Key takeaways
- Fixed pricing is only survivable once delivery is systematised - it forces every other discipline into place, which is precisely why it is the clearest external marker of an AI-native practice.
- You need duration data on named actions before you can quote with confidence. One month of rough timing across open matters is usually enough to start.
- Price the matter type, not the matter. Variance within a well-defined matter type is what your buffer covers; variance across types is what separate prices are for.
- Scope definition is the entire risk-management mechanism. A fixed fee with an undefined scope is not a price, it is an unfunded liability.
- Start with your highest-volume, lowest-variance matter type and expand only as duration data accumulates.
Why fixed pricing is a systems problem, not a pricing problem
The billable hour trained the profession to record time rather than manage it. That distinction sounds academic until you try to quote a fixed fee. Recording time tells you what you billed. Managing time tells you how long a given piece of work actually takes, how much that varies, and why. Only the second one lets you name a number in advance without gambling.
This is why firms that attempt fixed pricing as a marketing move usually retreat within a year. They publish a number derived from instinct, absorb three matters that run long, watch margin evaporate, and conclude that fixed fees do not work in their practice area. The pricing model was not the problem. The absence of duration data was.
Run the other way round, fixed pricing becomes almost mechanical. If you know that your standard commercial lease review takes between four and seven hours across the last twenty instances, with a median of five, you can price it, buffer it, and be right most of the time. The pricing question resolves itself once the measurement question is answered.
The dependency
Fixed fee depends on measured duration. Measured duration depends on named actions. Named actions depend on a defined matter. This is the MATTER sequence arriving from a different direction.
Step one: gather duration data on named actions
You cannot price what you have not measured, and you cannot measure what you have not named. If your action library does not yet exist, build it first - the sequence of named, repeatable moves that carry your most common matter type from open to close.
With that list in hand, start timing. This does not require precision instrumentation. Rough timing, logged consistently for one month across every open matter of that type, produces enough signal to price against. What you want at the end is, for each named action, a low, a median, and a high.
- 01
Log against actions, not against the matter as a whole
A total of 'eleven hours on the Sharma matter' tells you nothing reusable. 'Intake 0.5, document review 3.5, first draft 4, revisions 2, filing 1' tells you where the time actually goes, which action is worth automating, and how to price the next one.
- 02
Capture the outliers separately
When a matter runs long, note why in one sentence. Over a month you will find that outliers cluster into three or four recognisable causes - an unresponsive counterparty, a document set arriving in poor condition, a scope change the client did not flag. Each of those becomes either a scope exclusion or a pricing tier.
- 03
Include the invisible work
Intake calls, chasing documents, internal handoffs, and the administrative choreography around the work are real duration. Firms consistently under-price because they measure only the substantive drafting and forget the four hours of coordination surrounding it.
- 04
Stop at one month, then price
Perfect data is not the goal; sufficient data is. One month of consistent logging on a high-volume matter type gives you a defensible first price. You will refine it with every subsequent matter, and that refinement is the compounding advantage.
Step two: set the price
With duration data in hand, pricing becomes arithmetic plus a judgment call about risk appetite. The arithmetic is straightforward: median duration multiplied by your target effective rate, plus a buffer sized to the variance you observed.
The judgment call is how much buffer. A matter type where the low and high are four and seven hours needs a modest buffer. One where they are four and twenty-two hours is not yet a single matter type - it is two or three different matter types wearing the same name, and it should be split before it is priced.
| Input | Value | Note |
|---|---|---|
| Median duration (measured) | 5.0 hours | Across 20 instances over one month |
| Low / high | 4.0 / 7.0 hours | Variance is tight enough to price as one type |
| Target effective rate | Your figure | The rate you want to earn, not your list rate |
| Buffer | +15% | Sized to observed variance; wider variance needs a wider buffer or a split |
| Automation adjustment | Recalculate after | Once first-draft time drops, margin improves before the price does |
| Scope exclusions | Named explicitly | Multi-party negotiation, non-standard indemnities, etc. priced separately |
Step three: define scope, which is where the risk actually lives
A fixed fee with a vague scope is not a price. It is an open-ended commitment with a cap on your side and none on the client's. Every fixed-fee practice that works has an unusually precise definition of what is included, and an equally precise list of what triggers a separate conversation.
The good news is that the action library you built for pricing is also your scope definition. The included actions are the scope. Anything not on the list is out of scope by construction, which is a far easier conversation to have with a client than a retrospective argument about reasonableness.
- State the included actions explicitly, in client-readable language, in the engagement letter.
- Name the three or four most common scope-expansion triggers you observed in your duration data, and state what happens when one occurs - usually a fixed add-on rather than a reversion to hourly.
- Set a revision limit where relevant. 'Two rounds of revisions included' is normal, fair, and prevents the single most common margin leak.
- Say what happens if the client is unresponsive, because delay is a real cost and an unaddressed one in most fixed-fee agreements.
- Do not price uncertainty you cannot bound. Litigation with an unknown counterparty posture is a poor early candidate; transactional and process-driven work is a good one.
Step four: sequence the rollout
Do not convert the whole practice at once. Pick your highest-volume, lowest-variance matter type, price it, run it for a quarter, and check your realised margin against the model. The first cohort of fixed-fee matters is a calibration exercise as much as a commercial one.
Once that type is stable and profitable, add the next. Most practices find that within three or four matter types they have covered the majority of their volume, and the residual long tail can stay hourly or move to a capped-fee arrangement without undermining the positioning.
The strategic point is worth stating plainly: publishing fixed prices is the single most visible signal that a practice has systematised its delivery. Clients read it as confidence. Competitors read it, correctly, as evidence that you have measured something they have not.
What changes once you are pricing this way
| Dimension | Hourly billing | Fixed fee on measured data |
|---|---|---|
| Incentive on efficiency | Efficiency reduces revenue | Efficiency increases margin |
| Client experience | Open-ended risk, questions rationed by cost | Known cost up front, questions free |
| Effect of automation | Deletes billable time, so it threatens the model | Expands capacity and margin simultaneously |
| What gets measured | Time recorded for invoicing | Duration per action, for pricing and improvement |
| Where risk sits | With the client, uncapped | With the firm, bounded by scope definition |
| Improvement loop | None inherent | Every matter refines the duration model |
Frequently asked
How do I set a fixed fee for legal work I have never measured?
You do not, initially. Spend one month logging rough duration against named actions across every open matter of your target type. That produces a low, median, and high per action, which is enough to price the next matter with a defensible buffer. Pricing before measuring is how firms lose money on the transition and conclude fixed fees do not work.
What if a fixed-fee matter runs long?
Some will - that is what the buffer is for, and across a portfolio the overruns are covered by the matters that run short. What must not happen is uncontrolled overrun caused by undefined scope. If a matter is running long because the client expanded the work, that is a scope trigger you should have named in advance and priced as an add-on.
Does fixed-fee pricing work for litigation?
Selectively. Process-driven and stage-bounded litigation work prices well - a defined pleading stage, a discovery review of a bounded document set, a standard application. Open-ended adversarial work where the counterparty controls the pace is a poor early candidate. Most litigation practices price stages rather than the whole matter.
Should I publish my prices on my website?
Publishing is the strongest version of the signal and it is what the AI-native cohort tends to do, but it is not required to get the operational benefit. An instant quote generated at intake achieves most of the same client-experience effect. What matters is that the client knows the cost before the work starts.
How does AI change fixed-fee pricing?
It widens the margin on every matter type where a routine action has been automated, and it does so before you change the price. Firms typically hold price steady for a period after automating first-draft production, bank the margin improvement, and then decide whether to pass part of it on competitively. Because delivery is systematised, the duration data keeps updating and the price can follow it deliberately rather than by guess.