My Mobile AR Start-Up Died So Yours Doesn’t Have To

David Hyman
The Startup
Published in
7 min readSep 18, 2017

-- acquired a little over one million downloads in the months it was live, but if you’re not between the ages of 10 and 13, I’m pretty sure you didn’t see it. And now you definitely won’t. Because my mobile AR startup died. Here’s why yours doesn’t have to.

First attempts was a pivot from our earlier app, Chosen, which attempted to gamify the performance competition space. (Think mobile American Idol platform where spectators compete as judges. See my Medium post from its launch for the deets.)

Through our efforts building Chosen and securing what seemed an awesome marketing opportunity with Ellen Degeneres (who took an equity stake and integrated our app into her show), we had a good deal of data on our users to better understand what parts of the app were working.

(See Chosen on Ellen…)

The clearest lesson early on was that those who used the app as performers greatly outpaced on retention metrics those who participated as spectators or judges. It’s the same force driving apps like SnapChat and our civilization’s seemingly endless desire to use mobile tools as a vehicle for self-expression and sharing.

Although I believed in the concept of trying to turn spectators into judges, trying to crack that code with the cash we were sitting on, was a riskier proposition than figuring out how to improve our self-expression tools and make it a better place for performers.

Early Takeaways from Chosen:

  • Make sure you’ve built an application that is charting with high retention metrics BEFORE you give away any equity or pay for what you hope to be a large scale funnel. Your top-line installs are not important whatsoever if you did not build a better mouse trap to retain users.
  • Put ALL your effort into quickly cracking the retention code. Buy as few users as possible to get the minimum necessary data to reach statistical significance to project future performance metrics.. All of your blinding focus goes here. If north of 40% of your users are coming back the day after they install (d1), and more than 20% of users who install your app on a given day are going back 30 days later (d30), you are on a really good path. If your numbers are south of this, you can’t take your eye off this ball.
  • Be wary of celebrity hook-ups or any deals with potentially large distributors who talk up their socials. Those deals are better for brand builders than direct marketers looking for app installs. The bigger the celeb or distribution partner, the worse their conversion rates per follower are going to be.

The Pivot

Flush with the knowledge of our desired path to improve our self-expression tools, we came up with a fresh take on the plethora of AR-style apps that create visual effects based on face detection and tracking (e.g., giving you a bunny rabbit face or allowing for projectile rainbow vomit). Whereas those applications overlay fantasy onto reality, we thought it would be novel and differentiated to overlay reality onto fantasy.

We aimed to superimpose the user in real-time onto the background of their choice with nothing more than their mobile phone. Needless to say, this is really hard to pull off. Whereas those facial morphing tools simply need to approximate your eyes, nose, or mouth, we need to know EXACTLY the edges of your body versus your background, without a green screen! Our pivot came with additional capital from our existing set of investors who continued to believe in the team. (Thank you DCM, Rhodium, Cumberland!)

We reached minimum baseline pretty quickly. Yet getting it to work well was incredibly hard. Weeks turned to months. I kept putting prototypes into my daughters’ hands, and getting consistent feedback that it wasn’t good enough. We needed an interim step where a user stood in front of a solid colored wall, in order to distinguish them from their background. That’s the only way they’d get good results from their video integrations.

Asking kids to do an additional step like this is really hard. I was concerned it wouldn’t fly. But with only four months of cash left, we had no choice. I had to get to market, gain traction, and find VCs. Stressful? You betcha!

Takeaway from pre-launch

  • Everything takes at least twice as long as you project. If you don’t have the run rate for double your engineers’ forecasts, plus the time needed to go to market and iterate, don’t move forward.

Just before launching, our engineers had a breakthrough. They’d been on a crash course in machine learning, training an AI engine using Google’s Tensorflow. We were bringing in kids and putting them in front of green screens to educate our algorithm, and were close to having a lightweight inference engine that would work with the graphics processor on a mobile phone. We were a few more weeks, maybe months, to optimization. But we ran out of time. We had to go live with the awkward interim step.

We desperately needed great retention metrics and some early growth in our daily active uniques –at next to no cost — showing the product’s viral, sticky nature. So we partnered with influencers who agreed to create “blin.gies” and upload them into to share with their millions of followers.

It worked! We got over a million installs at less than $0.10 per install. The videos were a perfect target to a perfect audience. The influencers on were still greatly underpriced relative to Instagram and YouTubers. And considering that does not allow embedding of external links, such as app store links, the conversion was even more amazing. All users saw was our watermark on the videos, and made the leap themselves to install the app.

(See in action!…)

So we started getting the users, but here’s where the rubber quickly met the road. Without the undesirable interim background detection step, many users saw less than stellar quality videos. This of course had a big impact on our retention metrics. We needed 40% day-one returns and were closer to 25%.

The clock kept ticking.

The maddening race

With just 60 days of cash left in the bank, I needed to get one VC to give us a term sheet, which would give us a bridge from existing investors to close. We had no choice but to go to the venture community with our sub-optimal retention metrics and unfinished product. We had to pitch our plans to improve the metrics with better tech –once our machine learning was perfected and the product worked like magic.

We set up 18 VC meetings and hit the road, hard.

The feedback was eye-opening and generally the same:

“Really great technology and vision. But how does this become a platform?’

Over and over, almost as if they were all had a collective hive mind. Our counter was that Instagram and started as tools and became platforms. Their rebuttal: “For every tool that becomes a platform, most don’t. How will yours?” We didn’t have a compelling answer.

Tick, tick, tick, tick.

Only one painful move left. Sell the technology and/or team. Do our best to recoup investment. Recast our sales mode for 18 VC meetings as sales mode for potential acquirers.

Many of the world’s most loved technology companies were at the table. One in particular had us building prototypes for them with our machine learning tech week after week, asking us to hit higher bars to demonstrate improvement.

We were told we were getting bought. My Tel Aviv engineers were flown out for a week of intensive grilling. We had 20+ meetings with our potential suitor and all signs and words said “Yes! Yes! Yes!” Love and admiration, redemption, and the whiff of impending victory. You try to keep calm and keep a lid on your expectations. But you also can’t help yourself. I’m warning my team in Tel Aviv NOT to start looking at places to live in California, and at the same time, I’m also starting to daydream.

A week later, a casual email is sent saying they’re passing. No reason given. Now, we’ve got two weeks left of cash in the bank. Enough to lay off staff and-pay severance. Game over.

Final takeaways:

  • Keep your expectations in check to preserve your psyche. It’s easier to set expectations low and exceed than to set them unrealistically high and get hurt. You know the expression “Until the money’s in the bank?” Well, I’ve had an investor pull their money out AFTER it was in our bank account. So believe me when I say it.
  • Have deep faith in what you’re doing so you can have the ammunition and energy to plow through the end. It takes a Herculean effort and that only comes when you’re fully committed to the vision. I’m still a frustrated believer in our vision.

It’s not easy to charge up a mountain and never quite reach the top. But you sure build up your stamina through all that hard work, and hopefully you gain new vision and clarity from having scaled those heights. It’s also good for the heart. Sometimes the pain of a defeat can inspire you to accept the next challenge.

This story is published in The Startup, Medium’s leading publication for entrepreneurs and startups.

You can subscribe to get the top stories here.