Should I pay my agency a fixed price or by the hour?

The wrong contract structure can cost you as much as the wrong agency.

Most founders pick a payment model the way they pick a hotel room, by whatever the agency puts in front of them first. But fixed price and time-and-materials billing have completely different risk profiles. Which one you choose should depend on how well you understand your own scope, not on which option the agency prefers.

Here is the honest breakdown of when each model works, when it backfires, and how to blend them so you are not absorbing risk you did not sign up for.

A fixed-price contract sets a total cost upfront, and the agency absorbs any cost overruns above that number. If building your product takes longer than they estimated, they eat the difference. That sounds like a good deal for you, and sometimes it is.

The mechanism is straightforward: the agency bids based on their estimate of hours required. If they estimate correctly, they profit. If they underestimate, they lose margin. To protect themselves, most agencies pad fixed-price quotes by 20–30% above what they actually expect the work to cost (Project Management Institute, 2021). You pay for their uncertainty before a single line of code is written.

The other protection agencies build in is a change-order clause. The fixed price covers exactly the scope documented in the contract. Any change, a new screen, a different user flow, a feature that turns out to be harder than expected, triggers a formal change order that costs extra. A 2021 Clutch survey found that 43% of fixed-price projects involved at least one change order, with an average cost increase of 18%.

Fixed price works well in one specific situation: when your scope is completely locked before the contract is signed. That means every screen is wireframed, every user interaction is documented, and you have no expectation of changing anything once development starts. If that describes your project, fixed price transfers real risk to the agency. If it does not, you are paying the padding without getting the protection.

Time-and-materials contracts charge you for actual hours worked at an agreed rate. The total cost is unknown at the start, because it depends on how the project unfolds.

That sounds scarier. It often is not.

The case for time-and-materials comes down to scope honesty. Most software projects change during development. You show the first version to a potential customer and realize the signup flow needs to be redesigned. You discover a competitor has a feature you did not plan for. You change your mind about how something works after you see it built. In a fixed-price contract, each of those changes triggers a change order and a bill. In a time-and-materials contract, the change just gets worked on.

A 2020 Standish Group report found that 45% of features originally specified in software projects are never used. Those features cost real money under a fixed-price contract, because you paid for them in the original quote even if they turned out to be wrong. Under time-and-materials, you can redirect that budget to features that actually matter.

Time-and-materials works best when you are building something genuinely new, when requirements are expected to evolve, or when you want the flexibility to stop, redirect, or pause the engagement without penalty. The risk is a runaway invoice. The mitigation is a weekly cap or a sprint-by-sprint authorization, which any reputable agency should offer.

Fixed price is not a bad model. It is a bad model applied to the wrong projects, which is most projects.

The danger zone is what project managers call scope ambiguity: the situation where you know what you want to build but not exactly how it should work. A founder who says "I want users to be able to book appointments" has a goal, not a specification. Does the booking system show real-time availability? Can users cancel? Do they get reminders? Can providers block certain time slots? Each of those questions has multiple valid answers, and the right answer might not be clear until you see the first version.

When scope is ambiguous, agencies quote fixed-price contracts by making assumptions and documenting them in an appendix that founders rarely read carefully. When the product gets built and the founder realizes the assumptions were wrong, the change-order clock starts ticking.

A 2019 McKinsey study on large software projects found that projects with poorly defined requirements at contract signing ran an average of 45% over budget, regardless of whether the contract was fixed price or time-and-materials. The model does not fix vague scope. It just determines who pays for the vagueness.

The specific conditions that make fixed price backfire:

You do not have wireframes or detailed specifications before signing. Without them, the agency is guessing, and padding the quote accordingly.

Your product depends on a third-party tool that the agency has not integrated before. Unknown integrations are impossible to price accurately.

You expect to get user feedback during development and adjust the product accordingly. That is by definition scope change.

You have a hard launch deadline. Fixed-price agencies facing a budget overrun have an incentive to cut corners on testing to stay within budget, because the contract does not reward them for spending extra time to get it right.

For comparison: a Western agency typically charges $150–$250 per hour on time-and-materials, or quotes fixed prices in the $50,000–$200,000 range for mid-complexity products. An experienced global engineering team with the same technical depth bills $35–$60 per hour on time-and-materials, or $15,000–$40,000 fixed for equivalent scope. The contract model matters, but so does the base rate.

Yes, and this is what sophisticated buyers actually do.

The blended model works like this: the discovery and scoping phase is fixed price, because the deliverable is completely defined (a specification document, wireframes, a project plan). The delivery phase is time-and-materials within a sprint-by-sprint budget, because software development always surfaces surprises.

The fixed discovery phase typically runs two to three weeks and costs $2,000–$5,000 at an experienced global engineering team versus $8,000–$15,000 at a Western agency. It produces a complete specification that both sides agree represents the actual product. That document becomes the foundation for an accurate delivery estimate.

Once discovery is complete, delivery can be structured in two ways. Sprint-locked time-and-materials means you authorize a fixed budget per sprint (typically two weeks) and the agency builds the highest-priority features within that budget. You can add, remove, or reprioritize features between sprints. Phase-fixed delivery means you take the post-discovery specification and price it as a fixed-price contract, because scope is now actually locked.

The blended model eliminates the two main failure modes: the padding that makes fixed-price quotes expensive when scope is fuzzy, and the runaway invoice that makes time-and-materials uncomfortable for founders without technical oversight.

One thing founders consistently get wrong: they treat the payment model as a negotiation with their agency rather than a structural decision about risk. The model you pick determines who is accountable when scope turns out to be wrong. That question is worth more attention than the hourly rate.

For a straightforward MVP with well-defined scope, fixed price at an experienced global team runs $8,000–$15,000 and ships in three to four weeks. For a product where you expect user feedback to shape the build, a sprint-locked time-and-materials model at $35–$60 per hour gives you flexibility without an open-ended invoice. The blended model works for most everything in between.

The contract model does not change the quality of what gets built. It determines who bears the cost when reality differs from the plan. Choose based on how confident you are in your own specification, not on which option the agency presents first.

Book a free discovery call