EdTech by Misadventure, Part 4

Nathan Allen

End of the line.

So how does a well-funded ed-tech start-up with seemingly experienced tech talent fail to the point of producing no value — as in, it can’t even get acquired? Its tech can’t get acquired? Its… nothing, zero nada no value at the end of the day?

That’s the question for IBM’s Watson Education, which came out of research after about three years (and tens-of-millions spent) and, after consuming tens-of-millions more over two more years, was put up for sale with no takers.

It’s been oft said that IBM can’t dev software — I heard that many times even within IBM. But what about all their software? Well, first, much (maybe most) of their software was acquired. What was the last notable and good internally devved IBM software? OS/2 is a common response. But I’d point out that (1) we’re talking about the 1980s and (2) that was co-devved with Microsoft, and Microsoft threw a fit over IBM’s dev skills and processes (or lack thereof). IBM was supposed to dev Apple’s next-gen OS in the 1990s, but that crashed and burned. Then there’s Lotus. (j/k I said “notable and good.”) Just the other day John Kelly, IBM EVP, was talking to a few IBM programmers from the 1960s that had worked on the Apollo software. So, maybe the answer is: the 1960s Apollo software.

Watson Education’s first problem is that they didn’t have any products — even after all that time and money. I said often at IBM that I didn’t think many there knew what the difference was between a product and a project. Watson Education’s products were really collections of technology with some services thrown on. Some people think a product “solves a problem.” Nope. A product is something people will pay for that has economies of scale. It’s just something you can make and put on a shelf, and a customer will show up and buy it off the shelf, and the more you make, the less it costs to make. Simple. And good tech has enormous economies of scale wherein adding an incremental customer costs almost nothing (apart from perhaps run-time costs and maybe customer service — but very low incremental margins.

Watson Education’s main “products” were a teacher-focused content and data aggregator and an intelligent tutor (for higher ed). The former required significant install cost+time and the later wasn’t scalable — so, really, neither were scalable in the usual software definition of the term. You couldn’t release a v1.0 on day 1 and a million people start using it on day 2. Given all the work it would take to install a single new customer, both of these were projects, not products. It’s not surprising that when these failed within IBM that no one else was interested — they just didn’t want to acquire a project.

In a recent post, I made a point about $9/hour devs killing people, literally. This happens when software is a cost-center and not a value-creator, and that happens when accountants run your business. And accountants ran Watson Education. One of the problems is that Watson Education had maybe ten good devs and 40 terrible ones. At a certain point your team reaches a critical mass of crap devs from which the project cannot recover. I once said to a Watson Education dev leader that he should fire his whole team and outsource development to some local shop (local=California). I was braced for him to take this as an insult. He just looked at me and said, “I know.” It was that bad.

Perhaps most amazing, I began to realize that each dev team (and teams within teams) all had their own rules and tools. There was (is) no official IBM software Rules n’ Tools — no standard repo, no company-wide project management software or rules for using such software, no requirements for instrumenting v1.0 for usage (or anything). Nada el zippo.

Obviously I didn’t believe it, so I called around to 6 or 7 different PMs in entirely different groups and asked which tools, PM platform, etc., they used. I got a different answer from each of them.

Sure, IBM has a PM training course. One PM in Watson Education begged to be sent. Never happened (manager didn’t want to pay for it — IBM requires internal budget transfers). So there were no Rules n’ Tools. And none of the managers — none — had ever developed a software product before. And that’s how you get projects mistaken for products.

There are some people in IBM who know education; they were ignored and otherwise excluded from Watson Education. So, a bunch of people who have never released a software product wander into an industry in which they have little experience… hilarity did not ensue.

It’s no wonder they didn’t build something people would buy — how would they know?

Brings us to a key problem: how do a bunch of executives become convinced to spend millions on a business unit working in an industry about which they know nearly nothing?

That’s right: consultants! In IBM’s case, there was a McKinsey study of the education market that IBM commissioned that essentially said whatever you wanted it to say.

Let’s say you took a mentally challenged baby hippopotamus, loaded it up on meth, then unleashed it on a giant canvas with paint on its feet. The result is approximately that McKinsey study. You’ll see whatever you want to see, and the IBM executives who commissioned the meth-hippo saw a vision of transforming education, personalized education, and maybe even a promotion to General Manager. As Nassim Taleb (Black Swan) pointed out, these are rent-seekers and actors in training who saw their next rent or starring role: education innovator.

One of the issues in education is the structural fragmentation — funding structure, data structure, really everywhere. If you want to scale in any meaningful way, you either must address or avoid these issues. Google, for example, began in education by simply avoiding them; it was free (funding structure) and required no school data (data structure). Bam. Scalable. IBM launched Watson Education in ignorance that these issues existed; really there’s no other way to explain how a “product” would require massive quantities of SIS data just to be useful while claiming it’s scalable. You simply must not see the mentally challenged baby hippopotamus in front of you. Of course, this explains why (1) Watson Education never developed a customer base and (2) it failed to find an acquirer.

I mentioned to someone who worked in Watson Education that it was a shame that the net-result was no value. He said, “well, maybe it’s just a corporate proof-of-concept … tens of millions and you’d shown what you can do.” Another IBMer commented “we’ve wasted money on worse.” But here’s the problem: usually with a PoC, even if the concept doesn’t become a product, you develop internal knowledge and skills (relevant to the tech and the industry). But with Watson Education, the entire unit was disbanded and most of the senior technical members left IBM. A PoC may not produce a product, but it should produce value. And it’s not clear that Watson Education produced any value, anywhere.

Within IBM, the Watson Education management mantra was “we’re a start-up.” I heard it many times from their senior managers. It was a profoundly meaningless phrase — an ed-tech start-up with neither adequate technical skills nor adequate industry knowledge, run by accountants and guided by a McKinsey study. Even the mentally challenged baby hippopotamus has value at the end of the day.

NB. We’ll end the Misadventure series here — I’ve got way too many other items piling up (from Coleman to many A.I. ones I’ve been neglecting). Perhaps later we’ll look into some older edventures (none the failures of those in this past series) … like the time I was working on a national ed project in MX and trying to get HP to take the tablet contract (remember when HP made tablets? Well, they were assembled in Mexico … so you’d think selling your MX tablet in MX would make sense). At one point I discovered the hardware logistics guy was trying to bring in bootleg iPads…

Him: we’re missing 10,000 tablets.

Me: ?

Him: They’re stuck in a warehouse in Alaska and WE’RE NOT GETTING THEM OUT.

(These tablets were on a plane from China to Mexico that made a stop in Alaska, at which point U.S. Customs got cranky. And no, the tablets were never seen again. I can’t say that the logistics manager learned a ‘don’t buy bootleg iPads’ lesson.)

You’d think that HP would gladly step into the void with their MX-made tablets. Of course, you’re also think HP wouldn’t have paid $12B for Autonomy.

Or maybe another adventure into Central Asian math textbooks (the things State keeps track of…). Plot twist — even after two decades of post-Soviet independence, their math textbooks were still all in Russian.


read original article at https://medium.com/@nathan.a.allen/edtech-by-misadventure-part-4-3f61f70b63d3?source=rss——artificial_intelligence-5

Share