Why Education Startups Do Not Succeed (2011)
avichal.com
avichal.com
My impression was that the "educational" software was extremely crude and one-dimensional, basically glorified flash cards. Today, the "educational" software is Web based, and more flashy, but still retains that lack of breadth. It's hard and expensive to write interactive software, so a lot of the apps are basically one or two templates, with different sets of data / parameters for different lessons.
Here's a picture. Move your mouse around. When something lights up, click on it, and a little box of text will pop up for you to read. I kid you not. This is real. Today.
Now I have two kids who are in high school. We have acquired (can't say bought in most cases) piles of educational technology, yet little or none of it was "EdTech." The most glaring example was Microsoft Office. The kids used it to create reports, presentations, drawings, etc. It has been supplanted by Google Docs, which they now use heavily, including for collaborative assignments.
We've played with Jupyter/Python. My son is learning solid modeling using some free app. One of them uses DuoLingo daily, as does my spouse.
Some of their educational technology is still in the analog domain: Musical instruments. ;-)
I think the take-away for me is that we are providing our kids with technology that supports their education, but it isn't written for kids. It's the same stuff that grown-ups use. Jupyter/Python is an example -- it's my primary computing tool for my job. Another common feature is that these are creativity tools, not consumption tools.
This is why a lot of programming is taught through graphical games. I'm not knocking Minecraft style programming lessons, they can be very motivational and build up programming skills and thought processes.
Unfortunately, I'm not that kind of programmer. I'm more the type who enjoys scikit-learn and kaggle than xbox. Even as a kid, I was more interested in programming to know who had the advantage in 3v2 rolls in Risk than programming a game itself (in other words, I wanted to program to get insight into a game, not to play the game itself).
But I wanted to introduce my 12 year old to programming, so rather than trying to learn about Minecraft, I helped him install jupyter notebook and showed him how he could automate some of his math homework. We wrote a method to calculate mean, median, range, that sort of thing, and he learned how to feed a series of lists into it, and graph it.
Believe it or not, this pleased him a great deal. He left feeling slightly buoyed, like he'd gained an edge.
But if programming remains an add-on to a school curriculum, who wants to do more math in an after school program, or a summer camp? So "learning to program for kids" will probably remain heavily focused on graphics and games. Which may be ok, my interest in simulating a dice roll rather than programming a game per se may not be reflective of how most kids (who have great potential) are wired.
And honestly, programming Minecraft is among the best of the educational software. Zoombinis was pretty cool, too, even if it was purely a game. Truth his, much educational software just re-teaches math, or history, but with buzzers and rewarding little cartoons for wrong or right answers.
For example "Human resource machine" is entirely visual but if you happen to click 'export' you get a very basic assembly language style output. If you never clicked export you wouldn't even realise you were "programming", you were just solving the problems on screen. The game does at time explain the analogies used. If you prod the people they'll describe linked lists but such understanding isn't required to progress.
Moving across the spectrum a little there are games like TIS-100 and Shenzhen I/O which do involve programming but with very small instruction sets and graphical debugging and feedback.
These aren't aimed at children so are more complex than would be suitable for children but what I'm trying to get across is that they don't specifically teach programming. You won't come out of it knowing python or C. You won't know about stacks or function calls. But they do teach problem solving and the run/explode/debug loop of getting an instruction set to work on different sets of data.
By being games they gameify the process of wanting to reach acheivements while also solving the "what do I build?" aspect of more freeform/creative programming learning which can often be a dead-end without a sense of aiming toward a goal.
I think a Human Resources Machine style game aimed at a younger audience could be greatly impactful on teaching a programming mindset.
I think kids would ultimately have more fun with one or more Smullyan books.
They should however teach the basics as they do teach math.
I'm thinking the very basics of discrete mathematics. Like, we can make lists of things, then we can do things with lists, like adding, intersecting, indexing, traversing...
Honest question: what's wrong with Skinner boxing kids to get them addicted to memorizing an introductory corpus of data? At some point you want them to get creative, but creativity is impossible without some sort of foundation.
My n=1 anecdote is that I now know way more about geography than most of my peers, largely thanks to playing Carmen San Diego 20+ years ago. That game had me look up facts in an accompanying reference book in exchange for a little cartoon reward. I played it many times and eventually retained the facts.
PS: Kudos on teaching your kid to use jupyter notebooks! I taught myself how to do stats in college by implementing it in lisp.
My favourite software was The Crystal Rainforest. I think this was for 6-8 year olds. There was a long story (4 floppy discs in 1992!) leading through a rainforest, solving puzzles using increasingly complicated computer programming along the way. Here's a screenshot of rebuilding a rope bridge using a simplified LOGO/turtle graphics: [1]. (There is a playthrough on Youtube, which has skipped all the puzzles. I think someone missed the point...)
I suspect the general market for software has increased so much, that the money to be made on educational software is no longer appealing. It's probably easier to cut out the learning and sell a game, rather than made a few hundred or thousand sales to schools.
[1] https://images-na.ssl-images-amazon.com/images/I/51UoAfDY57L...
When it was my turn to play on the computer at school, it was a BBC micro, and the game "grannies garden".
I've recorded a playthrough of the demo of the sequel, which is from 1994: https://www.youtube.com/watch?v=VRWa6v-QEcY
I provided the disc images of my copy of the original (not the sequel) to the people at this forum [1] — I think the "provisional images" must be mine — but they haven't released the game to the public.
[1] https://forums.jaspp.org.uk/forum/viewtopic.php?f=7&t=204
Have you seen OpenSCAD? It's open source constructive solid geometry modeling software, and I use it in my math classes. If your son enjoys Python, he might really enjoy this approach to modeling.
Students do a separate, adaptive curriculum for their skills work. For modeling, they do a series of mini-projects that build their skills. They sketch a number of things like snowmen, robots, apartment buildings, skateboards, ice cream cones - things that can be represented through cubes, spheres, cylinders, and cones. Then they model these shapes and learn to translate and rotate elements. They also learn to use the difference function. The last required skill is using parameters, which is the core of the reason we can offer this for math credit. It makes them think about the relationships between all the parts in their model.
They also have the option of using modules, hull, minkowski, rotate_extrude, or any other feature that helps in their projects. They propose a final project of appropriate complexity. The range of projects is just fascinating.
We may open a Shapeways shop for our school in the next few months. I have two students designing a chess set as an extension project, and one student who came up with a functional design that hasn't really been done well on Shapeways.
I'm happy to share more if you're curious, my email is in my profile. Also, I wrote a student-friendly intro to openscad a few years back:
https://peak5390.wordpress.com/2013/01/08/an-introduction-to...
The limiting thing about OpenSCAD is its inability to import svg; for complex shapes it makes much more sense to draw them in a graphic software, import into modeling software and extrude them there. How do you go around this problem?
I also love that I can use it as a core part of my math curriculum, which I couldn't do as easily with most other modeling software. Students are learning measurement, geometry, and most importantly relationships through the use of parameters. It also makes for a great segue into more general programming languages.
I recently started working with polyhedrons for a terrain modeling project. Could you convert svg data to a polygon and then extrude it?
I switch accounts regularly here to not tie to my real life identity but hopefully I hve enough unrelated crap not to seem like a shill.
Just silly stuff like I want to print part of it but laser cut other parts of it and I can't find an easy way to get the surfaces out, or I have to go back and remodel because my fabrication approach has changed.
It's probably me. But I feel like I'm fighting the software sometimes more than using it. Also no fit or mechanical (altho those features are often crap).
I want to like openscad but I can't efficiently use it to make stuff.
Yes, there are some tools that do this, but they're a hassle to use and usually only support a subset of svg, so many times I use FreeCAD instead...
The one bit of very nice software that does get used some places is interactive constructive geometry tools like Geometer’s Sketchpad (1995), Geogebra, or the like, though we didn’t know about those in my high school geometry class (also circa 2000).
This tool taught me a lot about optics: https://chrome.google.com/webstore/detail/ray-optics-simulat...
Just look at khan academy. It is literally a guy drawing in something similar to MS paint and explaining stuff.
There is a lot of technology that exists which could be applied to education but it isn't.
Edit: I'll address the comments/thoughts here in a follow up blog post. Glad to see so many people thinking about this space. It's a great thing for the world to have entrepreneurs building here.
Given those 5 years have now passed, what's your assessement of the situation now?
One thing not mentioned which I think is important is the ease of paying (whether the edtech is app or web-based). It’s all very well to target those in poorer countries, but many of them (consumers) still don’t have online banking, and payment gateways are often local and clunky (I did notice in my last trip to Asia that this is improving.) I’d love to hear your thoughts on this :)
The way I remember it, middle class folks in Kansas City suburbs really, really care about the quality of their schools and levy property and sales taxes to fund it. Rural Ohio townships where dual earners are making 17k less than the national median for such households may not have a lot of middle class to speak of.
Cost and outcomes have always been a tradeoff. 1:1 tutoring is the gold standard for educational outcomes. We know that from there, outcomes fall as teacher:student ratios rise. This fundamentally puts a cap on returns to improved quality -- if your improved school outcome is comparable to adding one additional teacher per grade level, well, you better be cheaper than those teacher's salaries. And good luck finding that data in the first place. It's not easy to convince folks to run the experimental treatment on their children to get that data.
Is it? I'd be interested to see some data to support or refute this. I'd always assumed that students benefit somewhat from other students being around (due to observing how others learn, being pushed to be better etc.), so perhaps the ideal ratio would be in the 1:2 to 1:4 range.
If people have really figured out that 1:1 is better than 1:2, 1:3 and 1:4, I'd love to see that data.
I haven't read the paper myself and don't vouch for it, but I've heard two-sigma mentioned a lot.
It's not relevant to my specific question, as the three groups in the test had student:teacher ratios of (1) 30:1, (2) 30:1, (3) between 1:1 and 3:1
I'm interested in variation between the students within group 3, not the difference between the three groups.
Education is not what the teacher does: it's what the student does. Most people think that the job of a teacher is to organise information and explain it to students to make it "easy to understand". This is just plain wrong. In fact, by doing so you cheat your students because you steal their opportunity to organise information and discover how to understand it.
1:1, 1:2, 1:20, or 1:40 doesn't actually matter from a perspective of teaching material. If we make the teacher the font of all information, then we also make the teacher the bottleneck of learning. I was teaching language and if you want native level fluency and proficiency, then you might need to spend up to native level time acquiring the language. In other words, to get to a 10 year old's ability, it might take you up to 10 years of nearly full time practice. Sorry, but I'm not going to teach you 58,400 hours (10 years, 16 hours a day) unless you have Bill Gates level money to spend.
So, you're going to spend something like 3 hours a week in my class, and only something like 30 of those weeks are going to be productive in a year (if I'm lucky). So that's a pidly 90 hours of instruction per year. I'm going to teach you practically nothing in that time.
This is what is wrong with education.
But getting back to 1:1 vs 1:40. In a 1 hour class, if I have 40 students I can spend just over a minute paying attention to each student. In other words, I'm going to present my pidly amount of information and hope like hell that you're going to do something useful with it -- because I do not have enough time to interact with even one person in the class. If you assume that 1 in 40 students have a behaviour problem (just 2.5%!) you will see that I will expect to be interrupted in every single one of my classes. So probably I won't even be able to present my pidly amount of information. Probably I will just babysit and the students who know how to study will do well, while the others do poorly.
As we reduce the ratio, I have more time to interact with the students. But it's really important to realise that if I waste that time trying to teach the material, I will do no good. So in fact, what I want to do is let the students study and then watch them so that I can can correct their behaviour. By doing that, they can become good students and succeed in studying.
How many students I can handle depends a lot on my class management skills and also on the students. Some students require no intervention. They just need a friendly, welcoming, and engaging environment in which to study. Some students need considerable attention. 1:1 might be amazing for some students, but even 1:10 or 1:20 could be amazing for others (depending on the teacher). Some students like to study alone, some like to study in pairs, some like to study in groups. You have to be flexible.
We need teachers who will enable their students to learn for themselves. Only then can the teacher no longer be the bottleneck.
Well, there’s always not enough time to teach something properly :) I don’t think that teachers need to be so idealistic, just as long as they tick-box the curriculum (sadly exams do matter a lot) and plant plenty of seeds which, we hope, would set root soon if not later in their lives.
By that I mean spending significant classroom time on interaction and the “fun stuff”. For example if it’s ESL and the students are into pop music, okay let’s decipher some Beyoncé or let’s try and mimic some funky singer’s accent. Do some immersive stuff, make it real. I’m a big fan of flipped classrooms, where students study most of the formal stuff for homework - this will also help them construct the knowledge in their own way - and then present it in class and have more meaningful discussions.
But I would say that if the students are beginners, then yes you do need to organise information to make it as easy as possible. The initial learning curve is always the hardest! When they become more comfortable and confident, they can always go back and revise their mental models.
Edit: Sorry, I didn’t take into account that you might be a private tutor. If you were, then I can absolutely relate to the frustration! Parents’ (totally unrealistic) expectations also do not help at all
- For language learning, total study time should be a multiple of classroom time. (From personal experience, I've found 10:3 to work pretty well.)
- Teaching can be very effective in groups of 1:10 or 1:20
- Classroom time should not be spent on introducing new material
However:
- For language learning in particular, I don't believe that 'let the students study and then watch them...' approach is best. If you're doing 1:1 or 1:2, you can actively manage the learning process by asking the student questions in the target language, and (for example) making them reply using newly-acquired grammatical structures.
- You've made the point that teacher:student ratio isn't the only important factor (agreed!) but haven't made a strong case that it's insignificant, or addressed my question about whether 1:1 is better (assuming money no object) than all other ratios.
https://www.amazon.com/Equality-Achievement-Introduction-Soc...
This was well before dyslexia was a mainstream issue in schools - needles to say this did not go down well with the headmaster
If we are interested in how many students are required to support peer to peer learning, you might enjoy the following articles (popular and academic) about how Harvard physics professor, Dr. Eric Mazur has embraced and studied peer-to-peer instruction. This should provide a partial answer to your questions.
1) Eric's Harvard Profile https://www.physics.harvard.edu/people/facpages/mazur) 2) A good story/video about the development of his peer-to-peer, active learning methods (https://www.youtube.com/watch?v=WwslBPj8GgI). 3) One of Harvard's articles about his educational work. (https://www.harvardmagazine.com/2012/03/twilight-of-the-lect... 4) A list of Dr. Mazur's publications (https://academictree.org/physics/publications.php?pid=66072) 5) Eric's 2015 papers on peer learning a) 2015 - Dowd JE, Araujo I, Mazur E. Making sense of confusion: Relating performance, confidence, and self-efficacy to expressions of confusion in an introductory physics class. Physical Review Special Topics - Physics Education Research. 11. DOI: 10.1103/PhysRevSTPER.11.010107
b) 2015 - Miller K, Schell J, Ho A, Lukoff B, Mazur E. Response switching and self-efficacy in Peer Instruction classrooms Physical Review Special Topics - Physics Education Research. 11. DOI: 10.1103/PhysRevSTPER.11.010104
Googling for "ideal group size for peer to peer learning physics" brings up articles from Vanderbilt University and those mention groups of 2-4 or 5-7.
As there are many different models of peer learning, this article does a nice job of discussing practice and the theory behind peer to peer instruction (https://www.ncbi.nlm.nih.gov/pubmed/29128431)
My personal beliefs and experience is that people tend to like groups of 6 or fewer and the groups which speak the most among themselves range from 2-5 people. The limitations are that at least 1 person in the group must be advanced enough to help the others learn.
Last, if we look at apprenticeship models of learning, we are back to the 1:1 ratio (expert : learner). However, any single student can also have multiple 1:1 "teachers" and apprenticeships/internships usually include a peer group which engages after the 1:1 instruction. We see this in music and sports with individual sessions with a variety of experts followed by group work or performances with varied sized groups.
I hope you enjoy!
In a 1:1 interaction, it is possible to efficiently figure out where the two interlocutors’ common premises are, and then explain concepts in whatever amount of detail is necessary for understanding, skipping over any parts that are known, with frequent checks that the other party is following along (not just pretending, and not suffering some misconception).
In a 5-way interaction, it is typical for each participant to be coming from slightly different background / preparation, and have difficulty with different parts of the new material, which makes it necessary to take time for (sometimes extended) digressions to fill in whatever details are missing for anyone. It’s easy to accidentally not notice that 1 or more participants is missing crucial background, because it is time consuming to verify that each separate participant has a solid understanding of every idea. A problem that is too hard for one student will be too easy for another.
Things might be better if you had the same 4 students to 1 tutor for, say, 3–4 years at a stretch. Then you could assume a lot of shared history, and amortize digressions to fill knowledge gaps for particular students. Better hope the students learn at a similar pace.
When actively trying to solve a set of problems, with the students working independently with minimal guidance/help from an instructor, and if the only goal is problem-solving efficiency, then having more people is helpful, because different students will have trouble with one problem or another, and they can then confer and fill in the other answers at the end. Often the weaker students don’t fully understand everything to the extent they could solve a similar problem independently, but that’s often an acceptable standard. [Same story if you are trying to build some work project with many parts, etc.]
But if the goal is to get everyone to understand every answer deeply and accurately, then in a 1:1 tutorial setting the expert tutor can ask leading questions, give appropriate hints, or just outright explain the answer if the problem gets to be too big a time suck. And the student can get more direct personal experience with trying to solve all of the problems with a fast feedback loop and without getting bogged down, and hopefully feel like they personally figured out most of the main breakthroughs.
One issue is that often tutors (probably more than other types of teachers) are amateurs rather than experts in the subject or in teaching. For instance the tutor might be a slightly older student, a family member, a retired person volunteering, or the like. In that case 1:1 tutorial is still helpful vs. the student working alone, but obviously pales in comparison to highly skilled tutoring.
One problem is that this method of "teaching" is not appealing to most of us, so the ratio of teachers to students does not increase.
Another problem is that "traditional" lectures, homework, and testing are not the most effective methods of learning, they are simply more well-known, and known to be scalable.
It doesn't seem scalable for each student to follow her/his own curriculum, since that would create more contexts for teachers to respond to, thereby creating a need for more teachers. What is ignored here, is the redefinition of "teacher". Such a teacher is much more of a "mentor", and that is certainly a more appealing job to me.
There are new challenges there but it's also very high-reward. You can sell 1000 licenses from a single deal with a school. If you get a government public school contract you could be looking at 100k licenses from a single deal.
I think that companies which do big b2b deals have become greatly undervalued. VCs are so obsessed chasing exponential growth curves that they seem to have forgotten that the earth only has limited population and that buying power has become much more centralized and is becoming increasingly centralized.
I think most busy parents these days tend to see the school system as an end-to-end automatic education machine; you insert a young uneducated child, put some money in it and several years later an educated adult comes out.
With that mindset it's easy to see why it's better to do deals directly with schools than trying to get the attention of the parents.
1. Cost. When you deal with school districts, you usually sell for X number of devices. So for example, all devices for 5th-7th grade, something like that. When you sell alot, schools will usually ask for volume licensing. Schools don't want to see prices above $5. This goes for devices for the quality of the devices they purchase. Often, the student device is cheap, low-powered, and in the case of Windows devices, packed to the brim with all sorts of software running in the background: web filters, virus-scanners, and some educational apps. Top that off with apps that students install like Spotify and games, it slows the machine down to a crawl or causes problems that might interfere with your software solution.
2. Different school districts have different needs. Although our software solution was initially meant to be for general use by all schools, we kept running into school district that needed very specific needs to that school or district. That lead to developing new features that only a small handful of school district would use, sometimes confusing users who did not need such features, or didn't even know we had the feature.
3. If your software solution works with student information systems data, it can often be a nightmare getting school districts to create that data.
4. School IT teams are...pretty terrible. Or rather, they're swamped with so many issues from so many products and devices, that they care very little as to learn how to deploy your product properly. Have documentation? Trust me, they probably won't read it. You will most likely get an e-mail asking how to deploy.
5. Trials, trials, trials. Schools aren't going to pay up just because they see a video demo of your product. They're going to want a trial at minimum for half the school year, maybe more. Then...
6. ...Politics. Decision to purchase needs to be approved by some board.
1. Cost: Windows licenses for student PCs are just under $20/PC/year (for us). Is your $5 software 25% as useful to students as their operating system is? Even worse, the reality is if I put your software on my 1000 computers I'll be lucky if 300 of them use it. That's something to keep in mind when working with per-device licenses. I agree installed system software bloat is a problem. That results from every department having different needs. We might be a little different in that we don't allow students to install/run foreign software (using AppLocker).
2. I agree different schools are amazingly different. I collaborate with a few other schools around me and whenever we meet I'm pretty floored by how differently each one approaches basic problems. These significant differences really add up and I can understand how supporting schools might be a pain.
3. It seems like every vendor wants access to my SIS. That's a privilege not granted lightly. I'm generally okay with automated CSV dumps of bare minimum data uploaded over a secure channel, but some vendors have the nerve to request direct LDAP or Oracle database access over the public Internet. No way!
4. I'll agree school IT teams are generally terrible. Among my peers and schools I collaborate with, I'm the only one with a degree in IT. Many are older and migrated to the role from something else (former math teachers surprisingly often). Some find themselves shoehorned into the role because they happened to be the most technically-inclined among the staff. These guys are uncomfortable in powershell/cmd and typically use contractors or local PFYs to handle technical details. Schools do not pay enough to attract decent talent. I could work for a local company and make 1.5-2x more doing less than I do here, but for now I like this environment.
5. While video demos can be helpful product overviews. I distrust video demos just as I distrust other marketing wank. I want to test drive the car before I decide to purchase it. That said, I wouldn't expect anyone to let me demo something longer than 60 days. The long time in my case is that exploring new software is usually a backburner project, and I am likely to have other things come up that take priority. I never put production data into trial software however.
6. Yes, purchasing processes are a pain. I 100% agree. I currently have a request for $60 of printer toner that's been sitting in the requisition process for over a week. Trust me, it's just as bad for us on the inside.
1. I totally understand this. I've actually often wondered what a waste of funds it is when I see school districts purchase several hundred licenses of our product, but only a handful of teachers and students end up using it. I understand though that it's hard to just purchase the right amount of licenses, since you never know if another teacher/student down the line might want to start using it.
4. Yes, I've been to a few local schools for support and there's quite a few teachers who are handling the simpler IT side of things. From the IT staff that I've met in schools, I think I would like the environment too.
Disclosure: I have two kids in high school, both of whom need graphing calculators for their schoolwork. Both would prefer to use Python/Jupyter.
Students and parents are often fed a misleading impression of what tools they will actually need to work through a course, and pressured to buy calculators that are not necessary, preying on their desire to succeed at school. I think I was the only student in the class with enough moral outrage about TI’s calculator scam to try such a thing, but my impression was that it really had no practical consequences, except for me spending a lot less time than the other students playing low-quality snake, brickout, and tetris clones.
It could even be relevant: I used the projectile equations from Physics class for my Artillery clone.
We all complained about the TI-83 being a prime example of school boards being clueless and putting technology in the classroom for technology's sake.
Sorry to hear about your tough financial situation in the past. It’s a tragedy that kids go hungry in such a rich country. If an expensive calculator is going to be mandated in school, there should be financial aid available for it.
I'm a math teacher and a programmer, and I can't wait to see the end of the TI dominance in education. I would happily let students in my classes use Python/ Jupyter for most of their work, and use calculators as needed to get ready for any tests that require their use.
I think there are a couple of broader issues here. First of all, I'd like to see more of what I call computation for lack of a better term, in the K-12 math curriculum. This isn't for turning people into professional programmers, but because computation has had a huge influence on how people do both practical and theoretical math. Thus, a computation-free K-12 curriculum misrepresents what math is.
Of course the other issue is teacher training, and standardized testing. Changing the curriculum requires changes to both teacher training and the tests. If it's not on the test, then it can't be taught. Remember proofs in geometry? Gone.
There are a ton of popular websites that teach people coding/software development/design skills online now that have grown incredibly over the last few years, but still not really getting to mainstream status. Treehouse, Code School, DataCamp, Udemy, Frontend Masters, Udacity, Thinkful, Codecademy are just a few popular websites where people can learn to code today and all of them have a subscription plan. I don't think any of these have reached "mainstream" status, but there are definitely a lot more websites and applications today teaching people to code via subscription plans than there were 6 years ago.
Whatever your thoughts are on the whole "learn to code" movement, it's clear that the landscape for edtech companies teaching programming/software development has changed since 2011. Will any of these companies teaching people to code ever reach "mainstream" status, I don't know.
I still believe we could have, but we couldn’t convince others. Mostly because a real MVP would have cost millions to create.
But we could have saved states billions by replacing textbooks with online, interactive, cross-discipline content.
The market is closed to startups. The industry is fighting disruption in every way they can.
However, I am an optimist and believe that even with well entrenched interest groups (big textbook publishers, school districts, unions etc.) there is still paths to disruption. At a fundamental level, education is simply teaching someone something new. That's it! If you are providing something quality then people will naturally flock to it or share it.
I think the future education startup simply needs to focus on building high quality learning materials. Then either give those products away for free or innovative monetization models (ie. watch 5 video ads to unlock the content etc.). Don't focus on trying to sell to school districts. Focus on going straight to the consumer: students or parents. Every student I know is already going online to get extra homework help, resources, youtube videos etc.
This is where Pearson, Britannica, et al fail. They have a library of content created over decades with a guaranteed state-driven pipeline into classrooms. This content is specifically created by old school experts specifically for printed chapters in a bound book.
Creating real interactive content would require content experts work within an interactive digital agency framework, which would require a sizable R&D budget with a lot of trial and error. We're talking millions of dollars spent before a completed product was available.
But here's the upside. Textbooks cost each state roughly $100 per student per year. An online system might charge a fraction of that (Textfyre targeted $1/student/year).
Imagine the state of Illinois saving $99 per student per year over several years.
The investment is risky to a degree, but I view it no differently than Kennedy choosing to go to the moon. You decide to do it. You find the best and the brightest to enable the vision. You enact the vision. There is an upfront cost, but the vision and results would alter the course of history in the world.
Interactive, adaptable education systems are the future. They would be able to reach everyone on the planet regardless of age, gender, politics, class, or any other identity labels.
I have a second start-up (Wizely). I'll probably keep swinging if it doesn't hit a nerve. If I ever have one succeed and make (a lot of) money, I will re-invest my profits into education and what I call positive social engagement.
My advice to ed-tech startups. Find a way to make money, but immediately redirect all profits to attacking the classroom. That's where we need the R&D to go. As the OP said, consumer-based ed-tech has its limits.
The classroom is where we need to drive innovation.
- It's very hard to measure to value add of education; people have spent lives working on this
- Only half the value of educational institutions comes from educating people; the other half comes from the credentialing where they pick out the 'brighter' individuals, but measuring how much a credential is worth is almost impossible for most individuals
And in the USA in particular what point are BAR exams when you get people with ZERO experience of law appointed to senior legal positions - No offence to one of Obamas Supreme nominations but they had no real world judging experience - shades of E l Wisty
- As it says in the post, there is a lot of generalization. And usually generalized advice for something as specific as building a startup is not good
- This is 6 years old - not sure if it still applies. A lot has changed in the tech possibilities since then
One could say that his degree was entirely a waste of time, yet in order to qualify for a job in investment banking you need to go to a prestigious school and excel.
I see many spaces where education is largely irrelevant and functions only as a filter to reduce the supply of people who are qualified for a job.
I think an interesting problem in edtech today is there are a lot of "startups" that fit in that dangerous area between a small business and a high-growth startup. You'll find dozens of companies that help improve mathematical skills, create STEM games, or build products for improving reading and writing comprehension. A lot of these companies might grow to be multi-million dollar companies over several years, but they won't fit that "classic" startup growth curve. Since venture money is flowing relatively well (and a few VCs focus specifically on edtech), a lot of these companies get funded. But there's a dilemma for these entrepreneurs: what do you do when you realize that your company is growing 2x each year (which is great!), but not at 10x where your investors want you.
Unfortunately, MECC got spun-off from the actual Minnesota education system, then purchased by a giant ed-tech company and left to wither. The endowment that MECC (in its purist form) left behind faded in time.
There is still a lot of great educational software out there. It's just all locked away on a MacOS 9 box, tied to libraries and runtimes that never made the jump to Mac OS X and then iOS.
From the business side of it, we'd honestly need something of a PBS-like software company to reconstruct what we lost with MECC. I think that's the only way it'd work. The software would have to be made, not because it was profitable, but because it needed to be made and was a good thing for our society to have.
One would think there would be a market for Duolingo style apps for other subjects, such as mathematics, where the user is simply charged on a subscription bases. Could such a business not survive?
> While it is “teaching,” it’s also squarely in its own domain
Could this idea simply be because, for the most part, language learning applications have been the most successful? Other than Khan Academy, which is mostly videos, i've never heard of another company building similar products for other subjects.
I'm not sure whether it's because the market doesn't exist or that it hasn't been tried.
As humanity requires language, learning a new one can happen at any stage of life. Rarely do people spend casual time learning what is typically taught in youth. Yes there are always people interested in TED talks but that’s effectively TV. Yes people go back to school but that’s focused structured time. Language is better if learned interactively, in both casual and formal settings. I can read a history book and feel accomplished. Reading a French textbook without application won’t work as well. Language requires active learning, where as many other subjects do not. Language apps such as duolingo are well suited accordingly (even better would be a really great conversational bot that could help you better pronounce your words).
I’m not saying there aren’t other markets, but I am positing it’s def not evenly distributed across subject matter.
Except we have no idea whether these companies are actual businesses or not (i.e. have profits). Anectodal - Clever has just as much funding and number of employees as another EdTech business I know about that also interacts with "millions of students" but is barely pushing beyond a few million dollars in revenue. This isn't enough to sustain a business with 100+ employees in the bay area.
Source: I work at Clever
This is because "the tech" is just the tip of the iceberg and there's a disconnect between the districts that purchase software and the teachers who are left to try and figure it out on their own.
I wrote in more detail about this issue here:
https://www.forbes.com/sites/forbestechcouncil/2017/11/13/wa...
Coding is one of the things that parents are increasingly willing to pay for, but it is still in the process of becoming a middle-class family demand.
I was looking at the apple app store today as we were waiting for an appointment. It is quite flooded with all sorts of educational apps, but most that I tried were of low quality. Performing a search for the best reading or math app did not improve much.
I think there may be an opportunity here, but it would take a team with diverse talents and not just a team of developers to build something good.
Reminds me of the AirBnB guys pitching their original idea of people renting out their couch for strangers to sleep on and getting thrown out by virtually all of them. “Who would EVER want that??”
I can think of at least a dozen educational services and products and it's possible/likely the optimal growth strategy would be different for each of these. For example, tutoring service (in person or remote), learning web application, learning mobile app, classic content products (print textbook, digital textbook, recorded video lessons), new-media content products (e.g. new apps or learning modality), etc. Then there is the whole "credentialing" aspect, which we all hope will go away but currently it's still a very important: people are easily impressed by big-name universities, and three-letter degree acronyms.
The "edtech space" is very interesting. The OP is totally right that it's not obvious how to make money given that neither schools nor students have money... (Sure there are rich kids and motivated parents who can pay and they will enjoy the value, but it's a very ugly thing to deny access to knowledge to someone behind a paywall/subscription just because they can't pull out a credit card and follow the funnel.)
In addition to the traditional for-profit startup model, there are big developments in the ed-tech sector by non-profits like Khan Academy. Some of the best learning UX/UI and content to come out in last decade is from KA. More generally, the open educational resources (OERs) are starting to be a thing now... there lots of good open content and there is potential for companies to be "service providers" on top of the vast sea of "commons." Open source gave us an interesting last two decades. Now imagine what will happen when open content becomes mainstream too?
On the other hand being too open with the content and starting with a "yes take it all for free" strategy is a sure way to end up out of runway. How many non-profit projects shut down because of lack of funding. Aren't there always strings attached when receiving funding from governments, NGOs, and philanthropists. What if they change their mind next year? Or two years form now? There is something beautiful and robust about a self-sustaining project like a for-profit startup. I know teacherspayteachers and other for-profit projects that are doing very well. Definitely something to learn there.
It all depends on what you want to optimize for, but the fundamental algorithm in the edtech space remains the same as in all other spaces: deliver more value than you capture and be in the right place to ride the wave up.
The future of education is something along these lines I think. One could really make a lot of money with this because the costs will be fractional compared with real high school. Also no "athletic departments" fuck that shit.
At our school, everyone's in classes like any other school. But towards the end of a session, after most of the teaching has been done and everyone's working on final projects, students are free to go at their own pace. If you finish before the end of the session, you can do an extension and earn more credit, not just a higher grade. If you don't finish your work by the end of the session you keep working on that class until you're finished, and you don't start a new class until you've finished your existing work.
This is working really well. The culture of our school has shifted from students feeling like teachers are giving them work, to students looking at us as the educational resources we're supposed to be. Students are in charge of their own pace to a large degree; there's no concrete expectation that everyone will graduate in exactly four years. The range is closer to 3-5 years.
Coworking spaces sounds great for a high school. Access to remote experts sounds great. But "all teachers are remote" sounds really stifling.
In my school, 5% of the class (probably less) were actually there to learn, were on a college track of some sort, and would take advantage of educational resources. Their parents took an active role in their education and pushed them. A good 70% or so were pretty much furniture. They were there because they had to be, and their parents saw school as free babysitting service. They would end up in a trade school afterward if they were lucky, or a more likely future drifting between unemployment and Walmart/McDonalds employment. Then there were the 25% who were actively opposed to learning. They existed to disrupt everything, start fights, and waste all of the teachers' and administrators' time. Most of them were into gangs, selling drugs, etc. and their presence detracted from everyone else's experience. This group is likely dead or in prison today.
Aside from the political reasons already covered by other posters (school boards playing politics and being more interested in mandating and banning things than furthering education), there is this reality that education startups ignore: The vast majority of their end users will not want to be end users.
There is so much here that can be looked at critically if we want to change how education is done. Looking at people as furniture is really telling about what you think of their potential. If you're people like this from a student's perspective, I can understand this. But anyone carrying this mindset as an educator should leave the profession.
I have spent my life teaching the kinds of students you describe. They're the people whose father died when they were 8 or 10 years old, and whose mom had to try and hold their family together. They're the kids who have never known sober parents. They're the kids of parents who are angry at the world. They're the kids of parents who have been arguing and hitting each other their entire lives. They're the kids whose parents never opened a book with them by the time they got to first grade.
Yes, there are some kids who grew up like this and inherently took school seriously and flourished. But they're the exception, and we can't expect everyone to respond like that. Schools really need to be restructured to truly meet kids where they are at, and work from there. It can be done, and it's not overly difficult. We tend to meet kids where they're at for a little bit, and then expect them to catch up to everyone else and stay there.
There is a small portion of the population that is really difficult to reach. But schools can do a lot more to bend to students' needs than they're currently doing, which is asking everyone to change to meet the schools' needs.
what's inhuman about this btw?
If online education got people cushy jobs the way university degrees get people cushy jobs - online education would be massively profitable.
That's not the case. There's nothing else to it.
Regarding online education in general:
Universities/colleges should have to post their full curriculum, lecture notes and all textbooks and chapters used for each class - that's enough.
There's no need for businesses here - just tell us what you're teaching your class, and anyone interested can go read the textbook. Make video lectures available as a bonus for a little bit of money, that's it! A lot of students already don't attend classes and just read the lecture notes and chapters needed.
These online education start-ups remind me of companies that take Craigslist and start adding unnecessary bells and whistles to justify their existence. We already have craigslist and it works folks. Stop trying to make a buck by being so incredibly boring.
So the last mile left to accomplish in 'online education' is to standardize the presentation of the syllabus/outline and inform people that that is a viable option.
Basically consumer facing spending is about delivering cheap services, whether that be babysitting style tutoring, textbooks or slinging a minimally viable educational experience that aligns with government programs.
Targeting schools is an enterprise play. You either have the ability to survive the public school sales cycle, or the connections to get the money people behind charter schools to buy your crap.
Personally, I think this is a shitty startup market. It’s an insular industry, most non-content plays have entrenched competition or a bunch of players (Microsoft, google, apple) doing a half ass job for free/cheap. It’s easier to do enterpsie style business with normal government entities.
Targeting poor people is dumb because poor folk give zero shits about education. If you have the political juice, targeting a government agency with money to burn on helping poor people makes more sense.
Don’t even know where to start on this...
You may find that harsh, but it’s reality. Doing what the author suggested (producing low quality cheap crap) may sound more sensitive, but in my mind is unethical.
Things like Khan academy have traction because third parties pay the bill. Online degree mills have traction because an uncritical third party (ie the government) loans the cash to students.