Walking Backwards From a Missed Deadline

The date was 30 April. On 24 April, Ines told her director the migration would land on 12 May.

Nobody was shocked. Two weeks isn't a scandal; the client accepted it, and everyone moved on. The retrospective produced three lines: we underestimated the data cleanup, the vendor was slow, and we need better planning next time.

This retrospective is worth nothing, and the reason is that it starts at the end. Missed deadlines are almost never decided in the final fortnight. By the time you can see a slip, the decision that caused it is six weeks behind you, and it looked completely reasonable on the day it was made.

So run it backwards instead. Take the slip date, walk back through the calendar, and stop at every point where a different choice was available. Let’s learn how to do that through Ines’s story.

Six weeks earlier

Week one of the project, mid-March. Ines built the plan forward: start here, work through the steps, arrive at 30 April with a little room. Standard practice, and the arithmetic worked on the page.

Working forward hides two things. It hides the steps that belong to other people, such as approvals, vendor lead times, a security review that runs on its own schedule, because you plan your own work and assume theirs fits around it. And it hides how long the last mile takes, because the end of a plan is always the vaguest part of it.

Tina Larsson learned this lesson from a public goal of 100 podcast appearances in a year. She planned forward, believed she had time, and missed it. The year she reverse-engineered the same goal from December backwards into monthly and weekly numbers, she hit it. (Tina Larsson, "Working Backwards from Your Deadlines," January 2026)​

The fork here: on the day the plan was built, nobody asked what had to be true on 29 April and a week before that. The plan had a start date and an end date and no reverse checkpoints in between.

The signal everyone saw and ignored

Week three. The vendor's data extract, due Monday, arrived Thursday.

Three days. Ines noted it, absorbed it, and moved the internal dates along by three days without telling anyone, because the end date still floated it and raising a three-day vendor delay felt like noise.

That was actually the signal. Not the delay itself, which was small, but the fact that the buffer had started paying out in week three of an eight-week plan. Buffer spent that early is buffer that won't be there at the end, and the end is when you need it.

There's a documented version of this playing out in New York right now. Every existing traction elevator in the city with a single-plunger brake has to be altered to a dual-plunger type or fitted with Unintended Car Movement Protection by 1 January 2027. The requirement is retroactive and sits in Appendix K, Chapter K3, Section 3.8.4.1 of the New York City Building Code, with the filing and permit process set out in the Department of Buildings service notice. (NYC.Gov)​

Boards looking at their calendars in early 2026 saw a year of runway ahead. Walk it backwards, and the picture inverts. No contractor does non-emergency elevator work on New Year's Day, so the working deadline is 31 December 2026. Elevator companies across the city are already booked deep into the year, and the DOB requires the work finished and acceptance-tested before the date, so a filed permit or work in progress won't count. A full modernization runs to about twelve months of planning, funding, and execution. This leaves the first quarter of 2026 as the window to engage a consultant and get it into the budget without panic.

Same date, two ways to go about it. One of them says you're fine.

The moment it was still fixable

Tuesday 1 April. Ines had the extract, the cleanup had begun, and the first sample showed roughly 18% of records with a field mismatch nobody had scoped.

She had a call that afternoon with her director and used it to give a status: cleanup underway, some data quality issues, and still tracking to 30 April.

Every ingredient of the slip existed on that Tuesday. Three days of buffer are already gone, an unscoped 18%, a vendor with no urgency, and four weeks left. Anyone reading those four facts together could have told you the date was in trouble. They were never in the same sentence, so nobody read them together.

The Tuesday is the answer to the autopsy. It's the last day when four weeks was enough to fix it by cutting scope, buying vendor priority, or moving the date early while the client still had options. After that Tuesday, the outcome was set, and everything else was just discovering it slowly.

What you'd do differently

Run the autopsy properly on your last slip in forty minutes with the people who are there.

Put the slip date on the left of a whiteboard and the project start on the right so the timeline runs backward and nobody drifts into telling the story forward.

Mark every date where a choice existed. For each one ask two questions:

• What did we know that day, and
• What was still available to us that day.

Keep going until you find the last mark where the original date was still reachable. That's the finding. Record a single date and a single choice clearly.

Then change what you plan next time. Build the next project from its end date backwards, and be specific about what "done" means, because "submit the polished 1,200-word report" and "write the report" produce different calendars. Ask what must be true one day before delivery, then three days, then a week, until you reach today.

Name the shadow deadline out loud: the day you need to be functionally finished, ahead of the day you're contractually finished. For NYC elevators, that's 31 December 2026 rather than 1 January 2027. Every project has a version of that gap.

Then set one rule for buffer: any use of it gets said out loud the day it happens. Three days lost in week three go into the update as three days lost, not absorbed silently. Buffer that disappears quietly is the single most reliable early symptom you have.

Your version of this

Take the last thing you delivered late. Give it forty minutes and one whiteboard this week.

Find the Tuesday. Every slip has one, and yours is further back than you think.

Then look at the project you're running right now and ask what today would look like in that reconstruction. If you've already spent a buffer nobody knows about, this is the week the reconstruction is still cheap.