Every developer has performed this play. Most of us have performed it this month. It has three acts, a cast of one optimist, and an ending everybody in the audience saw coming except the lead actor.

Take your seats. Phones on silent. The curtain rises.

Act I: The Question

Monday morning. A stand-up meeting. Coffee is still hot. Hope is still alive.

Product manager: Small one. Can we add an "export to CSV" button on the payments page?

Developer (already picturing the code): Oh, that's easy. The query exists, I just loop and write rows. Two hours.

Somewhere backstage, a stagehand quietly cancels the developer's Wednesday.

Notice what happened. The developer didn't estimate the task. They estimated the happy path of the part they could already see: the loop. Nobody estimated the button, the permissions, the edge cases, the review or the deploy. Those weren't in the picture yet.

Act II: The Discovery

Monday afternoon, then Tuesday, then a bit of Wednesday.

The loop works in twenty minutes. The developer feels like a genius. Then the plot arrives:

• ❌ Some payments have a tenant name with a comma in it. The CSV now has an extra column for one row and Excel is furious.
• ❌ Arabic names come out as question marks in Excel until you add a byte order mark. Nobody has ever remembered the BOM on the first try.
• ❌ The page filters by date and status. The export should obviously respect the filters. "Obviously" was not in the ticket.
• ❌ Finance has a lot of payments. Exporting all of them in one request times out. It needs to be a queued job that emails a link.
• ❌ Who's allowed to export? Turns out not everyone who can see the page.
• ❌ The code review asks, reasonably, for a test.

None of this was hidden. All of it was predictable. It just wasn't visible from the stand-up, because at the stand-up we were looking at the loop.

Act III: The Reckoning

Thursday. The same stand-up. The coffee is cold. So is the room.

Product manager: How's the two-hour button?

Developer: Nearly there. Just waiting on review.

The audience nods. They have all said "nearly there". They have all been "nearly there" for a full day.

The button ships Friday. It's a good button. It handles commas, Arabic, filters, permissions and big exports. It just wasn't a two-hour button. It never was.

Curtain.

Behind the scenes: why the play always ends this way

Douglas Hofstadter summed it up decades ago: it always takes longer than you expect, even when you take into account Hofstadter's Law. A few reasons it keeps happening to smart people:

• We estimate the coding, not the work. Coding is often the smaller part. Understanding, edge cases, testing, review, deploying and checking it in production are the rest.
• Unknowns don't show up in the estimate because they're unknown. You can't add time for a problem you haven't imagined yet.
• Optimism is the default setting. We picture the version of the task where everything works first time.
• A single number sounds like a promise. "Two hours" leaves the meeting as a commitment, even if you meant "two hours if nothing weird happens".

Director's notes: how to write a better script

You can't remove uncertainty from software. You can stop hiding it.

✅ Break it down before you answer. Spend five minutes listing the pieces: UI, query, permissions, edge cases, tests, review, deploy. Estimate the pieces, not the feeling. The list itself usually exposes Act II before it happens.

✅ Give a range, not a number. "Half a day to two days, depending on how big the exports get" is more honest and more useful. It tells the listener where the risk is.

✅ Name your assumptions. "Assuming the export follows the current filters and only managers can use it." When an assumption breaks, the estimate changing is expected, not a failure.

✅ Add a buffer for the things you can't see. Not as padding to hide, but as a visible line. Everyone knows the unknowns exist.

✅ Spike the scary part. If one piece is genuinely unknown, time-box an hour to investigate it, then estimate.

✅ Compare with last time. Your memory of how long things take is optimistic. Your ticket history isn't.

Estimates will always be a bit of theatre. But a play where everyone knows Act II is coming is a lot less stressful than one where it's a surprise every single time.

What's the most innocent-sounding "two-hour" task that ate your whole week?