Every city has one: the temporary program in its ninth year. The trial parking rule still printed on a laminated sign. The interim fee schedule that outlived the committee that wrote it. Pilots are supposed to be experiments, and experiments end. In practice, a pilot without a written ending becomes a permanent program that never had to win the argument for its own existence.
This is a reference piece: what an exit plan contains, two national examples of pilots that ended properly, and the standard ways a trial drifts into a fixture. The core discipline fits in one sentence. Define the test period, metric, and decision date at launch.
What a pilot is actually for
A pilot buys information. That is its entire product. It answers a question the institution could not answer from a desk: does the bike lane slow emergency response, does the new intake form reduce errors, does anyone show up on Thursdays. If the question is never written down, the pilot cannot answer it, and whoever likes the program most will declare victory using whatever numbers happen to flatter it.
A street closure trial needs criteria before the first barrier is placed. Afterward is too late, not because people are dishonest, but because by then everyone has a stake. Merchants have adjusted, opponents have organized, and the staff who ran the trial would prefer not to have run it for nothing. Criteria written in advance are the only version anyone will trust.
Stockholm wrote the ending first
The cleanest large-scale example is Stockholm's congestion charge, proof at city scale that even small fees change behavior. The city ran it as a trial with a hard stop: charges on from January to July 2006, seven months, then off. Traffic across the charging cordon fell by about 20 percent during the trial. That September, residents voted in a referendum with the results in hand, and the charge came back permanently in August 2007. Transport researcher Jonas Eliasson's overview of the Stockholm charges documents the arc, including the detail that makes it remarkable: public opinion started fiercely hostile and ended with roughly two-thirds support.
The exit plan did the persuading. Because the trial genuinely ended, skeptics got to experience both worlds within a single year, congestion priced and congestion free, and then choose. A permanent launch would have been fought as a tax grab. A trial with a real off-switch was allowed to prove itself.
Finland stopped on schedule
Exit plans matter just as much when the result is mixed. Finland's basic income experiment paid 2,000 unemployed people 560 euros a month, unconditionally, for exactly two years, 2017 and 2018. When the clock ran out, the payments stopped and the evaluation began, comparing recipients against a control group. The published results showed small employment effects and clearer gains in reported wellbeing. Advocates wanted an extension. They did not get one, and that refusal is why the findings are credible. A pilot that cannot end cannot conclude anything.
The four parts of an exit plan
Strip away the branding and every sound pilot answers four questions before launch. This table is the reference card.
| Part | Decided at launch | Sign of drift |
|---|---|---|
| Question | The single uncertainty the pilot exists to resolve, written as a sentence anyone can repeat. | Six months in, supporters and critics describe the pilot's purpose differently. |
| Metric | What gets measured, how, and the baseline it will be compared against. | The success measure is chosen after the data arrives, and it happens to look good. |
| Deadline | A calendar date when the trial stops and a named body decides: scale, revise, or shut down. | The decision meeting is postponed, then postponed again, while the pilot keeps running. |
| Unwind | Who removes the barriers, reassigns the staff, and notifies participants if the answer is no. | Shutting down would strand hires or contracts, so shutting down is quietly off the table. |
How zombie pilots happen
Nobody sets out to run a nine-year trial. The drift follows a pattern. First the pilot launches with enthusiasm and a vague charter. Then the review date slips because the data is incomplete, which it always is. Meanwhile the program acquires dependents: two staff positions, a vendor contract, a constituency that likes it. By year three, ending the pilot has real victims while continuing it has none, so continuation wins every meeting by default. The label stays because removing it would trigger the scrutiny a permanent program requires. The word pilot has become armor against evaluation rather than a commitment to it.
Notice that no step in that chain involves bad faith. That is what makes exit plans necessary. They are not a defense against villains. They are a defense against the ordinary gravity of institutions, which pulls every temporary thing toward permanence unless someone bolted a deadline to it at the start.
There is also a vocabulary trick to watch for. When a trial's deadline gets awkward, the program rarely admits to going permanent. It gets renamed. The pilot becomes phase two, then an initiative, then simply the way things are done. Each rename resets the clock without ever forcing a decision. If you cannot find the meeting where a body with authority voted to make a temporary thing permanent, it never happened, and the program is running on inertia rather than consent, the same decay that sets in when shared rules go unmaintained.
The remedy does not require cynicism about the program itself. Plenty of zombie pilots are good programs that deserve to exist. What they are missing is legitimacy: the moment where the institution looked at real results and chose. Scheduling that moment retroactively, a review with published criteria and a genuine option to stop, is usually enough. Good programs survive the vote. The ones that dodge it were telling you something.
Questions to raise at the launch meeting
If you sit on a board, a council, a PTA, or a condo association, you will eventually be asked to approve a trial of something. Say yes if you like, but ask four questions first, and get the answers in the minutes, because trust is built in the follow-up. What exactly are we trying to learn? What number tells us, and what is it now? On what date do we decide, and who decides? And if the answer is no, who takes the barriers down, and what does that cost? A proposal that cannot answer these is not a pilot. It is a launch asking for a nicer name, and it should be debated as the permanent commitment it intends to become.