Category Archives: Startups

Startup Accelerators are dead

Startup Accelerators have been around since 2005 when Y Combinator was born and since then we have seen literally thousands of them created across the world. But I believe the model is broken and that they have had their day. Sure, we will see a few new ones, particularly large corporate-backed ones, but we will also see lots silently close.

To understand why the model is dead you have to understand why it was born in the first place. The ability for a couple of smart people with laptops to create a business that could grow to be worth a billion dollars with very small amounts of capital has only been around for the last 10 years. Before that you needed lots of capital, big teams and years to develop a product and take it to market. The Internet changed that – we could now build, deploy and scale a web or mobile application for a few thousand dollars.

Capital Injection

The first challenge that the Accelerator model solved was an injection of a small amount of capital into the startup, typically $20 – $30 thousand. This was money that founders often couldn’t lay their hands on, or required them hold a full-time job, restricting the time spent working on their startup. The injection of money brought them 100% focus on the startup, which meant they could do things much faster. They could also hire in the expertise they didn’t have. This capital injection was in the form of investment and typically took anywhere between 6-12% of the equity of the company. Accelerators were willing to take the high risk, providing capital to early stage startups when existing investors wouldn’t.

Expert Support

Accelerators provide a range of experts to support the startups they invest in. This support ranges from external volunteer mentors through to dedicated accelerator resource, available on tap. This is important because for the most part the startups joining accelerators have no previous experience of running a business, let alone the knowledge to scale one fast. I feel that the importance of this is often undermined, because it is difficult to quantify its value. In my experience this expert support is often the key driver for the progress of the startup – not the money.


Most accelerators last for 3 months, a marker laid down by Y Combinator when it ran its first programme in Boston during the summer college break. Everyone followed suit, presuming there was a good reason for the 3-month period – there wasn’t but it became the model none-the-less. Its seen as a short sharp intervention to accelerate the startup to the next round of funding.

What killed accelerators?

If accelerators solved a problem then what changed, when did they stop solving a problem? Its two-fold:

Firstly, accelerators changed – they stopped supporting people with ideas and started wanting more mature startups. Startups now need traction and probably even revenue to get on a top-tier accelerator programme. You will find that most startups now entering TechStars are already revenue generating; its not unheard of to have TechStars startups who already have $1million in revenue. That is a far cry from where it all began. Why did they make this change? Competition – as more and more accelerators entered the market they had to back winners to raise their profile and to keep the investors backing their programmes. You are more guaranteed to get winners not by starting at the beginning but by acting as a finishing school.

This change in criteria means that fewer startups fit the requirements of the accelerator, some are to early in their development, and some are too late. This increases the competition even further because, although accelerators are getting 1000+ applications, 80% are not suitable. This leaves accelerators fighting for the same small number of startups.

Secondly, the market matured which has had the following effect:
1- The accessibility of information for startups has grown exponentially – its all on the Internet and its free, so why go to an accelerator for it?

2- There is an accelerator in every major city so startups don’t need to re-locate half way across the world to get support and funding

3- Startup valuations have risen, making the average valuation by an accelerator look low for the best startups

4- Timing – many startups want to get cracking and don’t want to wait for an accelerators next round, so they either skip accelerators or go directly to one that is starting now.

5- Founders span a wider age group now. In the past it was more likely to be recent graduates but now we are seeing founders in their early 20’s through to those in their 50’s. These are less likely to re-locate because of commitments (mortgages, children, etc).

6- The availability of capital for early stage startups has significantly increased, particularly in the UK where tax incentives have played a key part. This capital is available for startups at a fairly early stage, but probably not as early as accelerators delivered when they first started.

What will happen?

For lots of accelerators they will die a slow painful death – they will find it harder to attract the quality of startups they need and the knock on effect will be worse results, making it difficult to find their next round of investors & startups. Slowly over time they will just find it impossible to run their next programme so they will silently slip away.

The top tier will increasingly morph into something that resembles a VC. YCombinator & SeedCamp have already made that transition by raising funds and supporting their startups through much more of the journey than just 3 months.

Some accelerators (StartupBootcamp) are trying to go for volume of startups in the hope that they will find a few winners across the total portfolio. Whilst that may work, the ratios will look poor and most analyses would suggest that this is the wrong model.

The rise of the corporates

We will see more and more corporates enter the market as late adopters. They don’t have the same requirement for economic return and therefore can have a wider criteria of acceptance.

The problem hasn’t gone away!

The problem that accelerators set out to resolve hasn’t gone away. Startups still need that really early investment to create an minimum viable product (MVP) now more than ever. Today we talk about bootstrapping a startup through to MVP and early revenue; in years gone by we would have been saying ‘apply to an accelerator’. Its no longer economically viable for an accelerator to operate in the super early space unless they are corporate or government-backed – this is a market failure.

Paid email newsletter subscriptions

Email newsletters have been around since the internet began and in most cases they have been provided for free. However, increasingly people want to monetise their email content and there are a few way to do this.

Here are some technical services to help you start creating email subscriptions: – mixes stripe & mailchimp to create a solution $20 per month

A DIY solution using Zappier, Fancy Hands, Mailchimp & Paypal to keep the costs down

Campaignzee – uses stripe & mailchimp but costs you 10% commission on each subscription


Evernote session times

Sessions UP, Time DOWN

I was listening to the HBR Idealist podcast where they interviewed Phil Libin of Evernote about new ways we will work. In the interview he revealed some interesting stats that I think is great insight for all startups who are developing multi platform aps (web, mobile, wearables).

Its probably fairly obvious to say the more mobile the device, the smaller the form the shorter the session time (time spent using the app), your unlikely to read a 50 page document on your Apple watch but you might glance at a txt message. What was interesting but not something I had really thought about was that as the device shrinks the session time goes down but the number of sessions goes up. So when you are designing for mobile or designing for a wearable you are really not designing for a smaller screen you are designing for less time. Phil Libin used the term designing snackable content which I think explains the user behaviour well. This means if you have a web app today you are deciding what elements are important for the experience on a wearable not how do we cram what we have into the small screen space.

Evernote figures show the difference between platform session times and session numbers and its very different. How do you build a product that is used 200 time a day for just 5 seconds per session, that’s a very different product to one that is used for 40 minutes once per day.

Evernote session times

It’s certainly something I had not really thought about but it shows a lot more thought and therefore cost is going to be required when you are developing across platforms as diverse as desktop, mobile and wearables.

Why Cost Per Acquisition (CPA) doesn’t matter at the start

What is cost per acquisition?
A quick explainer for those not familiar with the term. Cost per acquisition is the cost of an activity that generates traffic to your website or downloads of your app. If I buy a banner advert (who buys banner adds these days?) for $60 and 3 people click on that banner add per day and visit my website then the cost per acquisition is $60/3 = $20. My Cost Per Acquisition (CPA) is therefore $20 per customer/user.

You may have heard people say “do things that don’t scale when starting out”, this means doing things that are resource intensive that you know you couldn’t scale because it wouldn’t make financial sense.

So why do non scaling stuff in the early days?
You are seeding your startup trying to build momentum so you can afford to do things that will not scale because you know or believe their will be a tipping point where momentum takes over and you no longer need to put as much effort in to meet your targets. Think of it like pushing a giant rock down a hill, getting it going requires huge effort probably by a group of people pushing with all they have but once it starts to roll it requires little effort to keep it going.

In the early days of testing an idea you are acquiring customers/users to test your hypothesis and iterating the product based on the feedback from those customers/users. Therefore you are also going to be doing things that don’t scale to acquire those customers/users (social media, blog posts, content marketing etc) and if you are looking for quick tests then you are almost certainly going to have to do some paid acquisition. Given that you are in testing hypothesis mode and you are in no way optimising your customer acquisition strategy there is no need to measure your cost per acquisition. Of course it does start to give you some valuable data and some ball park figures but its far from accurate and shouldn’t be used to base your financial model around.

It is very possible to go from $10 cost per acquisition down to $0.20 after you learn what works and optimise your strategy. So to base your financial model on $10 when it could end up at $0.20 is going to make it way out, of course if the financial model works at $10 per customer then its going to kick arse at $0.20.

Therefore while you are in the early stages don’t worry about the cost worry about which methods are effective, spend your time tracking which blog post and which tweets were more effective rather than the relative cost. By effective I mean activity which drove traffic to your website or to downloaded your app.

Startup Signal to noise

It’s true that the cost of starting a tech startup has plummeted over the last ten years; now all you need is a couple of talented people, internet access and a laptop or two. Once upon a time you would need servers, offices, expensive hosting and bandwidth, and lots of time to write bespoke code from scratch. In years gone by you would build the product with every feature you can imagine before you launched. Minimum viable product wasn’t even on the radar.

So now it’s easy right? Anyone can start the next big tech startup, no problem. Well I think this brave new world has created a new barrier and that barrier hits you a bit further into your startup journey.

Getting started is relatively easy and the number of new startups being created across the world shows this. We have millions more startups than we did this time ten years ago, particularly in the case of tech startups.

What’s the problem then?

You can get started, build a minimum viable product, test the idea with early beta testers in quick time. The biggest barrier facing startups today is acquiring customers. There is so much noise that getting your message (signal) through is becoming increasingly more difficult and expensive. If you are a Business to Consumer (B2C) startup, then you need a budget of hundreds of thousands to get any sensible market penetration. The days where £20k from an accelerator would get you there are well and truly over.

The world of startups has developed mechanisms to help with this new barrier and it all loosely comes under the heading of ‘growth hacking’. Growth hacking techniques help you acquire customers cost effectively. The problem with growth hacking is at the moment it’s very labour intensive and slow, compared to paid acquisition of customers. Time is often not a luxury a startup has, and in the words of Ben Franklin, “Time is money”. That’s not to say paid acquisition is not a great bootstrap technique – because it is – and in most cases it’s the only option a startup has for customer acquisition.

How best to tackle this problem then? I think the collective wisdom in the industry (particularly out of Silicon Valley) is that you should build an amazing product and concentrate on that; the users/customers will follow. Whilst I don’t disagree with that wisdom, I think it lacks a few stages of detail on the path to ‘just build a great product’.

Building a great product is only achieved in conjunction and with feedback from customers, it is never done in isolation. With a B2C product that means talking and listening to thousands of customers to help build your great product. I believe you should be fanatical about getting customer feedback and your first hire or co-founder should be an experienced growth hacker. The difference between someone with experience in growth hacking vs a few co-founders giving it a go is quite remarkable.

After founder divorce the biggest killer of startups is not being able to acquire enough customers at the right price. Therefore, when starting out, think about how you will solve the customer acquisition problem in as much detail as possible. Don’t just say, “We will growth hack it” – that lack of detail is a sure fire path to failure.

Pivot or Persevere – the data points

The question for lots of startups is whether to Pivot or Persevere on your journey towards product market fit. A pivot is a change in strategy and comes in many different flavours, it could be a technology pivot (web to mobile), it could be a market pivot or more likely its a business model pivot. You thought your users would pay for x but all the data says they will not so you need to try something completely different.

Persevere means continue to iterate the product making small changes based on customer feedback. The difficult question to answer is always are users not paying for the service because its not quite right and a few more iterations will fix that or are they not paying for it because they never will, it doesn’t solve a problem that’s painful enough for them to part with money.

Pivoting for the first time is a little like admitting defeat because you sold the business on your initial plan and research, now it turns out it wasn’t correct and you are admitting it by changing direction. In reality no amount of planning or research can predict or replace customer engagement, what you now have is live data to base decisions on.

I have recently spoken to two startups who have both pivoted and asked them what took them so long to figure it out (one took 12-months the other 24-months) and what data finally made them realise they needed to make the pivot.

In both cases they were tracking all the metrics you could think of so it wasn’t a lack of data or the wrong data it was simply knowing which data really mattered. So in both cases the acquired user figures were rising month on month at healthy rates. The issue was around retention, the retention figures were not good in either case, they had the classic leaky bucket, they were putting new users into the top of the bucket but they were leaking out the sides before generating any revenue. Both startups could see this was the issue and both interpreted that data as meaning they were missing a feature or some of the existing features needed improving, while this might be a reasonable assumption it can lead you into an iteration loop. An interaction loop is where you just keep iterating and you are moving a metric but very slightly each time so you are encouraged and think a bit more iteration will keep this moving up. This can lead to months of iteration and months of lost time when what you should have done is pivot. The difficult thing is still knowing if you need to iterate or pivot.

The answer is that at the point you have a retention issue you should talk to customers again, work out what they want, what they are using and why. Dig deeper with customers as this data is the most valuable, get them in your office, go sit next to them, get a very deep understanding of their needs.

To show you a worked example of this problem I sat down with Jeremy Walker, CEO of Meducation ( who has just pivoted after 12-months fighting the iteration loop. Here is his breakdown:

We have the following per month:

New users per month = 1000
Returning users = 2000 (from previous signups)
Total active users = 3000

But the devil is in the detail because if you break the monthly down into day/week/month you get this:

Daily active = 200
Weekly active = 500
Monthly active = 3000

Which gives you a 1/5/15 ratio. So although the monthly looks like a good figure it is skewed by the 1000 new users per month, you really need to look at the daily & weekly to get an accurate picture. The 200/500/3000 tells you that the product is not that sticky you don’t have regular returning users (of course it does depend on the product type to know if this is really a problem) but as a general rule if people are using it regularly (daily/weekly) they will pay for it and its solving a problem.

Jeremy suggests, If we’d had 200/300/400 or something similar, we’d have been on much more solid ground, even though our actual numbers would have been lower. Because then it would have just been a scaling problem where as 1/5/15 as a ratio means it’s a retention problem. I’d aim for 1/2/3 as my minimum successful ratio now.

As a startup your focus should be on retention and not user/customer growth, work with your customer to solve a problem and once you are doing that your product becomes sticky. Once sticky stick your foot on the gas peddle and go hell for leather on growth.