Welcome to the inaugural Startup Of The Week – a feature dedicated to exposing the most exciting and innovative startups from the UK and beyond.
The first startup to star in our weekly "Startup of the Week" feature is Mendeley, a free reference manager and academic social network founded by three students from the Bauhaus-University of Weimar.
Less than a decade old in its development, the cross-device app has been adopted by the likes of Oxford, Cambridge and Stanford universities, and through a network of student ambassadors to spread the word, the future is looking good for the startup.
Mendeley CEO and co-founder Dr Victor Henning achieved a Ph.D from the Bauhaus-University of Weimar, where he researched the role of emotion in consumer decision making. He has also worked as a talent scout for Sony Music/Columbia Records in Berlin. He co-founded Mendeley in 2007.
All highs, lows, innovations and deviations - Victor took ITProPortal through the Mendeley story.
How did the idea behind Mendeley come about?
Mendeley grew out of an idea that I had together with my co-founders seven years ago.
Let me just give you a timeline; I was a PhD student in 2005 and every researcher I knew had the problem of managing documents on their computer. PDFs, research papers, all flying around everywhere in different folders on the hard drive etc., and no good way to keep track of them or make sense of them.
So I started talking to a friend of mine, Jan, who was also a PhD student at the time and we decided, why don't we try and develop a software that can automatically extract all the relevant information from pdf documents like the author, title, volume, journal, issue, everything you need to keep track of your research papers and have that software build an automatic database from the files in the hard drive?
That was the initial idea and, at that point in 2006/2007, we just thought of this as a side business. We found a company in Belarus to build a prototype and thought we'd just continue down our regular career paths.
I was on track to become a professor and Jan was doing consulting work for SAP, the German software company, and I think it was around that time we realised that if we actually had software that could extract this information automatically from the research papers on a user's computer, if we then get the software into the hands of millions of researchers, then we could crowd source all of the data that they extract locally from their hard drives and pool it to build an amazing research database, allowing us to keep track of what's happening in research in any given field of science today – what is it that biologists are reading today, what are mathematicians reading today, what are the overlaps between different fields, what are the relationships between papers?
How did the business develop?
That was really the starting point for us to go out and raise venture capital and relocate to London where we started the company with a third co-founder, Paul, who did a masters at the university where I had done my PhD.
So in 2008 we raised funding from the founders of Skype, some of the early investors in Last.fm, the music service and then subsequently also from the owner of Warner music, who we got introduced to through the last.fm connection. And they stayed as our investors right until the point where we got acquired last year.
It grew from the three of us to about eight people in the end of 2008. Then 20 by the end of 2009, 30 by 2010. We opened an office in New York around that time to do more outreach over there because our user base was largely from the US.
In fact, the very first office that we rented in Covent garden was Michael Palin's office from Monty Python and he would stop in from time to time. If you were going to the copy room to get some paper and there'd be a Spanish inquisition wind up doll!
How far does your user base extend?
Our biggest user base, in terms of universities, is the top universities in the world, so Stanford, Harvard, MIT are actually paying customers of ours. In the UK; Cambridge, Oxford, UCL and Imperial are our biggest clients, so we've really managed to capture the high end of the research market.
Simultaneously, our expectations of growing this data base have actually worked out exactly as we had hoped. Our users have uploaded over half a billion documents to our database, so it's probably one of the largest – if not the largest – academic database in the world, and the vision that always pursued was that you could do much more with this data than we could do ourselves.
It's quite similar to the Amazon collaborative filtering approach, where people who have bought this book have also bought that book, and we can say if you've read these research papers then those papers may complement the stuff that you've been working on. So that's one application of the data that we have.
Another one is that we have become the largest source of real time metrics of what is happening in science today – research trends around particular papers as well as topics and authors.
We opened an API that anyone can access under a creative commons license available at http://dev.mendeley.com, and there are now more than 300 third party apps that tap into our database, and they have built a wide range of scientific applications, from being able to synchronise your Mendeley library to an Amazon Kindle, to Android, to analysing the relationships between papers in your library to get a mind map of how the research in your library relates to each other, to matching up your library with genetic information.
There was even a hack where people could upload their raw genetic information to a website and could then retrieve from Mendeley the research related to their particular genotype.
The social network
What is an academic social network? How does it differ from the ones that we know as consumers?
The idea behind it was always that, we would use the data in your Mendeley library, so the documents that you're reading and that you're researching, to connect you with like-minded academics.
When you sign up to Mendeley you get a free profile on the site, like LinkedIn or Facebook, except that it's about your academic CV and you can upload your publications, which is done automatically for you so it saves you a lot of hassle. You can then start to connect and follow people in your field.
One of the most popular features of Mendeley is actually a group functionality where if people are collating information around particular topics of research, say there's a group about the latest trends in neuroscience which has thousands of followers. So it's a couple of academics getting together and deciding that these are the papers that are really interesting for neuroscience and then anyone can follow that group and get the updates straight to them and you can discuss why is a paper relevant.
[Mendeley is] available on the web so you can just access everything in a browser but the main document functionality resides in our apps, which are available on Windows, Mac, Linux, iPhone and we have third party Android apps and are currently working on our own Android app as well. Also, through those third-party apps you can synchronise everything to a kindle.
What were the main challenges you faced in developing the product?
I'd say the first challenge was finding good staff. Especially if you're growing as rapidly as we are, trying to fill a recruiting pipeline with 30-40 good engineers per year is quite difficult. We've recruited quite a lot from abroad; we've brought Americans over to Europe but also from other European countries, so I think at last count there were about 20-25 different nationalities working at Mendeley.
In terms of the actual software development, I think it's the complexity of what we've built. If you imagine, you have basically a couple of different clients across all the different platforms – apps, analytics, mobile and web – you have 3.5 million users in total, and in any 24-hour span they're adding more than a million documents to your system, the system has to be able to, in real time, ingest those documents and de-duplicate them.
What we do is, say that you and I upload the same paper that's just been published last week, the system obviously needs to realise that that's the same paper so that we can then generate those research trends and statistics and say that this paper has now been read by, say, 6000 people and they're from those fields, those academic disciplines, those geographic regions.
There's a lot of data processing happening in real time in the background. You have a database of more than 500 million documents to de-duplicate against each other and build recommendations – where you basically compare each item to another item to create an enormous matrix of 500 million document comparisons.
We've invested a lot in our data science team and they've managed to bring down the processing time on Amazon Web Services for running a new batch of recommendations from more than 24 hours initially to now a couple of minutes, so we now run it several times a day.
Those were the big challenges – scaling a system like this across many different platforms and making it all work seamlessly, because its mission critical for people. Research labs at Stanford and Harvard are trusting you with their data and they have tight deadlines they're submitting for conferences or for grants and they have to trust that you keep they're data safe and keep it available, so setting up a system like this was the main challenge.
Do you have a particular approach to get universities interested in the service? Can these institutions quickly just switch it on and go?
We were quite aware of the slow decision cycles in universities when we launched, and so we tried to do a bottom-up, grassroots approach rather than a top-down adoption. From the start we tried to target end users first.
When we launched I did a speaking tour around the colleges and universities of the US east coast, but we quickly figured out that the best way to get adoption was to use our own users as advocates. We'd always involve them in the development process by being very open on our blog about what we were planning next, what the future roadmap looked like.
People then got engaged and said: "I can see where you're trying to go, how can I help you get there faster, how can I help you spread the word." So we set up something called the Mendeley advisor programme.
We now have a community team of three full time staff, and those three people manage basically a volunteer organisation of more than 2,000 volunteers of phd students, post docs, professors on campuses around the globe, and in their spare time they actually teach how to use Mendeley to undergraduates, they put up posters and presentations, and they introduce us to librarians.
That was how Stanford turned out to be our first customer. I had a couple of Stanford professors come into the meeting and tell the librarian that "I've been using this in my lab and I make all of my phd students use this and I think we should get it for the whole campus." I think that was how we were able to, in a pretty short time, sign up lots of these major institutions.
Do you have any advice for would-be start-ups?
I think the most important thing is choosing your co-founders wisely. I think most of the start-up failures that I've seen went back to dysfunction in the founder team, when the founders just didn't share the same vision, didn't share the same work ethic, didn't share a team spirit.
I've seen start-ups break apart because founders tried to just build their own little teams within the start-up to try and increase their influence, whereas most of the successes I've seen were teams that just worked together extremely well.
My second most important advice would be, likewise, choosing your investors wisely because, I've seen many start-ups make the mistake that they just take the first money available because they just think money is money. Conversely, I've seen other start-ups be really smart about how they pick their first investor. You have to aim high with your first investor and they aimed high with the first one.
The third most important one is, I think you have to realise that a start-up is going to be an ultra-marathon. When you start a company you have to be prepared to run it for eight, nine, 10 years or even longer because that's how long it takes on average to really build something of lasting value that turns out to be a success.
Along that route you will have turbulence. From the outside people might think that everything always went smoothly for us, we raised successful amounts of venture capital, we grew with the company, we won tonnes of awards and we then had a fairly successful acquisition, but we went through awful lows where at some points the company didn't grow or the user base didn't grow and we had to lay off some of the staff.
We made wrong choices in how we structured the organisation, we've almost been screwed over by venture capital firms, we were close to running out of money and the company going bankrupt, even though we were growing, so you have to prepare yourself for these incredible lows and be aware that every company, no matter how successful they are at the end, goes through this.
Mendeley is available to both individual students and institutions and operates within major academic establishments across the world.
Look out for our next Startup of the Week feature, next week, and of course if you're part of a growing business yourself, let us know using the Contact tab below.