Why Ed-Tech Startups Don't Scale (2022)

I’m not anti-screen, I think it’s good if my son uses a computer at home with things like kidpix, scratch, castle of dr brain, etc… But I think devices at school at that age are a potent distraction and I do not trust a school to be thoughtful about the usage of tech (the school I mentioned thinks it’s also a good idea to have some instant messaging app in grade 1 betweeen all the kids).

It takes a crazy amount of time to extensively research and prove things in education for a few reasons:

There is opposition to doing the research because teachers/parents don't want their students to be the first.

It's really hard to make random control groups because schools want to put kids in classrooms by their process, not the whims of the researchers.

It's expensive and mostly controlled by the big granters (ex. Gates Foundation).

Results are often looked at via testing measures (which if you look at assessments, sometimes they were proven by having 400 white kids in an affluent school use, aka not representative).

And so, we have a bit of an issue.

Almost no one teaches math via a spreadsheet even though most adults do. It isn't researched on the implications.

Data science as a class to qualify for credit to get into college is greatly contentious because it isn't proven. Even though it is clearly important to our current era.

In addition, teachers are human and so they don't necessarily teach with the 100% best proven methodologies day in and day out.

So I think it's important to realize that education is human. It's flawed and messy and isn't aligned to the best methods already. Sometimes you get lucky with a teacher that works for you and sometimes not.

I like that I have an automated gradebook and communications tool. I like that I can e-mail students and accept work done outside the room via Google Docs. And it's pretty great to be able to present slides from me or students, or occasionally look at a Desmos or Jupyter notebook. Khan is neat to get an instant read or to allow students to hear a topic explained a different way than I do. Playing a game in Gimkit is a way to break up the monotony of a class. I privately advise students they might want to consider Anki as a way to increase their long term retention of information.

Even though I've named a lot of technology, these things are about 10% of what we do. Even when I'm teaching about technology, a huge portion we do is pen, paper, careful thought, and discussion. I know students aren't trying to game; I know they've produced their own work; I know they've not become over-reliant on a crutch. And beyond this, the fundamental reason why a classroom works is not some assumption of my authority or my use of grades for leverage: it works based on the social connections formed in a classroom between the teacher and students, and between students with each other. Too much time on screens interrupts this.

Of course, I'd love high quality research to help guide me in making better choices in the room, but actionable educational research that has good enough methodology to be believed is rare.

Most of my edtech founder friends were motivated to serve and transform education because it didn’t work for them and they realized it didn’t work for most students.

Codeacademy is a good example of a successful one.

Companies that succeeded, in some cases enabling inner city school children to perform on at a college level, died. CiteLighter is a good example. There have been countless ones in math.

The reason the bad edtech products win is because they are built to survive the incredibly disjointed buying groups who all have, explicitly, their own agendas when it comes to anything in the classroom.

If students & teachers could decide where their classroom’s tech budget went, there would be a revolution in education in half a generation.

But they are dead last on the list today.

I've worked briefly in school where kids had iPads since 1st grade and in my opinion it was a catastrophe. Most obvious problem is simple – it's very difficult to get kids concentrated on anything. Tiny working memory and control over cognitive load makes it especially hard for kids to learn and no distraction helps. And iPad with especially funny, "engaged" and other material meant to be helpful created a tons on these – "Look, how colors change in the animation!", "I wonder what I can do with fingers on screen?", "Can I scroll slower and faster?", "Copmetition! Who is the fastest scroller!", "What math?".

This is an example of a standard problem in enterprise computing where the people with purchasing authority are far removed from the actual users.

Spence signalling[1] won the econ Nobel back in 2001. It says that the cost of signalling in a market for imperfect information ends on the one emitting the signal.

For education and jobs, the signal is your degree (hence the huge career difference between dropping out one week before getting a degree and completing). The cost of the degree will fall on the student.

The employer can easily diminish the hiring problem by pre-sorting applicants by education. Students know there is a large difference in career outcomes between university education required jobs and the ones below that.

Colleges want to extract as close to 100% of this lifetime earning difference as they can.

The main question at this point in education research is "how much" signalling is a component of education versus actually teaching skills.

For tech, almost no certificate provides a comparable signal to a real degree.

1. https://en.m.wikipedia.org/wiki/Signalling_(economics)

Many companies that target PreK-12 also know there is a lot of incompetent admins with boat loads of cash. Pricing gouging is very common.

People are shocked when they learn about 70% of property tax bills goes to the school district, not the municipality or county.

The UK had a number of competitive and teacher-led EdTech AI startups delivering lesson planning, like briskteaching.com & teachematic. They have some inroads with schools and colleges now. And probably would have disrupted up into the procedural level naturally.

However the government decided to give £2m without any open procedure to a new company to build similar tech from scratch, and their PR acted as if this was groundbreaking, while the tech they promised, largely already existed and was in use by many teachers: https://www.gov.uk/government/news/new-support-for-teachers-...

Who wants to invest in teachematic or brisk teaching, when your competition can get millions from the government without process and launch the same ideas over a time frame to starve you out? With the market overcrowded, prices are forced to lower, and many companies will now go bust because of this kind of unnecessary government intervention. Brisk Teaching has already had to make most things free. You cannot really be more cooperative with the system than that, but tough luck they still won't get the £2m despite proving out the idea before those funded.

Shortly we will likely see the institutions in control of education like Ofsted mandate, approve or limit the tooling used, and most startups will die out. Rather than several low cost startups bringing high amounts of value, tax payers have spent millions on salaries for people with no AI background to create duplicate products, and the net value to education for actual students is therefore lowered as a result. So you can't disrupt this procedural level very effectively when the government intervenes early on.

I love everything about this statement. I've seen this from both sides, as a seller of software and buyer/user of software. All of the stickiest changes happen procedural level, usually not through full-blown change but instead moderate tweaks and improvements.

One way I like to think about it is this: if my software is all of the sudden unavailable, what will users do? Can they keep working? Will they keep working?

Specific to education, I would hope that many core school functions like teaching would simply revert to simpler methods or alternatives rather than grind to a halt.

But I assume that the process of convincing the buyers to use your software runs headfirst into confusion over how much work it will be to learn something new rather than stick with (mostly) status quo.

Academic institutions, even very poor ones, have a lot of local leverage through employing academics who are integrated in the community and that is largely lost in distance education.

If you go and try to make an awful school somewhere I think you will get a great deal of assistance compared to if you offer a much better online education to the area built somewhere else.

the problem is that even innovations that do work and can be replicated successfully do not scale because the system as it is is highly resistive to change and that is what limits the potential to scale for any change.

take a look at montessori. an often praised and proven system of teaching and learning that absolutely revolutionizes education, if we could only convince those in charge to adopt it.

This money is used to pay for 13,500 adults (government employees).

Their total compensation is higher than you'd think. You can look it up on the Transparent California web site.

Paying those people seems to be a higher priority than children or learning outcomes.

The problem is not a lack of expenditure.

Flashcards and Anki were incredibly effective for me. I am building a study tool which automatically generates them from lecture slides. Beyond that, it applies other psychology of learning techniques - It breaks up content into manageable chunks you can see your progress on (to boost your confidence of exam coverage) and has an exam-conditions like mode where it shows you relevant slides, then asks you questions and to freely recall by blurting onto a blank page everything you remember. It's then able to give you a realistic judgement of your learning and this way of learning by actively recalling freely is, though more intense, ultimately more effective than either reading, watching or quizzing in the long term.

The part of learning that other flashcard apps miss, even Anki, is that students must be motivated. They must enjoy, visually engage, and be confident as they study. They must feel they are on a sustainable realistic path as they gather knowledge when on the way to their goals. And they don't have to study every day to do that - they just need to genuinely learn and be able to see the state of their progress...

I am excited to be working on EdTech now[Revision.ai] which unveils and detects misconceptions to aid the student in eliminating them. Misconceptions are a massive opportunity - as much as spaced repetition itself [https://www.youtube.com/watch?v=eVtCO84MDj8]. We believe they will genuinely learn, feel comfortable and see their progress, when they overcome misconceptions with tech. Other tools really aren't identifying misconceptions; answering is still a "scale of wrong-to-right". And that's a big opportunity. To truly reduce Test Anxiety by studying and building your confidence. I think Tech can be useful in this way, even without teachers intervening, if enough of the learning content is there.