Most agencies have a number they use on everything: 10%.
It is simple, defensible, and wrong in both directions. A flat 10% on a novel project with a brand-new client gets eaten in week two. A flat 10% on a repeat sprint with a client you have served for three years is money you left on the table. The right approach is a scoring exercise, not a fixed percentage. Here is the version I would run.
The five-factor scoring exercise
Score five risk factors from 0 to 2 and add them up. Score honestly, not aspirationally; the exercise only works if the 2s stay 2s.
| Factor | Scores 0 | Scores 1 | Scores 2 |
|---|---|---|---|
| Client familiarity | Long-term repeat | Worked once before | Brand new |
| Scope clarity | Fully written SOW | Some ambiguity | Still being discussed |
| Project novelty | Done 5+ times | Similar, new tech or industry | Genuinely new work |
| Dependencies | All internal | One external | Multiple or third-party platforms |
| Timeline pressure | Comfortable | Tight but doable | Hard deadline, cannot move |
A total of 0 to 2 maps to 5%, 3 to 4 to 8%, 5 to 6 to 12%, 7 to 8 to 15%. A score of 9 to 10 gets 20%, and at that point you should ask whether the project is worth quoting at all.
Worked: an $80,000 website build
An $80,000 website build. Brand-new client (2), ambiguous SOW (1), new CMS (1), one external dependency (1), hard launch date (2): total 7, which maps to 15%, or $12,000 of contingency.
Your target margin is 30%, so $24,000 of the $80,000 is supposed to be profit. The project runs 10% over on delivery: $6,800 in extra hours. Without contingency, margin drops from $24,000 to $17,200, about 21.5%. With the $12,000 contingency priced in, the overrun is covered and $5,200 drops straight to profit, leaving you at $29,200, above plan. The contingency did not make the project more expensive for anyone. It made the quote honest.
Three rules that make contingency work
Three rules make it work. Contingency is not margin: the moment you treat it as margin, you cut it to win work, and you win the work and lose the money. Track what gets consumed on every closed project; that record turns the scoring from a guess into a system. Unspent contingency is profit, so do not invent ways to spend it.
The scoring conversation is the real product.
When the team scores scope clarity a 2 and the account lead scored it a 1, that gap is telling you something about how uncertain the work actually is. The agency consultant Michael Farmer has argued that agencies routinely deliver 10 to 20% of work that adds no value to the client's program simply because the scope was not engineered correctly. Scoring contingency is scope engineering with a number attached. On fixed-fee quotes, keep the contingency internal; the client buys one price and one outcome. On time-and-materials work, a visible contingency line is fine and often appreciated. For the pricing-model context, see fixed fee vs hourly and what a good agency margin looks like.
Frequently asked questions
What percentage contingency should an agency add to a project quote?
Between 5% and 20%, depending on risk, not a flat number. Score five factors (client familiarity, scope clarity, project novelty, dependencies, timeline pressure) from 0 to 2 each: totals of 0-2 map to 5%, 3-4 to 8%, 5-6 to 12%, 7-8 to 15%, and 9-10 to 20%.
Is contingency the same as profit margin?
No. Contingency is a priced allowance for things you know can go wrong without knowing whether they will. Margin is what remains after all delivery costs. Treating contingency as margin leads agencies to cut it to win work, then lose money delivering it.
Should the client see the contingency line in the quote?
On fixed-fee quotes, keep it internal: the client buys one price and one outcome. On time-and-materials work, a visible contingency line is fine and often appreciated. Either way, build it into the quote math from the start.
What happens to unspent contingency?
It drops to profit. Do not invent ways to spend it. Tracking how much contingency each closed project actually consumed is what turns the scoring from a guess into a system.
What contingency for a brand-new client on a brand-new type of work?
That combination typically scores 9 to 10, which maps to 20% contingency. At that level, seriously consider whether the project is worth quoting at all.