Pricing the Queue: A Week-Long Tariff Review for UK Data Teams
Photo: 奔流沙, CC BY-SA 3.0, via Wikimedia Commons
Most UK data teams can produce a platform bill to the penny. Far fewer can say what it costs to answer a routine question about customer churn, stock availability or pipeline conversion. That gap matters in 2026 because data budgets are being scrutinised less for their size than for their yield. A team that cannot price its own output is poorly placed to defend it.
The remedy is not a grand activity-based costing programme. It is a short, disciplined review of the queue that already exists. Over one working week, a small group can sample common requests, trace the people and elapsed time involved, and produce a simple tariff for three kinds of answer: self-serve, analyst-led and bespoke. The point is not to charge colleagues for asking questions. It is to make the cost of an answer visible enough to support decisions about automation, retirement and demand management.
Why the question is the right unit
Infrastructure spend is a poor proxy for value because it is several steps removed from the business outcome. A warehouse bill includes capacity for queries that may never be run, dashboards that may never be opened and pipelines that may serve no current decision. The request, by contrast, is where business intent meets data work. It has a requester, a purpose, a delivery route and a cost.
Treating the request as the unit of analysis also exposes hidden effort. A simple extract may consume twenty minutes of analyst time but three days of waiting for clarification. A dashboard change may be trivial to build but expensive to govern because it alters a board metric. Unit economics gives the team a language for these differences that finance and operations colleagues already understand.
The review, day by day
Monday: choose the sample and define the answer
Start with a defined window, such as the previous eight to twelve weeks, and draw a sample of requests from the service desk, inbox, planning sessions and informal channels. Include the unticketed work that consumes time but never appears in a queue. Aim for breadth rather than statistical purity: routine extracts, recurring reports, ad hoc analysis, dashboard changes and one or two genuinely novel questions.
For each item, define what counted as the answer. This sounds pedantic until the team tries it. Was the deliverable a spreadsheet, a revised metric definition, a meeting in which the number was explained, or all three? The definition should mark the point at which the requester had what they needed, not the point at which the team stopped polishing.
Tuesday: trace the work
For each sampled request, reconstruct the route from arrival to answer. Record the roles involved, active effort, waiting time, hand-offs and rework. Waiting is often the largest component and the least visible. It includes time spent clarifying the question, waiting for access, waiting for a subject-matter expert, and waiting for the requester to respond.
A simple timeline is enough. Avoid turning this into a time-and-motion study. The aim is a defensible estimate, not a forensic reconstruction. Where memory is uncertain, ask the people involved to mark whether effort was under an hour, half a day or more. Precision can come later if the tariff proves useful.
Wednesday: convert time into cost
Use a fully loaded hourly rate for each role, including salary, employer costs, tools and an allocation of management overhead. Be explicit about what the rate includes and what it excludes. A tariff that carries every central cost will be disputed; one that carries none will be ignored. A reasonable middle path is to include direct labour, contracted support and a simple overhead percentage agreed with finance.
Do not allocate platform spend per query. That way lies false precision. Instead, present platform cost separately as a fixed capacity charge and use the tariff to show the variable human cost of answering. The two numbers tell different stories and should not be blended into one misleading figure.
Thursday: set the tariff bands
Group the sampled requests into three bands. Self-serve covers answers the business can obtain from an existing governed tool without analyst intervention. Analyst-led covers requests that need a defined piece of work, usually within an existing dataset or dashboard. Bespoke covers novel analysis, new data, complex joins or work that requires significant interpretation.
For each band, state a typical elapsed time, a typical active effort range and the management action the band should trigger. The tariff should be simple enough to quote in a conversation and stable enough to survive the next reorganisation.
| Band | Typical signal | Management action | | Self-serve | Existing governed dataset, standard question, no new logic | Redirect to the catalogue or tool; track repeat demand | | Analyst-led | Known dataset, defined output, moderate interpretation | Schedule against capacity; consider reusable template | | Bespoke | New data, contested definition, significant judgement | Require a named decision owner and a stated value at stake |
Friday: decide what changes
The final day is not for publishing a price list. It is for deciding what to automate, retire or push back. Automation candidates are repeatable analyst-led requests that could become governed self-serve products. Retirement candidates are reports nobody uses or questions already answered by an existing tool. Push-back candidates are bespoke requests with no named decision owner or no discernible consequence.
Invite one finance colleague and one senior business stakeholder to the Friday session. Their presence prevents the tariff from becoming a data-team artefact and forces the conversation towards decisions rather than definitions.
Using the tariff without turning it into a toll booth
A tariff can easily become a barrier. If business colleagues experience it as a charge for curiosity, they will route around the data team and build shadow answers in spreadsheets. That is worse than an unmeasured queue.
Use the tariff as a routing and prioritisation instrument, not an invoice. Publish the bands, explain what makes a request self-serve or bespoke, and give people a clear path to challenge a classification. Review the tariff quarterly, because automation and new tools will change the boundaries. Most importantly, keep a small allowance for exploratory questions. Some of the most valuable work begins as an ill-formed request.
What to do with the result
The first output should be a one-page schedule showing the three bands, their typical cost and the action each band triggers. The second should be a short list of requests to automate, retire or return to the business with a named owner. The third should be a commitment to repeat the sample in six months.
Unit economics will not settle every argument about data funding. It will, however, replace assertion with evidence. For UK data leaders facing another year of constrained budgets, that is a modest but useful advantage: the ability to say not merely what the platform costs, but what a question costs, and what the organisation proposes to do about it.