When Ed-Tech Isn’t Ed-Tech

And why there are so many failures in ed-tech…

So, why did the hottest ed-tech start-ups from a decade ago — Schoology, Canvas, Edmodo — fail?

At IBM, an SVP once asked me (with that rhetorical annoyance that some SVP-types have) why ed-tech doesn’t scale like med-tech. To further complicate the question, he was specifically referring to China.

Education and healthcare have a lot in common, but education generally has two complicating factors.

1. Balkanization of funding (and related terms). Go from district to district, state to state, university to university, and you can’t always predict exactly who (or how) is paying for the product or what the terms may be. Idiosyncratic is a kind way to describe education funding and deal terms. Exactly who you’re invoicing and who’s receiving the services isn’t always the same entity, and term-sheet cacophony only gets worse from there.

2. Balkanization of standards (and customization). Again: district to district, state to state, university to university, and everyone’s got their own list of standards, requirements, regulations. Keep in mind we’re talking about deals within the same country. Going from country to country, standards, learning metrics/outcomes, etc., can differ wildly.

If you and I go to a bookstore and buy a “same” U.S. history book, we’ll both actually buy the same book and make the purchase with the same terms. That’s a scalable business.

But the U.S. history textbook you sell to Texas can’t be sold to California. You’ll need to negotiate changes, then negotiate terms, and, at the end, you fundamentally have two different deals. This causes friction, which translates into cost, time, resources.

In the end, these complicating factors massively drive up prices in education, but it also means something else: none of this is fundamentally scalable. Any business that has at least one of those two education-specific issues is an education company … and those issues means it’s not scalable. Which also means they’re terrible businesses. (Is there a textbook publisher that hasn’t gone out bankrupt at some point?)

Though med-tech funding and deal terms can get wonky, there’s a hard backstop to med-tech variability: the human body. The universal truths of biology mean that there’s only so much possible variation within healthcare solutions. The Chinese may have idiosyncratic ideas about A.I. image rec for cancer, but, at the end of the day, we’re still talking about the same cancer that Canadians get. That’s not true in education.

So who’s ed-tech and who’s not?

Google Apps for Ed (or “G-Suite for Education”) isn’t “ed-tech.” Google has entirely avoided the funding and standards issues in education. (So ‘Apps for Ed’ is just ‘tech’ not ‘ed-tech,’ which means it’s scalable — which is why it has scaled.)

Municipal Business Model

But what about, for example, Schoology? It’s an LMS that doesn’t really deal in the idiosyncrasies of education standards or even, generally, education funding/terms?

I’d argue that Schoology’s industry sector isn’t education but rather government. That’s an important distinction though there’s much overlap. Schoology was in something like 1,800 schools/districts with average revenue of about $19,000 per,[1] and yet they were never profitable (just as Edmodo, Canvas, etc. were never profitable).

Clearly, Schoology needs scale. Generally, when you need scale with municipal contracts, your entire business is driven by one thing: municipal budgets. That’s all. You can have a $2 value prop, convince everyone your product is worth $2, and demonstrate that your product is a great $2 value — but if the budget only has $1, then that’s all you’re getting. If $1 contracts aren’t profitable for you, then you wind up on the auction block.

Given municipal budgets (city, school, waste treatment plant — whatever municipal entity you’re selling to), you must build your entire business around being profitable at the typical budgeted amount at the smallest scaled customer. Most school districts in the U.S. are small (<5k students total), and those generally have no restrictions below $10k and restrictions above that amount; this means if you want to scale across all those districts (e.g. your broadest customer base) with minimal friction, you need to build a product that you can sell profitably for under $10k/year. Everything above that amount creates more friction and less scalability. The obvious chicken/egg issue becomes: more friction = more cost = higher prices = more friction. (Friction in government contracting means local salespeople, lobbying, state/local funding/reg experts, etc.)

A lot of ed-tech companies are really in the government contracting sector and don’t realize it, which means the feasibility of your entire business is built around its cost structure.

Retail Business Model

What about MasterClass or Coursera? They generally have no standards or funding idiosyncrasies and no municipal contracts.

Such companies are mostly or entirely general public end-user transactional (e.g. “come into my store and buy something”), and are therefore consumer retail companies. Their KPIs, customer acquisition costs/metrics, scalability are generally similar to other retailers. An online shoe store and MasterClass have roughly the same business model. (The substantive difference — if there is one — is likely in user acquisition costs; I suspect a lot of online retail ed requires repeat business to be profitable. Subscription services best survive high acquisition costs, so 1x purchase/retail with high acquisition costs are high-failure rate businesses.)

Retail means a possibly quick revenue ramp (unlike ed or municipal) and flexible pricing, but requires lots of marketing/hype/sizzle and suffers from low stickiness (with high CAC, which is lethal) and fickle customers. (In contrast, if you sign a municipal contract, your software can totally suck and the municipality will still be paying you for it a decade later … just ask IBM.)

The cases above define industries by the kind of business plan and financials they’ll produce, the way products are developed (e.g. whether the product is developed to solve certain educational problems or to be profitable at certain budget limits), the time required to become cashflow positive, and the way that sales are handled. None of it really considers exactly what the product is — which is the source of the confusion. People think “it’s educational so we’re in education,” yet that has nothing to do with the business model. Netflix may have documentaries, but they aren’t ed-tech. Further, while “educational” implies certain user experience features, most such features are the same or similar to other UX-dependent applications. (Are the UX considerations for Edmodo or Schoology substantially different from Facebook?)

Finally, surveying the landscape of failed ventures — Grockit ($45M raised, now dead), Learnist (“Pinterest for education,” dead but revised as Seesaw), Livemocha (hottest language learning app of 2010, murdered by Rosetta Stone), Infrastructure/Canvas (hobbled profit-less into private equity), Edmodo (auctioned to the Chinese), Schoology (snared by bottom-feeding PowerSchool) — there are other lessons aside from knowing which industry you’re really in.

First, an advertising business model doesn’t work, which we learned back in the 1990s from Channel One — too much friction (ads in ed), not enough scale. Ten million MAUs won’t drive enough advertising revenue. You just got to charge the people who use your product. (And no, selling data won’t work, PowerSchool.)

Second, selling into “education” is massively resource intensive and usually only worth pursuing (if at all) if the contract is very long (or, there’s a ridiculously massive moat around your contract). It’s conceivable that SIS’s are a reasonable business (they’re not, which is why PowerSchool gets passed around), but the average district holds onto their SIS for 18 years. To some degree, this is the cable model — very high user acquisition cost but decades of revenue. (For this reason, teaching & learning/classroom products are usually bad businesses whereas ed infrastructure can be good businesses.) Textbooks attempt this moat strategy (which is the only reason they might make sense) — textbook publishers aren’t in the content business; they’re in the standards/regulations business. And all those regs create the moat that justifies all that friction.

So why did IBM’s Watson Education fail? It was pitched as the impossible — a scalable education business.

[1] It was never really clear to me whether Schoology was referring to schools (90k+ of them in the US) or districts (14k+).