Shared values predict startup success? An interview with Saberr

Introduction

As part of our ongoing research we have been looking at the best ways to go into entrepreneurship. When we talked to Matt Clifford, of Entrepreneur First about the question, he suggested talking to Saberr. Saberr are a small startup focused on the question of predicting the success of teams in business settings, and they have already had some impressive successes.

We spoke to Alistair Shepherd by phone, one of the two original founders of Saberr, about their perspective on forming a successful entrepreneurial team. The following is a selection of highlights from the call, edited and reorganised for clarity.

Key points

According to research by Noam Wasserman most startups fail because of their team, suggesting team composition is important for entrepreneurial success.
While standard personality tests have not been shown to be very successful at predicting success in careers, Saberr have achieved some impressive, if small scale, predictive success using a model based on value alignment and behavioural diversity.

The interview

What does Saberr do?

We help organisations optimise their workforce by designing internal teams, or by recruiting candidates who are actually more likely to fit well with the rest of the team. In
Jim Collins’ book “Good to Great” one of the principles is that companies whose workforce is aligned on values is 6 times more productive than their competition. A rather startling statistic but it’s not new or revolutionary. The question is how do you find employees who are aligned on their values or how do you hire with that in mind?

The problem is that there is no real way to do that at the moment. If I were to ask what your values are or if I were to ask myself what my values are i would probably list every value known to man, because humans aren’t very good at knowing what our values are explicitly. But, when we meet someone, we’re very good at knowing whether we share the same values. And we don’t ever ask them what their values are, we just sort of know it, and we actually know it very quickly, within seconds. And our goal was to understand, from a data perspective, when two people’s values are aligned, without them explicitly telling us what their values were, and we built an algorithm to do that.

When I was studying in the states one of my professors was Noam Wasserman who has written quite a few books about entrepreneurship including “The Founder’s Dilemma”. His research found that 83% of startups fail, which is a fairly standard statistic, but the one that was interesting was that 65% of the time they failed because of the team; the team members just didn’t get on with each other or there was some issue that caused the team to fragment and ultimately cause the startup to fail.

Now my background is engineering and my attitude is “why do we keep failing at this whole team thing? You know humans aren’t magic we’re just really, really complex. So is there a way to predict team success and is there data that can help us to do that? Obviously there is a lot of psychometric tools out there at the moment. There’s a lot of research around what makes a good team, but on its own it’s not a good predictor of team performance and it also turns out to be fairly conflicting if you read the academic literature. But there is one common theme which is basically that a diverse team makes a more successful team. The other piece though is this value alignment.

So we looked for patterns in online dating to see if we could establish what makes people click. Because online dating is a numbers game, you go on enough dates with enough people and eventually you find someone who is significantly better than all the other people you went on dates with. And when you do that you leave the dating site because you are now in a relationship, so the dating site serves no purpose for you anymore. A lot of people tell the dating site why they left. There’s an option that says “I left because I found somebody on your platform”. If you then look at all the questions that they answered that were the same and you have enough people in this data set there are very clear patterns.

So we managed to figure at a way of asking seemingly arbitrary questions to understand whether people were aligned on their values or not. What we’ve been asking is what makes a successful entrepreneurial team? One of our philosophies is that anybody can be an entrepreneur. Now that might be wrong but it’s something that we have tried to include in our research. The thing that makes or breaks you as an entrepreneur is whether you are with the right team or not. We haven’t done any research into what types of people make more successful entrepreneurs.

Who are your current clients mostly at the moment?

Mostly high growth businesses – business that are growing quickly and so are hiring a lot of people. We’re targeting series A and series B funded startups, but we’re operating on a minimum viable product at the moment. But as soon as we’ve cleaned it up and made it a lot slicker it will be applicable to much larger companies. PwC has approached us three times to ask “have you finished yet? Have you finished making it a slick product?” and we’ve said “No”. We’re working on it, we’ve done some work with Coca-Cola, we’ve done some work with Microsoft but our product isn’t at a stage that they can rely on, so we’re working with young companies.

When we spoke to Matt Clifford he said you are working with Entrepreneur First at the moment, what sort of things do you expect to be able to predict there?

We’ve been working with Matt to help him think about the team design in their cohort and really we’ve just been giving them feedback, saying “this team is going to do well, this team is going to do badly” and telling them not to do anything about it. So we had a lot of discussion around it – to get validation and build their trust in the system. Because originally when we spoke to Matt about a year ago, he was originally very skeptical and he said “I don’t believe you” so I said “well, invest enough time in us to have your cohort be in our data set and we’ll just keep drip feeding you the information and you can tell us if we are right or wrong” And we’ve been right every time. Going forward what we’d like to help them do is design their teams from the outset and to design their cohort.

So your product is essentially a piece of research at the minute?

Yes. We built an algorithm to see if we could assess whether we could predict a successful team or not. We took it to a lot of short term entrepreneurial competitions like startup weekends or the Microsoft imagine cup or a bunch of internal events that different universities were running. And the interesting thing was that our very first iteration of this, we took to Bristol University and there were 8 teams competing in a week long business plan competition. And we made predictions right at the start of the week as to which team was going to win. So we didn’t know what their business was, their plan was, or what their skills or experience were. We simply knew the answers to the questions that we had asked them which were thing like “Do you like horror movies?”, “Would you consider having an open relationship?” so we said “we think this team is going to win and we think this team is going to lose” and we ranked the teams 1 to 8, and we got the ranking perfectly correct. Which is interesting because it’s difficult to do by chance, 8! (40320), that’s a fairly big number!”

Do you have any idea how your algorithm compares to other predictors of success in a job, like the other recruitment literature on how predictive things like structured interviews, IQ, conscientiousness, etc. are?

I don’t have hard metrics on how good that is. All I know is that our software is incredibly good at predicting the outcome of short-term team success. I don’t know how that translates into long term career success. Having said that, we did a few tests with corporations, where we went in to their product development team. For example there was one company in particular we worked with in Australia, with a product development team of 20 people. We said to them “We’re going to tell you who your top performers are and who your bottom performers are, and you’re going to tell us whether we’re correct based on your own internal KPI’s”. Some of the people had been working there for 20 years, some of the people had been working there for 6 months and there was a range in between. We had no idea about their seniority, their length of service or anything like that – we only found out about that afterwards. And we got it perfectly correct. We ranked their employees from “this is your top performer” to “this is your bottom performer” and they were like “wow that’s magic because we know after they’ve been working here for 20 years and we’ve got KPIs for their performance. You know that after they answered a couple of questions for you guys and you clicked a button on your computer.”

You’ve mentioned some pretty impressive successes, have you had setbacks as well?

Unfortunately not.¹ I wish we had. We’ve tested about 21 times now, at different events. All short term entrepreneurial competitions or corporate teams. I find the top performers and the bottom performers, and every time we’ve been greater than 95% correct. It’s a good problem to have, but it’s annoying because if you think about the way people perceive information or build trust, as humans we like to see people overcome certain barriers or walk up a staircase in terms of success. This failed but we tried this and then it succeeded. That’s much more compelling a story and it’s much more convincing than if you waltzed right in there and get it right on the first go. So from a psychology perspective I wish we had failed but we haven’t.

It is a bit surprising that fit with team alone could predict overall performance so well because you would think that there would be some other factors like IQ that would be quite important in something like that.

Sure. Originally I thought it was quite surprising, but the more l looked into it the less surprising it became. Yes, there are a bunch of factors that contribute to your performance within an organisation: the quality of your skill set, the breadth of your experience and your fit with the team – to pick three of the top high level ones. Our hypothesis, and we’re constantly seeing that we’re correct on this, is that your fit with the team is the biggest predictor of your performance. So if you have a team that fit equally well with each other – they have equal fit – you have other factors that become the biggest indicators of performance. But more often than not you have a range of fit within the team and that is the biggest indicator, so the other ones pale in comparison. It’s not that they’re not important, it’s just that they’re not as significant in predicting performance as something else, and we think that something else is team fit. So when we predict performance, we’re not actually analysing an individual’s personality, we’re aggregating they’re fit with their team.

Are there any other companies like you out there?

There’s a lot of companies who are doing stuff around the behaviour side of things. You’ve heard of Myers-Briggs, Belbin and psychology tests like that. There are lots of organisations who are basing their methods and their tools on personality insight. I don’t think it shows anything and I think our method is not only measurable in terms of its success, but also a smarter approach. But there are a bunch of companies who are using the behaviour aspect or the personality aspect in their tool. There’s only one other company that we know that’s using a data driven approach to value alignment. That’s eHarmony – they’re about to enter the recruitment field in a couple of months, which is potentially a game changer, but they have a sort of stigma about them in that they’re a dating site. A lot of companies will be concerned about creating lots of relationships in the workplace because that could go badly wrong.

Based on this, say for somebody who is interested in starting a startup themselves. Do you have practical advice on what they should do based on the data that you’ve got?

At the moment we’re attacking a very complex problem and so a lot of the advice we give to teams is generated by a computer because it is just too complex for a human to think about and solve. So we don’t have any hard and fast rules like “when you’re thinking about making a team you should think about X” because we have a computer to tell us whether you’re going to be a good fit or not. But I guess that the overriding piece of advice is “build a team out of people that you like”, they don’t need to be your best friends, but go through some really shit times together and if you’re still friends at the end of it then you’ve got a good team. But of course that’s logical advice, it’s not groundbreaking.

The algorithm is finding two things: it’s finding people who behave differently from you but people whose values are aligned with yours. That’s a large part of what we do, the computation of how those two things are interrelated. A bad team is one where everybody behaves the same. But if everybody behaves the same but has unbelievably aligned values, your chances of success are better than a team that is behaviourally really diverse and has a great behaviour split, but value wise absolutely hate each other. So it’s not a simple black and white problem, there’s a lot of overlap.

Can you give an example of what you would do to measure behavioural diversity?

At the moment we ask them questions again. That’s based on all the research around the big 5, Belbin, Myers-Briggs, all these other psychometric models we looked at them and said, although they disagree on the nuances, where do they agree and can we build a model out of that?

Have you been using your algorithm for yourselves?

We definitely use it for ourselves. In fact we don’t interview somebody unless they fit with us well in our algorithm. When we do interview them it’s just to find out what their skills are, because before that we don’t know. It’s just to find out whether their skills are precisely what we’re looking for and basically to see if their goals are aligned with our goals. You know, is the salary that we’re talking about something that’s attractive to them, is the seniority something that’s attractive to them? We do the interview in reverse really and it works quite well for us.

How well does your team go on your own tests?

For a long time we were the best team that we had ever analysed which was convenient. But it didn’t start that way, we had two different people who were in our team at the beginning. So there were 2 cofounders including myself and we had two other guys who were working with us, neither of whom worked out well. Once we had figured out our algorithm we used that to hire our team, and since then we’ve had the dream team which has been great.

Finally, can you use your algorithm to predict dating success?

Probably, but we haven’t looked at it, and we’re staying very well away from it! It’s not a field that I want to touch.

Notes