Startups win by learning fast

Robbert van Geldrop
June 19, 2018

In the past 3 years we have been tracking the progress of many startup closely, because we organised mentoring across multiple startup accelerators and programs.

This weekend I decided to aggregate and slice the data of our 528 log reports of coaching 111 startups in the past 2 years across 7 different programs. Most of these startups were really young when they entered these programs.

I found some useful insights while digging through all this data thoroughly:

  • Successful startups are in constant flux with their customers and produce learnings fast.
  • When building a technically complex product, successful startups make their customers do things which are leading indicators for purchase intent.
  • Disruption can only work if the value and type of customer is very clear from the beginning.

Below is a detailed description of the data we got, how I analysed in and came to the above conclusions — including some examples. Also, we have some recommendations for accelerators to improve their program.

What we tracked

The log reports are mostly filed by mentors and program managers. We tracked the following:

  • Milestones the startups agreed on with the program managers
  • The gaps in the business model, based on the Business Model Canvas
  • The progress on the milestones made and reported by the startups on a bi-weekly basis

I enriched the data with a couple of public data points per startup:

  • Public information on funding rounds
  • Number of employees on LinkedIn and/or their own Team page
  • If I could not find traces of the startups online or had a clear sign they were out of business, I marked them as such.

Here are some highlights on the data:

  • 21 out of 111 startups are out of business. That’s 19%.
  • The startups raised approximately 8,5 million euro in funding. However, the majority of the startups have not raised funding yet. I’m only sure about 14 startups, but the data listed on the sites of some accelerators show higher amounts.
  • The startups employee approximately 4 people on average, which includes the founders. Only 12 startups employ more then 10 people. Only 2 of those companies have announced funding rounds publicly.

Let’s have a look at some aggregate data of the log reports. The question ‘On which part of the business model should the startup focus?’ delivered these results:

  • Value Proposition = 225 times
  • Customer Segments = 225 times
  • Revenue = 162 times
  • Channel = 149 times
  • Customer Relations = 109 times
  • Key Activities = 97 times
  • Key Partners = 51 times
  • Key Resources = 36 times
  • Costs = 35 times

We can compare that to the milestones the startups came up with:

We use milestone categories which startups can associate to actions. This is not entirely the same as the business model focus, though.

  • Customer Understanding = 103 times
  • Product = 131 times
  • Marketing and Sales = 128 times
  • Partnerships = 66 times
  • Metrics = 53 times
  • Work Method = 43 times
  • Finance = 30 times
  • Basic and legal stuff = 30 times

What does the data tell us?

On aggregate level, it’s clear that most early-stage companies struggle to find the right target customers and articulate their value proposition. It’s also interesting to compare that to the milestones these startups set regarding funding and the concerns the mentors have regarding costs.

Funding is not the main topic of discussion in most of the checkins and since most startups in the program are effectively only the founders, it means their runway is probably long enough to make it until the end of the program.

Hence, the aggregate data doesn’t teach us something we wouldn’t already know. Most startups fail, because of ‘No market need’. This is what CB Insights has proved already and if you look further down the list you’ll find more failure causes in the same category, such as ‘Ignoring customers’ or ‘Product without a business model’.

The true value of acceleration is education

Since most accelerator program managers are well aware of these failures, they craft curriculums, which train the founders to avoid these problems. In some cases they do that so well that these founders can skip the follow-on funding stage in favour of paying customers.

Looking for patterns

In order to get to insights beyond what’s known in the industry, I had to dig deeper and look for patterns. Also, it’s hard to strictly draw conclusions from the data by itself. Here’s what I started doing:

  • I tracked paths of successful and failing startups. What were their coaches advising across multiple checkins?
  • Correlated those finds with my personal experience, to pair the data with qualitative insights. I’ve coached some of these teams myself. What was the sentiment of those meetings?

However, when analysing the patterns across batches and pair that with my personal experience and those of the mentors I’ve worked with closely, here is what I found.

1 — Successful startups are in constant flux with their customers and produce learnings fast.

The pattern across these startups is that the mentor checks off a different part of the business model after each session every few weeks. They move from Value Proposition and Customer Segment to Channel, Revenue and other things.

Basically it means that the founders have identified potential customers and have a clear understanding what value they bring to these customers. They do so by following through on engaging with customers with the goal to produce learnings and insights. This only works if their customers are interested enough to also appreciate the company’s efforts.

From personal experience I also know that I had a good relationship with the startups who followed this pattern. The founders tend to be good listeners who follow through on advice. This doesn’t mean that they’ll blindly accept advice I’d give them, but they do process it often by triangulating it.

2 — When building a technically complex product, successful startups make their customers do things which are leading indicators for purchase intent.

We all know the classical examples of how startups proved purchase intent through landing pages. The most known two examples are:

However, for many startups it is not feasible to copy these online experiments. For example, many business-to-business startups have a hard time to get potential customers to come to their website or landing page in the first place.

What remains is to simply go back to these potential customers and keep hoping that they’ll buy the product. The risk of that is that these customers will not buy and are simply pushing the conversation to learn from you just as much as you want to learn from them. What to do to get real commitment in case you can’t sell them a product yet?

Here are some examples out of Tristan Kromer’s Real Startup Book.

  • Broken Promise Smoke Test — Send something in confidentiality and track whether they still decide to share it with others, because their excitement trumps your request not to share.
  • High Bar Smoke Test — Make the customer go through tremendous effort to get something from you. For instance, organising an unveiling of your prototype on a very remote location.
  • Concierge Test — Testing the value proposition by delivering it as a service and getting paid for that.
  • Selling information products as MVPs — If you have interesting findings based on the industry knowledge and insights you generated, sometimes you can sell this to your target customers to see if they’ll transact. (This one is not listed as an experiment by Tristan Kromer specifically).

The principle behind these experiments is to match the customer’s words with actual behaviour. Rob Fitzpatrick, author of The Mom Test calls this Learn & Confirm. It’s an indication of** applying Lean Startup as a mindset.**

3 — Disruption can only work if the value and type of customer is very clear from the beginning

A few startups have a very interesting pattern, in which desirability (basically the right side of the canvas) was not in question at all. Basically, the big challenge for these startups are related to their Key Activities.

Here are three examples:

  • Social networks like Facebook and Twitter sell advertising space to customers. Their biggest challenge is to get enough users. This is a Key Activity on the Business Model Canvas.
  • Uber offers services people use every day. Their biggest challenge was to get drivers so the customers would be able to book a taxi via their app. Again, it’s about the Key Activities and Resources.
  • One startup I coached wanted to build up a recruitment network through friends and ambassadors. It was clear that if they could pull that off the customers would be willing to pay for it. Their service would be much cheaper then comparable recruiting services. However, building that network of ambassadors turned out to be really hard. Again, the most risky part is a Key Activity.
  • Nerdalize, a cloud computing startup, wanted to distribute computers to homes as free heaters. It was clear that if they’d be able to pull that off they’d have a very competitively priced cloud computing service. They established a partnership with Eneco for distribution of the computers in the homes of Eneco’s customers. Nerdalize still had to figure out who would potentially buy their cloud computing capacity and what concerns they would have.

Lots of marketplace startups have this particular challenge, where the hardest part is building up inventory instead of finding interested customers.

Conclusion - Accelerators need to help founders build up the right mindset and focus

I wrote a lot about the behaviour of successful founders, but lest we not forget that 21 startups went out of business and at least 66 did not get past the small team setup by now, which might indicate that they too are bound for failure. What makes them fail?

The data shows that many of these startups are wheel-spinning and do not get past the point of the right value proposition and the right customers willing to pay for that. During the accelerator programs the workshop and coaches do advise them on this, but yet it doesn’t happen. There are three main reasons for this:

  • They’re not comfortable engaging with customers before the product is ready.
  • They thought the program was about building the product and getting funding before anything else.
  • The team is not able to execute on the idea due to lack of skills or knowledge.

Accelerator programs can address this in various ways:

  • Help them do it. Just make yourself available and join the founders on their first customer visits. from a program perspective.
  • Invest in peer-to-peer learning that drives actionable outcomes and (again) helps founders overcome mental blockers. The folks at Source Institute have developed awesome tools and formats for that.
  • Consider and assess the team’s ability to execute very diligently. Introduce options in the program to form or expand their teams.
  • If your accelerator invests in startups, consider cutting the investments in half, creating a midway point and double down on those who are making great progress.

The purpose for accelerators is to… accelerate.

Yet, many teams who come in have nothing to accelerate in the first place. They’re still figuring out what their business is about, who their customers are and what problem they could solve. Startups who get past that point do benefit from accelerators as they now need help to focus on growth and can follow through on advice from the coaches, network introductions etc.

However, this is currently the exception, not the rule, unfortunately.