I Found Ten Tickets for a Decision Nobody Had Made

Mid-investigation on a redesign for a media company, I looked up and found ten engineering tickets drafted and ready to file. Eight of them pointed at a single engineer. They were for a direction that had not been chosen yet.

Nothing was wrong with the tickets. They were well-written, correctly scoped, and genuinely useful if the decision went that way. That is precisely what made them dangerous.

I killed the filing and parked the drafts in a working document instead. Zero tickets went into the tracker that day.

What an artifact does once it's in the system of record

Here's the mechanism, and I think it's underrated.

Filed, those ten tickets would have made one option look decided. Not argued for – decided. A backlog holding ten well-scoped tickets for Option A and nothing for Option B is an argument for Option A that nobody has to make out loud, nobody has to defend in a room, and nobody gets to vote on. The work arrived carrying a position, and the position came in under the door.

This is different from the thing everyone already knows about sunk cost. Sunk cost is about money you've already spent. Escalation of commitment is about doubling down after the bad news lands. Both of those require something to have happened first. What I'm describing costs nothing, happens before any decision exists, and is invisible precisely because it looks like diligence. Somebody was being helpful. Somebody got ahead of the work.

The generalizable version: work produced ahead of a decision does not stay neutral. Once it is in the system of record it starts voting, for the option it was written against, at the expense of the ones nobody drafted.

Why the tracker specifically

You could reasonably ask why it matters where the drafts sit. They're the same words either way.

It matters because the tracker is the system of record, and a system of record confers standing. Things in it are real in a way that things in a doc are not. They get counted in capacity planning, they show up in a sprint-planning view, they get assigned, they acquire an owner who now has an opinion about them. A stakeholder scrolling the backlog sees ten items and reads the direction as settled. Nobody told them it was. That's simply what a populated backlog means.

The drafts kept all their value in a working document. They lost their vote.

This is the same machine that costs companies billions

I'd have filed this under "a small thing I noticed" except that the organizational version of it is well documented, and it's expensive.

A 2026 Harvard Business School working paper on why international market entries fail makes almost exactly this argument at company scale. The authors find that failure is usually decided before entry, and – the part that stopped me – that the reason companies don't reverse course afterward isn't blindness. It's that once capital is committed, the board has approved the investment and the leases are signed, the structural question "should we be here at all?" cannot be re-asked "without admitting that it was never properly answered."

Their line about Home Depot's six years in China is the sharpest version I've read: people had noticed the problem, and the company stayed anyway, because the institutional cost of honesty under sunk-cost pressure exceeded the institutional cost of persisting.

Ten tickets are that same machine, running at desk scale, for free. The commitment doesn't need capital. It just needs standing.

The conditions that make this worse than it should be

Two numbers from the 2026 product-management survey data explain why a quiet argument wins so easily.

First, 60.2% of teams cite leadership escalations or new directives as the number one reason priorities change. Direction, in most organizations, does not get set in a deliberate forum where options are compared. It moves in response to whoever raises something.

Second, and worse: 7.4% of organizations report no clear decision-maker for prioritization at all. In roughly one team in thirteen, there is no person whose job it is to choose – which means the choice gets made by whatever's already in the system when the sprint starts.

Put those together and you get an environment where an undrafted option never gets a hearing. Not because it lost. Because nothing was ever convened where it could compete.

Tickets are just the obvious case

Once you're looking for it, the pattern shows up in every artifact a team produces early.

A design comp is the worst offender, because a picture of one option is enormously more persuasive than a description of three. I've watched a room spend forty minutes discussing the button placement in a mock for a direction that hadn't been agreed, and by the end of it the direction was agreed, silently, by everyone who'd been critiquing the buttons. Engagement with the details of an option reads as acceptance of the option.

Technical spikes do it too. A spike is supposed to reduce uncertainty about feasibility, which is genuinely useful, but a spike that comes back with "yes, we can build it, here's the architecture" has quietly reframed the question from should we to how. The answer to a feasibility question is not a decision, and it's remarkable how often it gets treated as one.

Job postings are the most expensive version I've seen. Hiring for a capability you haven't committed to is a bet with a twelve-month lock-in and a human being attached, and once the req is open the direction has a budget line, a hiring manager, and a candidate pipeline arguing for it.

None of these are bad artifacts. They're all reasonable things to produce while you're thinking. The question is only ever whether they're sitting somewhere that gives them a vote.

The test I run now

Before a direction is chosen, I ask one question about any work product: does this have standing?

If the answer is yes – it's in the tracker, it's in the roadmap tool, it's in the deck going to the exec team – then it's voting, and it needs to be treated as an argument rather than as preparation. Either it's ready to be defended out loud, or it shouldn't be there yet.

If you genuinely need to prepare ahead of a call, and often you do, put the preparation somewhere with no standing. A working doc. A branch nobody's watching. A file named after the question, not after the option. The preparation keeps every bit of its value and gives up its vote, which was the only thing wrong with it in the first place.

I'd also push back on the instinct that says drafting ahead for both options solves it. Sometimes it does. More often you'll find that one option is easy to draft and the other is vague, and you'll have produced a rigorous case for the tractable choice and a sketch for the important one. Effort asymmetry is just the same bias wearing a lab coat.

The uncomfortable part

The thing that makes this hard to catch is that the person who drafts ahead is doing a good job. They're being proactive. They're removing friction. In most organizations they'd be praised for it, and if you tell them to stop, it lands as a rebuke of exactly the behaviour you normally ask for.

So don't tell them to stop. Tell them where to put it. The instinct is right; the destination is wrong.

On that redesign, the engineer who wrote those tickets had done good work, and the work survived intact – it just sat somewhere that couldn't cast a ballot. Two weeks later we chose a different direction, and roughly half of what he'd drafted turned out to apply to it anyway. If those tickets had been in the tracker, I don't think we'd have chosen the other direction at all. Not because anyone would have argued for the first one. Because by then it would have stopped looking like a choice.

A decision nobody made is still a decision. It's just one you can't point at.