Pivot gates

Most pivots don’t fail because the new direction is wrong. They fail because the decision to pivot happens too late, made by someone too invested in the old answer to read the data straight.

A pivot gate fixes the timing, not the direction. It’s a checkpoint you set before you have a stake in the outcome: a metric, a date, and an action you’ve already agreed to take if the metric misses. When the date arrives, the gate does the deciding. You just execute.

Why gut calls fail here

Founders are trained to push through resistance. That instinct is right most of the time, which is exactly why it fails at the one moment a pivot is actually warranted. Sunk cost doesn’t feel like sunk cost from the inside. It feels like conviction. “One more quarter” is rarely a plan. It’s a way of not deciding yet.

The problem isn’t a lack of data. Most founders I know can recite their numbers cold. The problem is that the person who set the strategy is also the one grading it, in real time, under pressure, with identity wrapped up in the outcome. That produces predictable bias. You round the ambiguous case in your own favor. Everyone does.

The anatomy of a gate that has teeth

A gate that actually works has four parts.

One to three metrics, chosen in advance, that are hard to argue with after the fact. A simple pivot can run on one number. A complex enterprise pivot usually can’t — collapsing “automation,” “unit economics,” and “customer risk” into a single score just hides which assumption actually broke. Better to name each one and gate it separately than to average them into something vague enough to be argued with later.

A fixed date, far enough out to be fair and close enough to matter. Six weeks for a scrappy test. One or two quarters for something that needs real distribution to show signal.

A pre-committed action for the miss case, written down before you see the number. Not “we’ll reassess” — reassessing is what got you here. The action should be specific enough that a stranger could execute it: kill the initiative, cut the team by half, revert to the prior model, whatever it actually is.

Someone other than you who can hold you to it. A cofounder, a board member, an advisor with no emotional stake in the bet. Their job isn’t to make the call. It’s to notice when you’re about to move the goalpost and say so out loud.

Measure the assumption, not just the available number

A gate is only as good as what it actually tests. Missing a sales target can mean the thesis is wrong. It can also just mean distribution is slow, which is a different problem with a different fix. If the gate tracks whatever number happens to be sitting in a dashboard, you’ll get an answer, but not necessarily the answer to the question you needed asked.

The fix is to write the gate against the assumption you’re least sure of, not the metric that’s easiest to pull. If the pivot depends on work being standardizable, gate the standardization rate, not just revenue. A miss on revenue tells you the market is slow. A miss on standardization tells you the pivot’s premise doesn’t hold, which is a much more serious result wearing the same shape as a bad quarter.

Where gates break

Vague gates are the most common failure. “If we’re not seeing traction by end of quarter” isn’t a gate, it’s a mood. Write the number down.

Gates without consequence are the second. If the pre-committed action is “we’ll have a conversation about it,” nothing actually changes hands when the date arrives, and a checkpoint with no consequence isn’t a gate.

The quiet failure is the one that eventually catches almost everyone: moving the goalpost after the number comes in. It lands at sixty percent of target and the story becomes “but the trendline is improving” or “but two enterprise deals are close.” Maybe those things are true. They’re also exactly what you’d say if they weren’t going to save you. The entire point of setting the gate in advance is that future-you doesn’t get to renegotiate with present-you’s judgment.

Where this showed up for us

We are now running Routespring’s shift toward automated airline crew travel against three gates over a one-year window, not one: automation progress, measured as manual minutes per booking; margin, tracked as gross margin on the automated flow; and customer risk, tracked as concentration in a single carrier relationship.

These three aren’t symmetrical, and treating them as if they were would hide what each one is actually testing. Automation minutes and margin are operational outcomes we control directly. Customer concentration depends heavily on sales timing that isn’t fully in our hands — a slow quarter landing a second airline doesn’t necessarily mean the automation thesis is wrong. It may mean something narrower but still serious: that Routespring can’t build a viable independent company around this on the current timeline, even if the underlying automation works.

We’ve also written down the consequence of each miss, not just the measurement: whether to keep investing, narrow the product, reduce the team, or stop treating this as Routespring’s primary path.

The more important gate came earlier and wasn’t financial at all. At ninety days, the question was narrower: does this category of booking work actually standardize enough to automate, or does it stay irreducibly manual no matter how much process gets thrown at it? That’s the falsifier for the whole pivot. A soft first-quarter revenue number would have meant distribution was slow, which is a normal startup problem with normal startup fixes. Discovering the work resisted standardization would have meant the premise was wrong, which no amount of distribution fixes. The ninety-day gate was built to catch the second failure mode before a full year got spent chasing the first.

The work did standardize enough to justify continuing, which is why the one-year structure — automation minutes, margin, and landing a second contracted airline to bring concentration down — is the one running now.

The gate isn’t the pivot

The gate doesn’t make the decision easier emotionally. It makes it possible to make cleanly. You still have to live with cutting a team, killing a product line, or telling investors the story changed. What the gate buys you is certainty about when that conversation happens, and protection from the version of yourself who’d rather have one more quarter of ambiguity than one clear answer.

Set the gate before you need it. By the time you need it, you’re the wrong person to set it.