Recommend, a platform which enables its users to find recommendations such as restaurant reviews from trusted, human sources like friends and experts, was a finalist at LeWeb’12 Paris. Since doubling its VC investment post-LeWeb, the startup has come a long way. CEO Nicolas Mendiharat explains just how far…
LeWeb Blog: How do Recommend’s recommendations work?
Nicolas Mendiharat: Music, video and other media are just one part of Recommend – we’re looking at the global picture. Recommend was a standalone service at first, with 10-20 recommendations per day. Before, people would ask others for recommendations, or receive them over lunch or a phonecall. They might be “how can I find a babysitter for tonight?” or “a hotel room in Barcelona”. We built Recommend as a social network, more to rely on people you trust than on crowdsourced answers.
Indeed, when it comes to recommendations, there are two approaches. Some are crowdsourced, like on Tripadvisor; or algorithm-based, like on Netflix’. That’s not what Recommend does.
Instead, we use a tool called import.io, which allows you to turn any website into data, so you can extract anything you like from that site. That way, we can source recommendations from Time Out, for example. They are then brought into our site, and represent super-qualified data. Then, inside Recommend, you choose what sort of data you follow, and get the recommendations you want.
Not long ago, Mashable published a very interesting study about online recommendations. Based on 55k people, it showed that friends and experts are the the two top levels of trust. Not automatic recommendations! These work well right now, but only , notably because there’s a void in the top 2 layers of trust. That’s where we come in.
Another study shows that word of mouth is becoming even more important, as we’re completely bombarded with information these days. As such, 3 of your friends recommending a film is always more powerful than thousands of people you don’t know recommending it on Netflix.
The importance of focusing on the top two layers derives from the fact that you can’t have one without the other. It has to be social, so you’ll get thanked and/or rewarded when people take your recommendations (they’ll share them, they’ll like them, you’ll get more followers, etc.). In the end, you can gain influence as a recommender. We’re still building this 2-layer approach right now.
Our first 5,000 users taught us that our first version wasn’t good enough, and that we didn’t have enough technological resources. So last December, we decided to add the import.io element. This resulted in 60k recommendations by late September, versus 10k before. Our website will be public at the end of May, and we’ll launch the iOS app soon. Our major feature will be the “Top 10 Lists”, an example of how we’ve been listening to what users want. In other words, our responses will be based on feedback from hundreds of users.
> Did taking part in LeWeb help with this iterative approach?
Yes. Back then, we only had a prototype and explained what we wanted to do. In following months, we determined what we should start with. LeWeb gave us our first feedback – but that was just the initial step. Our first release was in November 2013, so LeWeb was like a strong push at the beginning: a lot of feedback told us how to solve problems were addressing, and thus confirmed our project was something a of people would want to use. It was basically a proof of concept.
LeWeb also gave us our first actual users: they were highly qualified, and gave us high-quality feedback. In the end, LeWeb exposure accelerated our development and our understanding of how we should execute this great ambition. We’ve had €1 million funding so far! And we’re now close to the launch, based on everything we’ve learned in the past year.
If you look at the big picture, building a tool that will be used as a Google alternative is quite a challenge. The user promise is “save time to make better choices”. It sounds great, but it’s not easy to make!
> What is Recommend’s situation in terms of funding?
We received €550k last February, just after LeWeb, from angels I’d already met. LeWeb enabled us to make that a bit bigger because it inspired more confidence from investors. We were on track to raise €300k, so taking part in LeWeb practically doubled that.
We’ve also built an algorithm that determines a trust score for each Recommend-er; it’s a bit like a Klout score. YOu win points for every followed or liked recommendation that can be traced back to you. Furthermore, it can be totally personalised and customised. Imagine you connect with Facebook, and say you like skiing, golf, or live in Paris: all future recommendations will be based on that profie. That way, two different people will never get the same recommendations.