Our DNA is written in Swift

ProductLayer Post Mortem

In the beginning, there was an idea: To provide a web service for app developers that would let them get basic product information – like a title and category – for a product which they only have a bar code for. I imagined that this could be the basis for a plethora of niche apps, each serving a different kind of niche: keep track of your books, CDs, games. Keep your inventory current, never again miss some food expiration date.

I thought that many app developers would have many users and those users would then add products to the global database if they encounter one that couldn’t be resolved. Monetization would be done from app developers who had the most web API traffic and at a later stage from letting product manufacturers analyse the sentiment about their products.

Since I couldn’t do all by myself – being specialized in iOS apps – I pitched this idea to several of my friends. The first to believe in it besides me was René Swoboda who developed the entire backend for ProductLayer. While he was working on that, I tinkered on the iOS app which I modelled after my favourite Twitter client.

Two more joined our team: Roland Moser as CEO and Werner Bayer as developer of the Android app and also versed in the backend magic to give René a helpful hand.

I was supposed to find and “evangelize” developers who would use our API in their own apps. But it soon became apparent that very few people would buy into my vision of a “common good” product database. But we pressed on, working to release the first versions of, hoping that some users would feel the need to share opinions (we called them “opines”, similar to “tweets”) about specific products, and in doing so help grow the database.

Unfortunately our user base never really got off the ground. My girlfriend and René’s mom where the most faithful users besides ourselves.

Deus Ex Machina?

My second idea to grow the database was to employ machine learning to the problem. Couldn’t you teach an AI what pieces of information are relevant to understand what a product is? Along the lines of, if it is in a bottle it must be liquid. If it says certain words on the bottle – like “body lotion” – then it is probably not potable, but cosmetics. Very briefly we spoke to a guy who could have been heading this, but he and us couldn’t agree on a mode of compensation for him.

Next we started scraping a number of affiliate feeds and that got us really many products. Mostly consumer electronics gadgets because those are the ones that are sold most over the Internet. This grew our database to more than 6 Million products today. This is nothing to sneeze at. Although a slight problem with this approach is that every now and then one of the affiliate partners decided that they didn’t want us to have their data and so we were asked to remove their product photos.

Early on, my partners convinced me that we’ll never be able to charge anybody for API access, if we don’t first have the most complete database possible. So, if app developers wouldn’t be our paying clients, who would?

We talked to a few companies building cashier terminal and online shop software. But they took issue with many of our product pictures not being “originals” from the manufacturer or not having white background. Those you could only get from the manufacturers but not from our crowd taking “in action” snapshots of the products they were testing.

Authentic, True and Guaranteed, Not

Another hitch with people being at liberty to saying anything about products would be that bad products would carry bad comments and that would make it hard to suck up to the manufacturer of said product to provide more complete product lists and information to us.

In particular with foodstuffs it became apparent that certain critical information (like ingredients or allergens) would have to come from the makers of the products themselves, because legally they would have to be responsible for it being absolutely correct and true. I know of only one company which is trying to do that: GS1 International is working on a cloud/server-based solution which aggregates and disseminates “originally sourced” information about products. But of course that comes with a cost, nothing is free.

There’s another (kind of) competitor to us, who had the unfair advantage of a lot of funding: Semantics3. So far these guys raised 2.2 Million Dollars – and they do exactly two of the things I mentioned above: scraping affiliate feeds (and web pages) and using machine learning to clean up the mess into coherent products. At the same time they gather pricing information. They send out extremely annoying emails laden with animated GIFs touting the benefits of their services.

Well, I tip my hat to them, because they more or less achieved the vision I had, minus a simple free API for app developers. Oh well, they go where the money is, online commerce.

An Amazon-Killer?

Around two years ago, when it dawned on us that our investments of time and money wouldn’t lead anywhere, we came up with an idea for a potential pivot. Everybody is angry with Amazon, but nobody does anything against it.

The idea for REGIZON (a regional Amazon) would be to let small local brick&mortar shops list their products, prices and location in a very convenient way. Let’s say you quickly need a certain toner cartridge so that you can continue working today. You take to REGIZON’s friendly user interface and find that the shop down the street has exactly that in stock. So your feed do the instant delivery and you don’t have to wait. You benefit and so does the shop.

Again we pitched this idea to several big players in Austria, but we never got further than some initial excitement from potential investors. It seems every does agree that “something must be done”, but nobody actually does.

Why we Shut Down

Those 4 years of ProductLayer were not a total waste of time. We three software guys agree that it certainly benefitted our skills to have to keep up a service like this. There are many things in the API/SDK design and the iOS app which I am personally proud off and I am sure René and Werner can say the same thing about their areas of expertise.

We even had a hand full of actual developers doing product queries via our API. But having taken 4 years to make it to 4 API users is a very sad statistic. In the months leading up to our shutdown we even had to turn away one or two people who where hoping to use our API, one for research, one for an actual app.

However much we love ProductLayer, we cannot keep it running. The main reason being that it costs time and money to maintain our redundant servers. We simply reached the end of our initial funding of 10k Euros and nobody is willing to keep putting down additional funds to keep our hobby alive and going. We were quite frugal dare I say?

At the same time this hard funding limit kept us from making a bigger mistake. If we had invested more, then our loss would probably have been much greater. As we see it now, there was no market receptive to our ideas. Let alone one that would pay us for them.


I’d say my own personal learnings are those:

  1. You need to fail fast and be able to quickly determine if an idea has a potential for paying clients.
  2. There is a very real danger that if you invested more than required by a minimum viable product, that you fall prey to the sunk cost fallacy.
  3. Trying to build a platform requiring extraordinary amounts of users to work is futile.
  4. You need to be able to afford investing time and money into something that is likely to fail. Better to be doing that on the side or as a hobby.
  5. You will see your own passion for an idea evaporate, once you have commitments.

We cloned our system in a virtual machine, so that maybe one day somebody could make use of the product data we collected. I was the former chief (and only) Evangelist for ProductLayer. You may contact me by email if you have any questions.

Also published on Medium.

Categories: Business

Leave a Comment