We sold edtech to school districts and lived to tell about it

Roster started in August 2022 in my apartment in Detroit. I had $14,000 in savings and a teaching pension I wasn't yet drawing from. My cofounder Aisha — a former engineer at an ed-publishing company — joined six months later for 35% equity.

The v1 was a differentiation engine: paste in a worksheet, get back four leveled versions calibrated to specific Common Core standards. Under the hood it ran on OpenAI GPT-4 with a custom fine-tune for elementary math. We charged schools $1,200 per teacher per year, with bulk pricing from $9,800/year for a small school of twelve teachers. Schools could buy directly via Stripe without a district RFP at any tier under $25k — that was deliberate.

§Selling school-by-school

We had three paying schools by Q1 2023, eleven by end-of-summer 2023, forty-two by start of 2024-25 school year. Roughly $640k ARR by December 2024. We raised a $3.4M seed in January 2025 from an education-focused fund. The product stack today is Next.js on Vercel, Postgres on Neon, Stripe for billing, OpenAI plus Anthropic Claude for differentiation generation, Linear for issues, and PostHog for analytics.

The district-level deals — six- and seven-figure contracts covering many schools at once — started in late 2025. As of April 2026 we have 4 district contracts and 211 individual school contracts, total ARR $2.6M.

The first summer almost killed us. June through August 2023 our monthly new-revenue dropped from $14,000 in May to $2,400 in June, $900 in July, $1,800 in August. I'd known intellectually that K-12 had a summer slowdown. I had not internalized what 90% revenue compression looks like when you have $19,000 in the bank.

I took a substitute teaching gig four days a week in July to cover personal expenses. Aisha cut her own pay by 40% for ten weeks. We made it to September because two existing schools, unprompted, prepaid their renewal in June — about $11,000 we hadn't been counting on — because the principal at one of them said, in an email I still have, "Roster is the first thing my teachers haven't quit using by October. I want you to still be here next year."

§The school-board fight

The second-hardest moment was a school-board political fight in a Texas district in February 2025. A board member had read our marketing copy, decided our product was "AI replacing teachers" (it wasn't), and tried to get our pilot canceled. Three teachers in the district wrote letters to the board defending us. We won the vote 4-3.

I almost gave up on the whole district market that night. Aisha talked me down. The same week I read a piece on EHR fights between front-line clinicians and procurement that crystallized what was happening: the buyer is not the user, and the user has more political weight than either side realizes.

The unlock that turned us into a real district-sale company came from a single hire. In April 2025 we brought on a former Director of Curriculum from a 28,000-student district as our Head of District Partnerships. Her name is Eleanor and she had spent fifteen years inside the kind of organization we were trying to sell to.

Eleanor changed how we wrote proposals, who we copied on emails, when we offered pilots, and what we agreed to during summer (almost nothing). She introduced two new line items into every district proposal: a "summer onboarding stipend" — paid by us, modest, allowing district staff to spend July training on Roster — and a "year-zero exit clause" giving the district a clean way out if the pilot didn't meet specific adoption metrics by spring.

The stipend cost us $4,000-12,000 per district. The exit clause cost nothing because we never triggered it. Combined, they took our district close rate from roughly 12% to 41% over four quarters.

The broader lesson: districts buy on trust signals and political cover, not just product fit. Eleanor knew exactly which signals mattered. None of them were features.

§What the third school year unlocked

By September 2025 — the start of our third district school year — Roster had thirty-one signed districts covering 412,000 students. The breakthrough was not pedagogical and it was not technical. It was a single operational change: we hired Tameka, a former Title I federal-funds compliance officer at a Texas ESC, as our Head of Procurement. Her job was to make sure every Roster contract was pre-mapped to a specific funding source — ESSER III, Title IV-A, IDEA Part B, or general fund — *before* the superintendent saw it.

Average sales cycle dropped from 11 months to 4.5. Win rate on procured deals went from 22% to 61%. We stopped losing deals to "we love it but we cannot find the budget" and started losing them to actual competitors, which was a much better problem. ARR hit $7.4M in March 2026 with twenty-eight people. We made one other hire that mattered: a former Clever integrations engineer named Pedro who rebuilt our OneRoster and SAML stack so that a district IT director could deploy us in under three school days.

1. 1
   Summer is a real fiscal event. Plan cash for it from day one.

   If you're going into K-12 SaaS, model 90% revenue compression for ten weeks. If your runway breaks under that, you don't have runway, you have a wishful budget.

2. 2
   Sell at the school level before the district level.

   Sub-$25k school deals close in weeks. Sub-$25k deals also produce the case studies that win district deals.

3. 3
   Hire from inside the buyer, not from sales.

   A former curriculum director closes more district business than any AE we ever interviewed.

4. 4
   Political cover is a feature.

   A "year-zero exit clause" is not a discount. It is a piece of paper a curriculum director can hand to a skeptical school board member. Ship the paper.

5. 5
   Teacher word-of-mouth is the only marketing channel that matters.

   Three teachers wrote letters to a Texas school board on our behalf. No paid campaign would have done that.

What this looked like in practice.

The thing nobody tells you about *selling to school districts* is that the buyer is the procurement officer, the user is the teacher, and the champion is the principal — and all three need a different artifact from you. We started shipping a "principal one-pager" alongside every demo and our champion conversion rate doubled.

The thing nobody tells you about *funding sources* is that the right line item turns a "no budget" into a "yes by Friday." ESSER III had a deadline of September 30, 2024 for obligation. We shipped a deal-acceleration pipeline that quarter that was 4x our previous record.

The thing nobody tells you about *teacher PD* is that you are not selling software, you are selling Tuesday-afternoon professional development time. If a teacher needs more than forty-five minutes of training to use your product, you have lost.

The thing nobody tells you about *district pilots* is that the pilot is a feature, not a stage. We turned the pilot into a productized 60-day onboarding with hard success criteria. Districts paid 40% of contract value upfront for the pilot and renewal rate moved from 67% to 91%.

The thing nobody tells you about *summer in edtech* is that nothing closes in July. We treat June through mid-August as a hiring and product-build window and accept that pipeline will compress. Forcing deals into July destroys trust with district staff who are, correctly, on vacation.