The Sunk Cost Fallacy Is Real

The Miracle Mile

Last week I floated the Miracle Mile — a legendary ten-mile stretch of the North Platte River in Wyoming, just below Seminoe Reservoir. If you fish, you've probably heard of it. If you haven't, it earns its name. The water runs cold, fast, and clear, and it holds some of the biggest wild trout you'll find anywhere in the West.

We went with guides, which turned out to matter more than I expected — not just for knowing where to go, but for how they fished it.

On a fast-moving river, the path of least resistance is obvious: drift through a hole, don't get a bite, and let the current carry you to the next one. It's efficient. It keeps things moving. But our guides kept rowing us back. Same hole, different approach — adjust the weight, change the depth, try a slightly different line through the drift. Over and over. And on many occasions, it was the third or fourth pass that produced the fish. Big fish. The kind you don't forget.

I've been writing a series of posts revisiting lessons from my Strongroom years — the company I co-founded in the early 2000s that built financial software for the real estate management industry. Lesson 4 hit me somewhere between casts on the North Platte: the sunk cost fallacy is real. But so is the temptation to move on too soon.

---

The Strongroom Story

By the time we started thinking seriously about a pivot, we'd spent years building.

The original vision was ambitious — an online AR/AP platform for the entire property management industry. Accounts receivable, accounts payable, treasury products: ACH, lockbox, credit cards, remote deposit capture. One platform for property managers everywhere.

In practice, that meant we had architectural decisions locked in, integrations built, a product roadmap we'd spent countless hours debating, and a team that had poured real time into something we genuinely believed in.

Then, as I described in the previous post, we started finding traction in HOA management — almost by accident. The problem was more bounded. The customer was more specific. The fit was real in a way it hadn't been across the broader market.

The right move was to focus there. To narrow down. To leave most of what we'd built behind and rebuild around the HOA customer.

Intellectually, that was obvious. Emotionally, it was brutal.

What we were feeling has a name: the sunk cost fallacy. It's the tendency to factor past investment — time, money, effort — into decisions that should only be about the future. Economists will tell you those costs are irrelevant. They're gone either way. The only question is what produces the best outcome from here.

But knowing that doesn't make it feel any less real when you're the one who built the thing.

We almost didn't pivot because of it. We kept telling ourselves we were "close" — that if we just finished a few more features, we'd find fit across more segments. What we were really doing was protecting our past work from the judgment of the market. It took a while to see that clearly.

---

The AI Era Changes the Math — But Not the Psychology

Fast forward to today, and the conversation around sunk costs in startups looks different in one important way: the actual investment required to build has dropped dramatically.

AI coding tools have compressed development timelines in ways that would have seemed implausible even five years ago. What once required a team of engineers working for a year can sometimes be prototyped in weeks. The raw cost — in time and money — of building a product has fallen significantly.

That's genuinely good news for founders grappling with sunk costs. If the investment is smaller to begin with, the psychological pull toward protecting it should be weaker. Pivoting feels less like dismantling something you spent years constructing and more like adjusting course on something you built last quarter. The emotional case for holding on to a bad bet gets harder to make.

On that dimension, the lesson softens. The sunk cost trap is less severe when the costs themselves are lower.

But I'd watch out for what I think is the opposite trap — one that AI makes more likely, not less.

---

Pivot Churn: The New Danger

When building is cheap and fast, pivoting becomes tempting. Maybe too tempting.

Here's what it looks like: you launch something, have a few conversations with customers that don't go great, and two weeks later you've rebuilt around a different hypothesis. Then you get mixed signals from a handful of other customers, and the cycle starts again. New direction, new build, new conversations, more ambiguity. Repeat.

This pattern looks like responsiveness. It feels like learning. But it can function like drift — a kind of motion that substitutes for the harder work of actually developing conviction about your customer.

The problem isn't the pivoting itself. The problem is the signal quality driving it. A handful of customer conversations isn't a pattern. Early feedback is noisy. Founders who move on every time they hit friction aren't necessarily learning faster — they might just be avoiding the discomfort of staying in one place long enough to find out if they're right.

This is where the Miracle Mile comes back.

---

The Guides Knew Something

Our guides on the North Platte weren't rowing us back upstream because they were stubborn. They weren't ignoring the possibility that the hole was empty. They were making a judgment call — informed by experience, by the water conditions, by how the fish had been behaving — that this particular hole had fish, and we hadn't found the right presentation yet.

That's a different thing from refusing to leave because you've already spent time there.

Rowing back upstream with an adjusted approach is not the same as holding on because you can't bear to leave. One is conviction. The other is the sunk cost fallacy.

They can look similar from the outside, but they come from different places — and they lead to different outcomes.

Founders navigating a market need that same kind of calibration. When customers aren't responding, the question isn't just "should we pivot?" It's "are we in an empty hole, or haven't we found the right rig yet?" Those are genuinely different situations, and the answer shapes everything about what you do next.

---

The Modified Lesson

The original lesson from Strongroom still holds: the sunk cost fallacy is real, and it will come for you. When you've invested years in something, the psychological gravity is powerful. Don't let past investment be the reason you keep going down a road that isn't working.

But the AI era adds a layer.

Sunk cost pressure may be lower, but conviction pressure is higher.

Just because you can build and pivot quickly doesn't mean you should. Speed reduces the cost of being wrong, but it doesn't replace the work of understanding your customer deeply enough to know when you're wrong versus when you just need to adjust your rig and try again.

The guides on the Miracle Mile didn't stay in every hole forever. They knew when to move. But they also knew the difference between a hole worth rowing back to and one that was genuinely done. That judgment — developed over years of reading water — is exactly the kind of thing that doesn't get faster just because the tools improve.