What to do when a project is 150% over estimate

The instinct when a project blows up is to work harder and hope. That instinct is expensive, because the only thing that reliably makes a runaway project worse is another four weeks of the same plan.

The moment you know the estimate is gone, stop delivering for a day and do the arithmetic. It's a two-hour job and it usually changes the decision.

The situation

A hypothetical fixed-price project: $60,000, estimated at 400 hours, loaded cost $95 an hour. On estimate that's $38,000 of cost and 36.7% margin.

It's week nine. 380 hours are logged, and an honest assessment puts the work at about 38% complete.

Projected total = 380 / 0.38 = 1,000 hours
1,000 / 400 = 2.5× the estimate — 150% over

At 1,000 hours the delivery cost is $95,000 against $60,000 of revenue. Finishing the project as planned loses $35,000.

Also worth computing: break-even is $60,000 / $95 = 632 hours. You passed the point where this project could make money somewhere around hour 632 of a projected 1,000 — which is to say, you are not going to recover this by being efficient from here.

Get an honest completion percentage first

Everything above depends on "38% complete", and that figure is where these analyses go wrong. Percentage complete is the most optimistically reported number in agency work, because the remaining work is the part nobody has looked at closely yet.

Get it by listing the deliverables from the SOW and marking each as done, in progress, or not started — then estimating the not-started items from scratch, today, with what you now know. Do not reuse the original per-item estimates. They are the thing that failed.

If the recount comes back saying you're further behind than 38%, that's not demoralising, it's the point of doing it.

The four options, with numbers

Option	Revenue	Total cost	Result
Finish as scoped (1,000h)	$60,000	$95,000	−$35,000
Cut to essential scope (700h)	$60,000	$66,500	−$6,500
Stop now, settle at 38%	$22,800	$36,100	−$13,300
Renegotiate to $75,000 + cut to 700h	$75,000	$66,500	+$8,500

The spread between the worst and best rows is $43,500 on a $60,000 project. That spread is the value of spending a day on this rather than pushing on.

Note that stopping now is not the cheapest option, despite feeling like the decisive one. Sunk cost works in both directions: the $36,100 you've already spent is gone whether you finish or not, so the only question is whether the remaining work earns more than it costs. From here, delivering a reduced scope adds $30,400 of cost and $37,200 of revenue you'd otherwise never collect.

Option four is usually available

Agencies assume renegotiating a fixed price is impossible. It's hard, and it's far from impossible, if three things are true.

Something changed, and you can name it. Not "we underestimated" — a specific factual change: the review group grew from three people to nine, the source material wasn't what was described, the requirements shifted in month two. If your SOW listed assumptions, this conversation is much easier, because you're pointing at a document rather than making a claim.

You're raising it before the deadline, not after. A client can act on information in week nine. In week fifteen, with a launch date public, you've removed their options and they'll resent it.

You're bringing a plan, not a problem. Never open with the overrun. Open with what you propose to do about it.

The shape of the conversation:

"I want to give you an early and honest update. Based on where we are, the remaining work is materially larger than we scoped — mainly because [specific change]. I don't want to surprise you at the deadline, so here are three options: reduce to the scope that matters most for launch and hold the current fee; keep the full scope with an additional $15,000 and a three-week extension; or launch with the reduced scope now and phase the rest. My recommendation is the first."

Three options, one recommendation, no blame. That's a conversation a competent client can have with you. Most will meet you partway, and even partway on a 150% overrun is a large number.

Cutting scope is the highest-value move

Look at the table again. Reducing from 1,000 projected hours to 700 saves $28,500 of cost. No client conversation about money is required — just one about priorities.

Ask which deliverables the launch genuinely depends on. On almost every runaway project, a meaningful fraction of the remaining work is polish, secondary variants, or documentation that nobody will read this quarter. Some of it can be deferred to a phase two that the client will pay for separately, which turns cost into future revenue.

The framing that gets a yes: "to hold the date, here's what I'd propose we ship now and what I'd propose we phase". Clients care about dates far more than they care about your original deliverable list.

Stop the bleeding while you decide

Two immediate actions, taken the day you spot the problem:

Take the expensive people off, unless they're the reason it finishes. A runaway project attracts senior attention. Sometimes that's the fix. Often it just raises the cost per hour of a project that is already losing money.

Freeze all new scope, formally. Say it out loud to the client: "until we've agreed a plan, I'm not adding anything new." A project that's 150% over cannot absorb another favour.

Afterwards, find the week it happened

Runaway projects don't go wrong at the end. There is always an identifiable moment — usually in the first third — where the plan stopped matching reality and nobody said anything.

Find that week. In a post-mortem, the useful question isn't "why did this take so long", it's "what was the earliest point at which we could have known?" The answer is normally: when the second deliverable slipped and we assumed we'd make it up.

The prevention is one number, weekly

Projected cost = cost to date + (burn rate × estimated weeks remaining)

Compare that to the quoted value every week. On the project above, the trajectory was visible by week four — with enough runway to cut scope, raise a change request or renegotiate while the client still had a comfortable amount of time.

Pick your largest active project and calculate that number today. If it's fine, you've lost fifteen minutes. If it isn't, you've just bought yourself the only thing that makes a difference on a project like this, which is time.