation belonged to neither.
The estates that recovered the most assigned the baseline, users, roles, capacity, and ILMT coverage, to one owner with a renewal calendar, which is the entire IBM playbook in one sentence.
Your first five moves
- Join the license table to activity: every Authorized User, their role, and their last quarter's actual usage. Retire the ghosts, downgrade the drifted.
- Verify ILMT coverage on the analytics servers specifically; they are the hosts the infrastructure sweep most often misses.
- Measure real concurrency before any Cloud Pak conversation, across a representative period, and quote against the curve rather than the peak.
- Price the conversion both ways: the measured estate on current metrics against the VPC bundle, with the strategic pressure discounted from both sides.
- Take the reconciled baseline into the renewal, where the role mix, the dormant recovery, and the capacity position negotiate as one package. The IBM practice runs it with you, on your side of the table.
Frequently asked questions
How is IBM Cognos Analytics licensed?
Two ways at once: role based Authorized Users, priced per named person by assigned capability from Analytics Administrator down to view only Analytics User, and capacity metrics, Processor Value Units on the servers, with sub capacity conditional on ILMT.
Cloud Pak for Data additionally repackages Cognos into Virtual Processor Core capacity.
What drives Cognos licensing cost the most?
The role mix. Each role carries a different rate, authoring and administration at multiples of view only consumption, and role assignment drifts upward over time.
Reconciling assigned roles against actual usage, and retiring the 12 to 30 percent of named users who are dormant or duplicated, is where most of the recoverable cost sits.
Should we move Cognos into Cloud Pak for Data?
It depends on measured concurrency. The pooled VPC capacity genuinely helps broad, spiky analytics usage and genuinely hurts narrow, steady estates.
Conversion quotes in our reviews ran 15 to 35 percent above measured concurrency because they priced peak capacity, so the measurement, run before the conversation, decides the answer.
Does Cognos need ILMT for sub capacity licensing?
Yes, wherever it runs on PVU metrics: the sub capacity discount requires ILMT deployed, scanning, and reported, and analytics servers are where the condition most often fails because they sit outside the infrastructure team's sweep.
One estate in three in our reviews forfeited the savings to exactly that gap.
How much Cognos spend is typically recoverable?
The dormant user layer alone ran 12 to 30 percent of the named population, and role downgrades against usage evidence add to it. On the capacity side, correcting a peak sized Cloud Pak quote to measured concurrency recovered 15 to 35 percent.
The reconciled baseline, users, roles, capacity, and coverage, is what converts the findings into renewal outcomes.
When should the Cognos license position be reviewed?
Annually against activity, and always before two events: the renewal, where the reconciled count and role mix become the negotiating position, and any Cloud Pak for Data conversion, where every error in the current estate reprices into the bundle at the new meter's rates.
