Most agencies choose a pricing model early, then keep it long after the work has changed shape. The model itself is rarely the problem. The problem is that the assumptions underneath it — how long work takes, how often scope moves, how much delivery time never gets billed — are never checked against what actually happened.
This guide walks through the pricing models agencies commonly use, where each one tends to leak margin, and how to test whether the model you are using still fits the work you are selling.
Where commercial risk sits in each model
Every pricing model is a decision about who carries the risk of the work taking longer than expected. That single question explains most of the differences between them.
- Hourly and day rates: the client carries overrun risk, so the agency is protected on cost but capped on upside. Margin is decided almost entirely by rate versus true cost of delivery, and by how much time is recorded accurately.
- Monthly retainers: risk is shared, but only if scope is defined. Undefined retainers drift — the same fee quietly absorbs more work each month until the effective rate falls below cost.
- Fixed fee or project pricing: the agency carries all overrun risk. This model rewards accurate estimating and punishes optimistic scoping. It is where most avoidable margin loss happens.
- Value-based pricing: price is anchored to the client's commercial outcome rather than effort. Upside is significant, but the agency still needs a cost floor, or a high-value engagement can be delivered at a loss.
- Performance or outcome-based pricing: revenue depends on results the agency only partly controls. Workable where results are measurable and attribution is clean; risky where either is arguable.
- Hybrid models: a base retainer or fixed fee plus variable elements. Usually the most realistic structure for mature agencies, provided each component is priced on its own logic rather than averaged.
The rate you quote is not the rate you earn
Almost every pricing conversation happens in quoted rates. Margin, though, is determined by the effective rate: fee divided by the hours actually delivered, including the hours nobody planned for.
Those unplanned hours are consistent across the industry — additional rounds of amends, stakeholders added mid-project, discovery that should have happened before the scope was signed, reporting and account management absorbed into the fee, and work delivered after the budget was exhausted because the relationship mattered more than the invoice.
A project quoted at a healthy rate can be delivered at a poor one without anybody making an obviously bad decision. The gap only becomes visible when the proposal, the delivered time and the financial outcome are looked at together.
How to choose a model for a given piece of work
Rather than adopting one model agency-wide, match the model to how predictable the work is.
- Well-understood, repeatable work with strong historical data: fixed fee is safe and commercially attractive, because your estimate is grounded in evidence.
- Novel or exploratory work: time and materials, or a paid discovery phase priced separately, so the fixed price is set after the unknowns are resolved.
- Ongoing work with a stable scope: retainer, with a defined inclusion list and an explicit process for anything outside it.
- Work with a clear, measurable commercial impact: value-based pricing, floored by a cost-plus minimum you will not go below.
- Work where the client wants shared risk: hybrid, with a base fee covering delivery cost and the variable element carrying the upside.
Discounting is a pricing model too
Discounting is usually treated as a commercial gesture rather than a pricing decision, which is why it is rarely measured. A ten per cent discount on a project running at a thirty per cent margin removes a third of the profit. Applied habitually to win work, it changes the economics of the agency without ever appearing in a pricing policy.
The same is true of rate cards that have not moved in three years, minimum engagement sizes that are quietly waived, and payment terms that stretch without a corresponding price adjustment.
Validating a pricing model against your own history
A pricing model is a hypothesis about your business. Testing it does not require new data — it requires connecting data you already hold.
- Compare the scope in the proposal with the scope in the signed SOW, and both with what was ultimately delivered.
- Compare the hours estimated with the hours recorded, by project type and by client.
- Compare the quoted rate with the effective rate once all delivered time is counted.
- Identify which project types, clients or deliverables repeatedly overrun, and by how much.
- Check whether discounting correlates with win rate at all, or only with lower margin.
Once those patterns are known, they can inform the next quote rather than the next post-mortem. That is the practical difference between reporting on margin and protecting it.
Making pricing decisions with evidence
Agencies rarely lose margin because their pricing model is conceptually wrong. They lose it because the model is applied without the delivery evidence that would have adjusted the number — the comparable project that took forty per cent longer, the deliverable that is always underestimated, the client whose review cycles double the timeline.
Keeve exists to bring that evidence into the moment a price is set. It connects the proposals, SOWs, resourcing data, task history and financial outcomes an agency already has, combines them with the agency's own rate cards, margin rules and discounting logic, and surfaces the relevant commercial context while the scope is still being written.
Choose the pricing model that fits the work. Then let your own history tell you whether the number is right.
More on scoping and margin in Keeve Insights.