Renewals are designed to be approved, not read. One figure, a year-on-year comparison, a discount if you commit to three years. The line items are available and almost nobody asks for them, because the total is within tolerance and everyone is busy.
Ask for them. An afternoon with the breakdown routinely finds more than the discount would have.
The discount on offer was 6% — about £18,000. The dormant seats and the stale capacity together were £109,832, and neither was on the table until somebody asked for the breakdown.
Five questions the breakdown answers
| Question | Where to look | What people usually find |
|---|---|---|
| What fraction is viewers? | Seat counts by licence tier. | The majority, often by a wide margin |
| How many of those seats were used? | Cross-reference seats against logins in the last quarter. | A meaningful share never logged in |
| What are we paying for capacity we do not use? | Reserved capacity, premium tiers, node hours. | Sized during a peak that has not recurred |
| What is bundled that we do not use? | Data prep, catalogue, ML add-ons folded into the tier. | One or two, usually replaced by something else years ago |
| What does leaving cost? | Export formats, notice period, whether your data is retrievable. | Rarely asked before it matters |
The second row is the one to run first, and it takes a single query on each side. Seats are allocated by process — a joiner gets one, a leaver's is reclaimed slowly, a project team gets a block — and processes drift.
Nobody is charged for a dashboard. They are charged for a person who might open one. That distinction is the whole pricing model, and it is the one to interrogate.
sankey-beta 1304 seats,Logged in last quarter,761 1304 seats,Dormant,543 Dormant,Viewer tier,478 Dormant,Explorer tier,62 Dormant,Creator tier,3
The question under the question
Seat counts are a symptom. The thing to establish is what the contract is actually buying, because three quite different things are usually bundled into one number.
| What you are buying | Is it worth paying for? | Notes |
|---|---|---|
| Authoring | Yes | People who build and maintain content are a real, skilled, small group. Paying per seat here is defensible. |
| Hosting and compute | Depends | Real cost to the vendor — but only because they hold your data. If it sat in your own storage, this line would not exist. |
| Reading | No | A viewer generates no content and costs the vendor a session. This is the line with no defensible basis, and it is usually the largest. |
flowchart LR
T["One renewal total"] --> A["Authoring
small group, real value"]
T --> H["Hosting and compute
exists because they hold your data"]
T --> R["Reading
the largest line"]
R --> Q["What is this recovering,
if the data is yours?"]
style A fill:#e6f7f2,stroke:#9fd8c9
style R fill:#fdecec,stroke:#eab6b6
style Q fill:#eef1ff,stroke:#c7cdf2
Do the shelfware count first
One number changes this conversation more than any other: how many of your dashboards were opened by nobody in the last ninety days. In most estates it is between half and four fifths.
Why it matters at renewal specifically
Seats are frequently allocated to the owners of unopened dashboards, and capacity is often sized against a refresh schedule that feeds them. Both lines shrink once you know the number, and the count takes an afternoon on the tool you already have. How to run it →
The exit question, asked while you still have leverage
Ask it at renewal rather than at migration, because at renewal it is a negotiating question and at migration it is a discovery.
| Ask | A good answer | A bad answer |
|---|---|---|
| Where does our data physically live? | Your storage account, in a format you can read. | Our platform, in our format |
| If we left, what could we take? | Everything, in an open format, today. | Exports, per artefact, manually |
| Can another tool read our extracts? | Yes — Parquet, ORC, something standard. | Only ours |
| Are the metric definitions portable? | They are text; take them. | Inside the workbooks |
| What notice do we owe? | A stated period. | Auto-renewal with a 90-day window you have already missed |
The fourth row is the one people underestimate. If your definitions live inside binary workbooks, they are not portable in any practical sense — every one has to be reopened and re-read by a person. That is the lock-in, far more than the file format. And it is the same property that lets them drift →
Leaving means re-extracting everything from source and rebuilding every definition by reading workbooks. The cost is high enough that renewal is not really a decision.
Extracts are Parquet in your own bucket; definitions are readable, versioned documents. Leaving costs you the tool, not the asset — which is also what makes the renewal a real negotiation.
An afternoon's work, in order
- Request the line items. Seats by tier, capacity, add-ons. You are entitled to them and asking is routine.
- Count logins in the last quarter, by tier. Compare against seats. This is the number with the largest immediate effect on the negotiation.
- Count dashboards opened by nobody in ninety days. It reframes both the seat count and the capacity line.
- Count distinct definitions of your top five metrics. Not a cost line — it tells you how much of what you are renewing is actually trusted.
- Write down what leaving would cost. Do this before you need it, because the answer determines how much of a negotiation you are actually in.
What we would say against ourselves
- Switching is not free. You pay for both estates for a quarter, and the definitions conversation takes weeks. The overlap, honestly →
- The cost moves rather than vanishing. Compute lands on your cloud bill instead of a licence line. It is usually far smaller, and it is a line you now own and have to watch.
- If your estate is small and your tool is working, renew. Fifty users and a handful of dashboards that everyone trusts is not a problem looking for a migration. The arguments here scale with the size of the reading population.