Nick QuirkEvery agency has one: the project that was quoted at 160 hours, has burned 210, is three weeks late, and nobody wants to open the folder. The instinct is to work harder and hope it closes the gap. It never does. Overruns compound because the same conditions that caused them are still running.
A project recovery plan is the alternative: a deliberate 48-hour stop, a re-baseline using real numbers, and a client conversation that turns a problem into a decision. Here's how to run one.
First, freeze the project for 48 hours
This feels insane when you're behind. Do it anyway. A project that's over budget is burning money on every hour worked, and you don't yet know which hours are productive.
Tell the team: no new work on this account for two days except client-facing emergencies. Tell the client: "We're doing a full status review on Thursday and I'll have a revised plan for you Friday." Nobody objects to that sentence.
Those 48 hours buy you the only thing you actually need — accurate information.
Step 1: Get the real burn numbers, not the vibes
Open the time tracking and answer four questions. Write the answers down.
What has actually been spent?
Hours logged to date, times your blended delivery cost. Not your billing rate — your cost. If your team costs you $85/hour loaded and you've burned 148 hours, you've spent $12,580 of real money.
What percentage of the deliverables are genuinely complete?
Be brutal. "Draft written, not reviewed" is 60% complete, not 100%. Count completed, approved deliverables against the original scope list. If 11 of 20 items are signed off, you're at 55%.
What's the estimate at completion?
This is the number that decides everything. Take hours burned divided by percent complete. 148 ÷ 0.55 = 269 hours to finish. On a $24,000 fixed-fee technical SEO migration, that's $22,865 of delivery cost against $24,000 of revenue — before any rework. You are running a 5% margin project that you sold at 45%.
Most agencies discover at this point that the overrun is worse than they thought. That's normal. The estimate-at-completion calculation is unforgiving because it assumes your remaining work will be as inefficient as your completed work, which is almost always true unless you change something.
What's the true remaining calendar time?
Not "three weeks of work." Three weeks of work at what capacity, with which people, around which other commitments? If your senior strategist has 6 hours a week free and the recovery needs 40 hours of her time, that's seven weeks, not three.
Step 2: Diagnose the cause in one sentence
Overruns on agency projects come from a short list. Name yours before you plan anything.
- Estimate error. You quoted 160 hours for work that was always 240. Nobody did anything wrong after the SOW was signed.
- Scope creep. Fourteen small "can you also" requests, each 90 minutes, nobody tracked them. This is the most common and the most preventable — see how to say no to out-of-scope requests for the language that stops it.
- Client-side delay. You waited 19 days for dev access and 11 days for content approval, and your team context-switched back in five times at an hour each.
- Rework. Direction changed after work was delivered, or quality wasn't right the first time.
- Resource churn. The person who scoped it left, and the replacement spent 30 hours getting up to speed.
The cause determines who pays. Estimate error and rework are yours to absorb. Client delay and scope creep are legitimately billable — if you documented them. If you didn't, you absorb this one and start documenting tomorrow.
Step 3: Rebuild the plan around what's left
Do not try to deliver the original scope faster. That's how projects die twice. Re-baseline instead.
Split the scope into three buckets
- Must ship. The work the client is actually paying for — the outcome they'll judge you on. On a migration, that's the redirect map, the crawl-clean launch, and the rankings-protection monitoring.
- Should ship, can shift. Valuable but not launch-blocking. Content optimisation on tier-2 pages. Move it to a follow-on phase or next month's retainer.
- Cut. Things that crept in, nice-to-haves, and anything you added to look generous during the pitch. A competitor gap analysis nobody has asked about since week two.
On a typical recovery, bucket three is 15-25% of remaining effort. On our 269-hour example, cutting 20% and shifting another 15% brings the remaining work from 121 hours down to roughly 79. That alone can take a 5% margin project back to 25%.
Compress the delivery mechanics
Two structural changes that reliably buy back hours:
Reduce review rounds from three to one. Set a single consolidated feedback window with a 48-hour deadline. Multiple review cycles are where agency projects quietly lose 20-30 hours.
Assign one owner per deliverable. Shared ownership on a struggling project means everyone waits. Name a person and a date for each remaining item. Most late projects are late for coordination reasons, not capability reasons — the patterns in why deadlines get missed show up in almost every recovery post-mortem.
Step 4: Have the recovery conversation
Call the client. Not email. Fifteen minutes, camera on, with the revised plan already written.
The structure that works:
"I want to give you a straight update. We're at 55% of the deliverables with 92% of the budget consumed. Two things drove that: we underestimated the redirect mapping by about 40 hours, and we lost 19 days waiting on staging access. I'm not asking you to cover the estimating miss — that's ours. Here are three ways forward."
Then present options, with numbers:
- Option A — reduced scope, original date and fee. We deliver the migration and monitoring by the 14th. Content optimisation moves to a separate phase.
- Option B — full scope, extended timeline. Everything ships, launch moves to the 28th, fee unchanged.
- Option C — full scope, original date, change order of $6,400 to bring in a second specialist.
Roughly three quarters of clients pick A or B. Almost none of them fire you. What damages relationships is the silent overrun that surfaces as a late, incomplete delivery. Being the agency that spots trouble at 60% and shows up with options is a genuine differentiator — the same principle behind good client boundary setting generally.
Step 5: Run the recovery with weekly checkpoints
A recovery plan without a cadence becomes the same drift you just escaped. Every Monday for the duration:
- Hours burned last week vs. planned
- Deliverables completed vs. planned
- Updated estimate at completion
- Any new request logged as a change order, not absorbed
Ten minutes, stand
