The licensed knowledge worker count ran 15 to 35 percent above measured active users
The bundle mixes calling, meetings, messaging, and contact centre behind a single headcount, and most buyers use a fraction of it. The count sounds like an administrative input and it is the entire commercial negotiation.
Prepared by Redress Compliance · August 16, 2026 · Cisco advisory. 20 to 30 Cisco collaboration agreements benchmarked, 2024 to 2025.
Executive summary
The knowledge worker count was set to total headcount in most estates, rather than to users who actually run Webex, which put the licensed count 15 to 35 percent above measured active users.
Premium meeting and calling tiers were bought for 20 to 40 percent of users who only needed messaging. The bundle prices uniformly and the population inside it does not use it uniformly.
Contact centre seats were bundled at a flat tier, missing 10 to 25 percent of savings available from agent right sizing against actual concurrency.
Scope the bundle to real adoption and the deal shrinks. Accept the standard knowledge worker definition and you pay for users who never log in, for the length of the term.
What the bundle contains, and who actually uses it
A Cisco collaboration agreement bundles calling, meetings, messaging, and contact centre behind one knowledge worker count. The components have very different adoption curves and the price does not reflect that.
| Component | Typical adoption | Where the waste sits |
|---|---|---|
| Messaging | Broad, close to the full population | Little, this is what most users actually need |
| Meetings | Partial, concentrated in knowledge roles | Premium tiers bought for users who only message |
| Calling | Partial, often role specific | Assigned by default rather than by need |
| Contact centre | Narrow, a defined agent population | Flat tier bundling instead of agent right sizing |
The knowledge worker definition is the negotiation, and it is presented as an administrative detail. Cisco proposes a count, the count is usually total headcount because that is the number the buyer can produce quickly, and everything downstream prices off it. Producing a measured active user figure instead is not a concession the vendor grants; it is a number you assemble before the conversation, and in the estates benchmarked it was 15 to 35 percent lower than the count that would otherwise have been signed.
Uniform pricing over a non uniform population
Cisco prices collaboration on a headcount that sounds simple and ends up high. The mechanism is not complicated: the bundle carries calling, meetings, messaging, and contact centre at a single per user rate, and the population it is applied to uses those four things at very different rates. Messaging approaches full adoption in most organisations. Meetings and calling concentrate in particular roles. Contact centre is a narrow, defined group. A uniform price over a non uniform population is a structurally favourable arrangement for the seller, and the count is where it is set.
What makes it persist is that the knowledge worker count is the easiest number in the deal to produce badly. Total headcount is available instantly from HR, it is defensible as a definition, and nobody has to do any work to obtain it. Measured active Webex users requires pulling telemetry, agreeing what active means, and reconciling it against the licence position, which is a week of effort nobody has budgeted during a renewal. So the fast number wins, and in the estates benchmarked it ran 15 to 35 percent above the accurate one.
The tier question sits on top and compounds it. Premium meeting and calling tiers were bought for 20 to 40 percent of users who only needed messaging, which is the same uniformity problem applied a level down. Once the count is set to headcount, tiering to the highest common denominator follows naturally, because differentiating would require the per user analysis that was skipped when the count was produced. The two errors are causally linked rather than independent.
Contact centre deserves separate handling because it prices differently and is usually bundled at a flat tier. Agent right sizing against actual concurrency rather than named agents recovered 10 to 25 percent in the agreements reviewed, and it is a distinct exercise from the knowledge worker count, requiring concurrency data rather than telemetry. Run it separately, because folding it into the same conversation lets a flat tier survive on the grounds that the bundle simplifies things. The ELA structure around all of this sits in the ELA renewal playbook, and the wider library in the Cisco practice.
- Usage exports analysed: inactive accounts, tier right sizing, per user reassignment
- Your quote benchmarked against real closed Cisco deals
- Every risky clause flagged with the exact quote, the page, and the replacement language
Where the count comes down
- Produce a measured active user figure before the count is agreed, since total headcount is the fast number and it ran 15 to 35 percent high in the estates benchmarked.
- Define active explicitly, because the definition decides the number and an undefined term defaults to the vendor reading of it.
- Tier by role rather than uniformly, since 20 to 40 percent of users on premium meeting and calling tiers needed only messaging.
- Right size contact centre agents against concurrency, not against named agents, which recovered 10 to 25 percent separately from the knowledge worker work.
- Treat the count as the commercial negotiation, because everything downstream, tiering, contact centre, and term value, prices off it.
- Do the measurement before the renewal opens, since it takes a week of effort and no renewal timetable includes it by default.
What the collaboration agreements showed, 2024 to 2025
Across roughly 20 to 30 Cisco collaboration agreements benchmarked:
How far the licensed knowledge worker count exceeded measured active users, because the count was set to total headcount.
Users on premium meeting and calling tiers who needed only messaging, which follows directly from setting the count uniformly.
Contact centre seats were bundled at a flat tier, missing 10 to 25 percent of savings available from agent right sizing against actual concurrency rather than named agents.
Scope the bundle to real adoption and the deal shrinks. Accept the standard knowledge worker definition and you pay for users who never log in, for the full length of the term.
Your first five moves
- Pull Webex telemetry and produce a measured active user count before any number is agreed with Cisco.
- Define active in writing, because an undefined term will default to the reading that produces the larger figure.
- Segment the population by what it uses, separating messaging only users from those genuinely needing premium meetings and calling.
- Size contact centre agents against concurrency, as a separate exercise from the knowledge worker count.
- Take the count into the negotiation as your number, not as a figure to be corrected. The Cisco practice runs the measurement with you.
Frequently asked questions
How does Cisco price collaboration?
On a knowledge worker count covering a bundle of calling, meetings, messaging, and contact centre at a single per user rate. The components have very different adoption curves, so a uniform price over a non uniform population favours the seller structurally.
How inflated is the typical count?
By 15 to 35 percent against measured active users. In most estates the knowledge worker count was set to total headcount rather than to people who actually run Webex, because headcount is the number a buyer can produce instantly.
Why does total headcount win as the definition?
Because it is available immediately from HR and requires no work. Measured active users requires telemetry, an agreed definition of active, and reconciliation against the licence position, which is a week of effort no renewal timetable includes.
How much premium tiering is unnecessary?
Premium meeting and calling tiers were bought for 20 to 40 percent of users who only needed messaging. That follows directly from the count being set uniformly, since differentiating tiers requires the per user analysis that was skipped.
Are the count error and the tier error related?
Causally, yes. Once the count is set to headcount, tiering to the highest common denominator follows naturally, because differentiating would need the analysis that was avoided when the count was produced. They are one decision with two costs.
How should contact centre be sized?
Against actual concurrency rather than named agents. Flat tier bundling missed 10 to 25 percent of available savings in the agreements reviewed, and it is a separate exercise requiring concurrency data rather than usage telemetry.
Should contact centre be negotiated in the same conversation?
Run it separately. Folding it into the knowledge worker discussion lets a flat tier survive on the argument that the bundle simplifies administration, which is true and is not a pricing justification.
What does active actually mean?
Whatever you define it as, in writing, before the count is agreed. Left undefined it defaults to the reading that produces the larger number, which is why the definition is worth more attention than the rate.
When should the measurement happen?
Before the renewal opens. It takes about a week of effort, no renewal timetable includes it by default, and a count produced during the negotiation arrives too late to be the anchor.
Is the count really the whole negotiation?
Close to it. Tiering, contact centre sizing, and total term value all price off the knowledge worker count, so a number agreed casually at the start determines outcomes that are then argued individually and unsuccessfully later.
5 Tips for Negotiating with Cisco
Commit to reality not history, cap the true forward first (3 to 4 percent vs the 5 to 8 opener, worth $80k to $160k on a $2M baseline), price Splunk standalone before it enters the bundle, attack the support line, and use the late-July fiscal year.
