Say a district has three high schools, and one of them spent last year piloting a new approach to ninth grade. The pilot organized teachers of the ninth-grade core courses into teams, gave them a shared planning period, and paired each team with a counselor. The teams met weekly to look at attendance and grades and to step in before a student fell too far behind. The results were as good as anyone hoped. Course failures among ninth graders dropped by about a third, attendance improved, and the teachers involved want to keep going. This spring the superintendent will recommend to the board that the model expand to the other two high schools in the fall. The board will ask what the expansion costs. Before the recommendation ever reaches the board, the people writing it have a harder question to answer: what exactly would the district be expanding?
Assume the pilot worked, that the comparison to the previous year is fair, that nothing else changed at that school, and that the drop in failures is real. The pilot asked whether the model could work. The expansion asks whether the district can make it work at two schools with different staff, different schedules, and different students. That question can get lost when the recommendation focuses on the pilot's results.
What a pilot is evidence of
A pilot shows that something can work in the school where it was tried, with everything that school had at the time. A recommendation that reports only the results leaves out what was in place to get them.
In this case, the pilot ran at the school whose principal asked for it. The teachers on the ninth-grade teams volunteered, or were chosen because the principal knew they would make it work. The shared planning period was carved out of a master schedule the principal rebuilt over the summer. A coordinator was funded for one year through a grant. Someone from the district office was in the building often enough that the teams expected it, and sat in on the data meetings. And the model was new, which meant it had the energy new things have.
None of those are flaws in the pilot. Any of them may have helped produce the result. For each one, the next school will either have it, get it in some other form, or go without, and the recommendation should say which. Nobody yet knows which of them the result depended on.
Research on how programs are put into practice lands in the same place. A large review of prevention and youth programs found that programs implemented well got noticeably better results than programs implemented poorly, and that how well a program was implemented depended on things like whether staff believed it was needed, whether someone championed it, and whether leadership visibly supported it. At the pilot school, all three were true. The recommendation needs to address how the district will build that commitment and support at the next two schools.
What actually made it work
It would be easy to conclude from all this that no pilot result can be trusted. I do not think that is true, and it would not help anyone deciding what to do next. Some things matter a great deal for one kind of program and hardly at all for another. The work of an expansion decision is figuring out which ones made this model work, one at a time.
Having teachers who volunteered for the pilot may matter when a program depends on what teachers do together, and this one does. The teams work because the people on them use the planning period to act on what they see. The same review notes that school staff pressured by the superintendent's office to offer new programs often implemented them poorly, citing a lack of commitment as a likely explanation. Assigning the model to a school is a different starting point from a principal and teachers asking for it.
The coordinator's work matters if it was part of what made the model work. If the coordinator pulled the data, scheduled the meetings, and followed up with families, that work was part of the model itself. The next school may not need the same position, but somebody there has to do that work and have the time and authority to do it. An expansion that drops the position without assigning the work is expanding something else.
District attention can be difficult to sustain as the work expands. One school can have someone from the district office in its data meetings most weeks. Three schools make that harder, and it thins out from there.
The school itself matters when the model relies on something the pilot school had. A model that depends on a stable teaching staff may need additional support at a school with high turnover. A model that depended on a principal rebuilding the master schedule may require more scheduling support under a principal in her first year.
So do the students. A school with more ninth graders who enroll midyear, more students whose attendance has been interrupted, or more families who need to hear from the school in a language other than English may need more or different kinds of follow-up after each weekly meeting than the pilot school did. The meeting can look identical in both buildings while the work that comes after it is very different. The district should expect as much of those students as it did of the pilot school's, and provide the additional support they need to get there. Giving the next school the same resources is not necessarily giving its students the same opportunity.
When the extra help goes away
A pilot can look stronger than what comes after it when the extra help disappears as the work expands. Cynthia Coburn, reviewing research on taking school reforms to scale, pointed out that externally developed reforms often arrive with a short burst of resources, training, and outside assistance, and that the help fades as the people providing it move on to other schools. Of the publications she reviewed, only one explicitly examined schools after the period of extra resources and attention had officially ended. That leaves an important gap: results from the supported years do not tell us whether the reform lasted after that support ended. For a district, the practical lesson is to plan for the grant money to end and the first-year energy and concentrated attention to fade, and to decide before expanding how it will replace what matters, whether that means funding a position, protecting time in the schedule, or keeping someone from the district office involved.
California shows the same problem at a much larger scale. When the state reduced class sizes in the early grades in the late 1990s, smaller classes did raise reading and math achievement, according to a study of the program by Jepsen and Rivkin. But the share of teachers with neither experience nor full certification went up as the state filled all those new classrooms, and that weakened the benefit, particularly in schools with high shares of economically disadvantaged minority students. The idea worked. What ran short was experienced, fully certified teachers for all the new classes. A statewide mandate is a long way from one district adding two high schools, and I would not borrow its numbers. The question still applies, though: what might this district run short of when one school becomes three?
A stronger case for expanding
So what would a stronger recommendation look like? It does not necessarily need a longer pilot, and it does not need to recreate every circumstance from the pilot school. It needs four things, using the ninth-grade example.
First, look for differences inside the pilot. If four teacher teams ran the model, compare their results and how they carried out the work. Where results differ, look into what might explain them. Differences might come from how the teams worked together, from which students they had, or from chance, and four teams may not be enough to tell for sure. That comparison can help the district identify what to preserve and what to watch as the model expands.
Second, decide which pieces of the work continue and say so in the recommendation. The shared planning period either goes into the master schedule at every high school as a standing cost or it does not. The coordinator's work either becomes a funded position or real, protected time in someone else's job, such as an assistant principal's, or it does not get done. These are real choices with real costs, and they change what is being expanded. A recommendation to "expand the ninth-grade model" that does not say which pieces are funded leaves the board voting on a name.
Third, choose the next school because it differs from the pilot in a way that matters to the model. Going next to a school much like the pilot can help establish whether the result can be repeated. A school with a relevant difference can tell us more about what the district will need to adapt. If the model depends on teams acting on data together, the useful difference might be a school with high teacher turnover, a schedule with no room for a shared planning period without giving something up, or more students who will need follow-up. Whether the staff asked for the model is one such difference, but it is not automatically the most important one. Whichever school is chosen, it should get what the model needs to work there, and more if its students need more. If the next school gets neither the shared planning time nor someone to do the coordinator's work, a disappointing result will leave the district unsure whether the approach failed or those missing supports explain the difference.
Fourth, watch whether the model is actually happening, along with the results. Results such as course failures arrive in June. Whether the teams are meeting every week, using the shared planning period, and following up with students and families can be checked in October, and if those things are not happening, the June results will be hard to interpret.
Waiting has a cost
Every semester the other two high schools run without the model is a semester of ninth graders who might have been caught earlier. That cost is easy to overlook in a conversation about how carefully to expand, and it should shape the recommendation.
If the district can provide what the model needs at both schools, including the staff, the schedule, the coordinator's work, and the support, then making either one wait is hard to justify. If it can provide those things at only one, I think supporting one school properly is likely to do more good than stretching the same resources across two, and it will tell the district more about whether the model can work beyond the pilot. What the cost of waiting does not justify is expanding on the strength of the pilot alone, without deciding what goes with it. If the results do not hold, the district may then have to rebuild the support it left out, with another group of students waiting while it does.
What the pilot proved
The pilot proved something real. It proved that this model can cut ninth-grade course failures when a school has a principal who wants it, teachers who volunteered or were selected for the pilot, time built into the schedule, someone doing the coordinator's work, and the district paying close attention. It did not prove that the district can provide those things at two more schools, or that the model works without them.
That is the decision actually in front of us. Expanding a pilot means deciding whether the idea is good, and it also means deciding which staffing, time, attention, and follow-up the district is prepared to provide, at each school, for as long as the model runs. If we cannot name those things, we do not yet know what we are expanding, and the board is being asked to approve a result instead of a plan.
For the ninth graders at the next school, the expansion will come down to whether someone notices in October that a student has stopped coming to second period, whether a teacher has time to bring it to the team, and whether someone calls home in the language the family speaks. The support the district commits to is the support those students will have.