EdTech Marketing Playbook

In EdTech, decide who pays before you decide what to say. A school district, a college and a parent buy with different money, under different rules and on different proof, so pick one buyer and build what that buyer needs first.

EdTech looks like one market and behaves like three. The same reading app can be bought by a district curriculum team, a university department or a parent at 9 p.m. on a phone. In my view, most early EdTech companies stall because they try to serve all three at once. The steps follow the order I would work through with a founder, and where the three buyers differ, the step says so.

In education the user, a student or a teacher, is rarely the buyer, and the buyer is often not one person. Someone uses the product, someone sees the results, someone pays and someone signs off, and in a district those are different people.

The three markets differ in shape more than in size. In 2023–24 the US had 13,335 regular school districts running 99,297 operating public schools^[1], and in 2022–23 it had 3,896 degree-granting colleges and universities^[2]. On the consumer side, about 5.2% of children aged 5 to 17 received academic instruction at home in 2022–23, up from 3.7% in 2018–19, but that count includes students in full-time virtual school; 3.4% were homeschooled^[3]. A district buys through a group, from a budget set once a year; a college adds a technical review run by IT; a parent decides alone, pays by card and can cancel in any month.

Then choose one buyer and stay with it until the sale repeats. A product sold to teachers, districts and parents at the same time usually ends up with three half-built funnels. If you sell to schools, the same test settles whether to start with teachers or administrators: if a teacher can pay and adopt alone, start with teachers; if the money sits in a school or district budget line, start with the administrator who owns that line, and treat teachers as the people who make the pilot work (Step 9).

US public schools spent $869.6 billion on current expenditures in fiscal year 2024, or $17,644 per pupil. Federal money was 11.7% of their revenue, and states and local sources provided the rest in almost equal parts^[4]. So most school purchases are paid from state and local budgets, set once a year.

The pandemic money that filled gaps for a few years is gone. Congress gave schools about $190 billion through three rounds of Elementary and Secondary School Emergency Relief (ESSER) funds^[5]. The last round had to be committed by September 30, 2024, and even approved extensions to pay it out ran no more than 18 months past that date, to March 2026^[6]. District leaders feel the loss. In RAND’s spring 2026 survey of 481 districts, the share ranking budget shortfalls among their three most pressing challenges rose by 21 points in a year, and leaders pointed first to the end of federal COVID-19 relief^[7]. In my view, every renewal now competes with staff and core programs.

Time follows the money. For most states the school fiscal year “begins on July 1 and ends on June 30”; Alabama and the District of Columbia start on October 1, and Nebraska and Washington on September 1^[8], while each Texas district chooses July 1 or September 1^[9]. Districts draft next year’s budget over the winter and adopt it in spring or early summer, so in my experience a product that is not in the draft budget by spring waits a year. The EdTech marketing calendar sets out that year season by season, with state deadlines.

Price a school product in the unit that matches the budget that pays and the way your value grows, and whatever the unit, also show the cost per student: that is how a district finance office compares quotes.

• Per student. Fits a product every student uses, paid from a core curriculum or technology budget. Easy to compare, but revenue falls when enrollment does.
• Per teacher. Fits teacher tools, such as planning or grading, paid from a school or professional-learning budget.
• Per school. Fits when a principal or a school-level fund pays. Some federal money is school-level: districts “target the Title I funds they receive to schools with the highest percentages of children from low-income families”^[10]. Band a flat site price by enrollment, or it undercharges large schools.
• Per district. Fits systems the whole district runs on: one contract, a larger number, a longer approval chain and often a board vote.

In my view, start with a price one school can say yes to, then offer a district licence once several schools in a district are paying, priced so the district saves something by consolidating. Set a minimum annual fee from what it costs you to onboard, train and support one account for a year, so the smallest customer does not lose you money.

Price parents as a separate product, not a derived one. A district pays by invoice, usually a year at a time: PowerSchool, which sold mainly to K-12 districts, reported that “Customers are billed annually in advance, with a small number of customers billed semiannually, quarterly or monthly”^[11]. A parent pays by card from household money, compares you with a tutor, a workbook or another app, and can cancel in any month, so the product has to show progress in the first week. Test parent prices at checkout; for schools, put two scopes at two prices in each proposal and record which one is approved.

Billing a year in advance also sets your payback. A district sale costs months of work before the first invoice, so track two numbers for each district: the cash gap, from first spend to first invoice, and payback, the months of gross margin it takes to earn back what the district cost to win.

A district purchase is decided by a group, usually led by the district office. RAND’s 2026 study of math curriculum adoption found that processes drew input from several groups, but “the final decision was most often made by district-level staff”^[12]. In my view, these are the people to write for:

• Curriculum and instruction leaders shortlist and judge standards fit and evidence. In the same study, principals named alignment with state standards as their top consideration.
• Teachers try the product in class and decide whether it gets used. They judge time saved and classroom fit.
• School leaders want the school’s goals met and staff trained. Most principals in the RAND study bought professional learning bundled from the curriculum vendor, so training is part of the product.
• The superintendent needs a link to district goals, and the board approves larger contracts.
• Finance and operations need a price that renews (Step 3).
• The technology leader vets privacy, security, accessibility and integration, and can stop a deal alone (Step 6).

One message will not serve them all, and two promises will not either: “fun AI” for teachers undersells, and “higher scores” for administrators overclaims unless the evidence is on file. Write one core promise about the learning problem and the teacher’s day, not the technology. Give it three pillars, and under each pillar a teacher line and an administrator line, each with its proof beside it. If two roles get the same sentence, one of them is not being served.

Colleges have a smaller, more technical group: IT, academic leaders and procurement, often after a department has tried the product. Their own agenda leads with value: the 2027 EDUCAUSE Top 10 opens with “Determining where AI adds real value” and puts cybersecurity third^[13]. So lead with an outcome they measure, such as retention, completion or staff time, and with integration into the systems they already run. Parents are a group of one, and in my experience a teacher’s recommendation is the most credible message they hear (Step 8).

Districts want evidence that students learn more, in a form they can file. The Every Student Succeeds Act (ESSA) defines four tiers^[14]. Strong evidence needs at least one well-designed and well-implemented experimental study; moderate, a quasi-experimental study; promising, a correlational study with statistical controls for selection bias; and the fourth tier, “demonstrates a rationale”, a logic model based on research, with ongoing efforts to study its effects. The Department of Education’s own grant rules add a sample of at least 350 students across more than one site for the top two tiers^[15], and school-improvement interventions funded under Title I, section 1003, must meet one of the top three^[16].

Few products have climbed far. In a 2026 analysis of 150 widely used classroom technologies, 40% of the tools built for education had any evidence aligned to ESSA, most of it at the lowest tier^[17]. It is vendor research from one company’s usage data, so read it as a picture of the field, not a benchmark.

For an early company, Tier 4 is the honest start: a one-page logic model linking the problem, what the product does, what should change for students and how you will measure it, each tied to published research, with a study planned or under way. Then climb. Compare users with similar non-users on a measure the school already trusts, adjusted for prior scores, for Tier 3, and later commission an independent quasi-experimental or randomized study. Users’ before-and-after scores show change, not cause. A school can share student records for a study without parental consent only under a written agreement “specifying the purposes of the study and the use and destruction of the information”^[18], so agree the design and the agreement before term starts.

Claim only what the study shows, and have the evidence on file first. The FTC requires “a reasonable basis for advertising claims before they are disseminated”, and if you say “studies show”, it expects “at least the advertised level of substantiation”^[19]. Education claims are not exempt: in 2016 the FTC charged that Lumosity’s maker deceived consumers with “unfounded claims that Lumosity games can help users perform better at work and in school”, and the settlement requires “competent and reliable scientific evidence” for future claims^[20]. Match the verb to the design, name the study and its limits beside the result, and name the tier.

Districts check before they buy. In CoSN’s 2026 survey of 607 district technology leaders, 86% vet digital tools before use, 61% keep a list of approved apps and 56% require vendors to supply information on the safety of their products; 29% ask for information on accessibility^[21]. Four sets of rules sit behind those checks. They are legal obligations, so treat this as orientation and get legal advice for your product.

• FERPA. When a school shares student records with you under FERPA’s school-official exception, you must be under its “direct control” in how you use them. The Department of Education warns that using the data for “targeted advertising directed to students or their parents” “will likely violate federal or state privacy laws”^[22].
• COPPA. For children under 13, the FTC’s amended rule took effect on June 23, 2025, and operators had to comply by April 22, 2026. It requires separate parental consent before sharing children’s data with third parties such as advertisers, unless the disclosure is integral to the service, and a written policy on how long data is kept^[23]. A recording of a child’s voice counts as personal information^[24]. A school can consent in place of parents only for its own educational purpose: data collected with its authorization cannot be used “for any commercial purpose, including marketing, advertising, or other commercial purposes unrelated to the provision of the school-requested online service”^[25]. If parents buy directly, their verifiable consent comes before any data collection.
• State student-privacy laws. California’s SOPIPA bars K-12 services from targeted advertising, building student profiles and selling student data^[26], and many states have their own rules. Many districts ask vendors to sign a standard contract such as the Student Data Privacy Consortium’s National Data Privacy Agreement; signing its general offer lets other districts adopt the same terms^[27].
• Accessibility. Under the Justice Department’s 2024 rule for Title II of the ADA, web content and apps a public school district or college provides, including through vendors, must meet WCAG 2.1 Level AA^[28]. An interim final rule of April 20, 2026 moved the compliance dates back a year, to April 26, 2027 for entities serving 50,000 people or more and April 26, 2028 for smaller ones and special districts^[29].

Colleges often ask for all of this at once through the Higher Education Community Vendor Assessment Toolkit (HECVAT), EDUCAUSE’s standard questionnaire on cybersecurity, privacy, IT accessibility and compliance, which added questions on AI in version 4^[30]. Completing it before you are asked shortens the review.

The practical result for marketing: student data collected through schools cannot feed your marketing. Build your marketing list from educators and parents who opt in to your content, not from students in your product.

AI is now part of most EdTech conversations, and buyers arrive worried. About half of students, and of English, math and science teachers, use it for school^[31], most teens say AI cheating is a regular occurrence at their school^[32], and most districts now have AI guidelines; the share whose guidelines include updated academic integrity policies rose from 32% to 46% in a year^[21]. Figure 7 has the numbers.

Lead with the learning problem and keep the AI in the second sentence. Then answer four questions in plain words: what the AI does, what it does not do (for example, whether it can produce finished work a student could hand in), what data it uses, and what teachers control and can see. Parents want the same answers; the Department of Education’s July 2025 guidance on AI in schools stresses “engaging affected stakeholders, especially parents, in decisions about the adoption and deployment of new technologies”^[33].

Make no claim you have not tested: not “cheat-proof”, and no accuracy figure you have not checked on real student work. The FTC alleged that Workado promoted its AI content detector as 98% accurate when independent testing put its accuracy on general-purpose content at 53%^[34], and its final order of August 2025 bars such claims without “competent and reliable evidence to support the claim at the time it is made”^[35]. The Marketing AI Products Playbook covers AI positioning in depth.

The channel follows the buyer you chose in Step 1.

• Teacher-first. A free or low-cost teacher version builds usage a district can later buy; the Product-Led Growth Playbook covers the motion. Free use still needs a path to the approved-app list and a data-privacy agreement (Step 6), because districts vet free tools too: in CoSN’s 2026 survey, 65% require a review by IT before approving one^[21].
• District-first. Account-based outreach to named districts, meetings at conferences, webinars for curriculum and technology leaders, and references from nearby districts. The ABM Playbook covers the account method, and How to sell EdTech to school districts covers the district sale step by step.
• Cooperatives and regional agencies. Districts can buy from a contract another public body has already competed, and when federal money pays, federal rules encourage it: recipients “are encouraged to enter into State and local intergovernmental agreements” for purchasing^[36]. NCES counts 1,143 service agencies, a category that includes regional education service agencies such as New York’s BOCES^[1], and many run shared purchasing for smaller districts. Rules vary by state: New York districts may “piggyback” on another government’s contract only if it was awarded competitively and opened to other buyers^[37].
• Conferences. FETC in January and the ASU+GSV Summit and CoSN in April fall while district budgets are drafted and set; ISTELive in late June comes after most are adopted, so it builds teacher awareness and next year’s pipeline; EDUCAUSE, in the autumn, is the college event. The EdTech marketing calendar has the next dates.
• Parents and learners. A free trial or free tier, a result the family can see in the first week, and word of mouth from teachers, parent communities and app-store reviews.

What Instructure’s, PowerSchool’s and Duolingo’s filings show

Three companies’ filings, made while they were listed, describe the two routes; Figure 8 quotes them. Instructure, the maker of the Canvas learning platform, and PowerSchool, which sold mainly to K-12 districts, described long evaluations, sales bunched around the school budget year and multi-year contracts renewed every year. Both have since been taken private, by KKR and Dragoneer^[38] and by Bain Capital^[39]. Duolingo, which sells to learners directly, described a free product with ads, a small paying share and growth by word of mouth, and it reports the same shape five years on at a larger scale. It is a language app, not a school product, so read it as a picture of the consumer route, not a template. None of this says how the companies decided internally.

The lesson I take from the filings: the district route rewards patience and renewals, and the family route rewards a free product that people recommend. They need different teams, prices and calendars, which is why Step 1 comes first.

In my view, a school pilot should carry a fee, a fixed scope, a length set by the school calendar, success measures agreed in writing, and the contract price written in before it starts. No reliable public data shows whether paid pilots convert better than free ones, so treat this as judgment, not a rule. A free pilot gets a faster yes but often becomes free use; a paid one proves that a budget exists and that someone who can spend it is involved. A middle path is a pilot fee credited against the first contract year.

• Scope. Small enough to approve quickly: one grade, one subject or a few classrooms, with training included.
• Length. Set from two dates: how long your measure needs to move, and when the district drafts next year’s budget. A pilot that starts in autumn and reaches a decision in winter can still make the spring budget.
• Success measures. One learning measure the school already uses, such as its screener or benchmark test, one teacher measure, such as a short survey, and one usage measure, with the result that counts as success written down before any classroom starts.
• Conversion. The full contract’s scope, price and decision date go into the pilot agreement, so the district can budget without a new negotiation. The person who holds the budget line signs; ask your champion who signs at each size, because the limits vary by state and district.

Classroom onboarding has more friction than ordinary onboarding, because one adult sets up many students at once, on school systems, under privacy rules. Load class lists automatically rather than asking teachers to type them: 1EdTech’s OneRoster standard “defines information about people (users), courses, enrollments, organizations, and grades” and “reduces the manual effort and saves the time of teachers and administrators”^[43], and 43% of districts in CoSN’s 2026 survey include interoperability requirements in their RFPs^[21]. Keep teacher setup inside one planning period, and give teachers a ready-made first lesson that works on school devices.

Setup gets a teacher started; a reason to come back keeps them using it. In my view teachers adopt a tool when it saves time in a lesson they already teach this week, when a colleague they trust uses it, and when the first try works in front of a class. So build training around their first three lessons, not around features: a short session in a staff meeting or planning period, one page per lesson that fits on a desk, and a two-minute video inside the product at the step where it is needed. Ask each school for one teacher who goes first and shows the others, and follow up in week 2 with that teacher’s own class results, not a feature tour. Judge the training by whether teachers reach the activation milestone (see Common questions), not by how many attended.

Then prove use before the renewal. Report the dose your logic model needs, not logins, next to the outcome measure the district trusts, and send that report before the district drafts its budget, not after.

Be careful what the renewal report claims. The US Department of Education’s guidance on evidence sets levels that start from a logic model informed by research and rise through correlational studies with statistical controls to quasi-experimental and experimental designs^[44]. One term of use in one district, with no comparison group, does not support a claim that the product improved learning. In my view the report should show planned-use completion, sessions completed against sessions assigned, beside the district’s own benchmark, and say nothing about cause. The sample engagement How an EdTech Product Wins Teacher Adoption in One District works through this on a fictional district.

What changes by stage?

What mistakes should I avoid?

• Counting teacher sign-ups as pipeline. Free use with no budget owner and no path to the approved-app list stays free (Steps 1 and 8).
• Pricing on ESSER-era budgets. Price for what a district can renew from core funds (Steps 2 and 3).
• Launching the big push in May. By then most budgets are set (Step 2).
• Two promises for two audiences. Keep one promise and change only the lead line and its proof (Step 4).
• Calling a product “research-based”. Name the tier and the study (Step 5).
• Treating privacy as paperwork after the sale. Districts check it first, and student data never feeds marketing (Step 6).
• An open-ended pilot. No measures, no decision date and no price means no contract (Step 9).