Fermi estimation for startup business models

cartoon6548

Early in a company’s life, you don’t know anything. Often your best estimate of any metric or market behavior or business model component is at best accurate within a power of ten, for example “expected conversion rate between 0.5% and 5%” or “cost to acquire a customer between $50 and $500” or “average monthly revenue per customer between $20 and $200.”

Estimating with these extremely wide ranges can be surprisingly useful. In physics it’s called “Fermi Estimation.” It’s useful because it’s easy to get to a rough answer to important questions, but also ensures you don’t erroneously ascribe too much precision to the result.

Let’s run an example, taken from real numbers from a startup I recently spoke with.

Question: How big does this company’s target market need to be?

Suppose the goal of this particular company is to achieve $1m in annual revenue for a two-person bootstrapped startup, with a goal to be so efficient as to never have to hire employees, and therefore produce a terrific little business that nets each of them a few million dollars over the next ten years.

Their product has a few usage tiers; on average a customer spends $50/mo. So they’ll need 1666 customers to achieve their revenue target. (We might as well use exact math when we’re not forced to estimate!) Including future cancellations, they’ll need to sign up a total of 2000 customers to net 1666.

As I’ve written before, an excellent rule of thumb is a 1% conversion rate of qualified traffic to signups, so they’ll need 200,000 unique, qualified visits. Click-through rates on paid advertisement is 1% at best[ref]For less-contested keywords, yes you can get into 2% or higher, but not typically at reasonable scale. By the time you average the better conversions with the worse, 1% is a good rule of thumb[/ref], so we’ll assume less; call it 0.3%, so they’ll need 60,000,000 impressions on qualified eyeballs if they use that sort of advertising.

That’s a lot of impressions! Not many search terms have that many impressions even over 3 years. So already our estimation is showing that standard advertising might not be a plausible strategy.

Of course not all people will find out about the product from advertisement! But if not that, then SEO or blog posts or tweets or word of mouth and so on, so it’s still going to take 10s of millions of impressions.

If there’s a grand total of 1m total potential users, there won’t be tens of millions of impressions. That quantity of people simply won’t generate enough activity.

If this company is going to have a shot, some of these numbers need to change, and not incrementally through A/B testing, but by powers of ten. Fermi-style.

Not all these numbers can move by a power of ten. You’re not going to get a 10% conversion rate from Adwords, for example.

Freemium is one way to increase conversions to free-customers by 10x, but typical conversion from “free” to “paid” is 1-5%, so actually that can be even worse.

So what can be shifted? Maybe between higher MRR, more targeted marketing, and stronger sales, you can whittle down the required number of customers and impressions by 10x. All that is quite difficult, however, and despite the Common Knowledge of the Internet, that’s not usually how a company succeeds.

So often the best answer is simply “more potential customers.” A product that addresses 100m people has a shot at creating 60m impressions.

This is why even a bootstrapped company with no desire to “scale” still benefits from being in a large and growing market. Bigger markets make things easier.

Another answer is to change the go-to-market strategy. For example, perhaps in this market there are enough affiliate-marketers that they become a way to get there, because they’re a pay-per-performance sales force rather than spray-and-pray advertisement. If there are 10 affiliate-marketers who you can get to shill for you, and each can cause 10 customers to sign up per month, this company could get to their goal. (Hint: It’s actually rare to find a niche where this is true, as small and simple as this sounds.)

Yet another answer might be developing a digital following through social media, email lists, blog posts, eBooks, and so on. Clearly this is a ton of work, and most companies who try to become “thought leaders” end up spending a lot of time and not getting much in the way of hard results. The break-out successes, however, show that it can work out (e.g. HubSpot and Buffer, although note that both are in the space of social media marketing; it might be harder to get lots of shares and Twitter chatter if the vertical is manufacturing).

In any case, it’s clear that you need use creative non-incrementalism, and focus on the one thing that is most likely to move a needle by 10x instead of 2x.

Fermi estimation is a simple way to make sure a company at least has a shot at success, or to see whether one strategy might be 10x more effective than another, and thus might be the better place to start. Try it yourself and let us know how it goes in the comments.

1,837 responses to “Fermi estimation for startup business models”

  1. Good post. So much of early stage planning is setting expectations with prospective and current partners. Fermi is especially supportive when setting expectations with investors: 9 times out of 10 you’ll meet/ exceed expectation when it’s based on a range, that at worst, is an order of magnitude higher than optimal.

  2. Very helpful. Many founders are either paralysed by a search for precision or wildly over optimistic when estimating sales. This offers a simple approach to get off the ground.

  3. Agreed. Now that we are entering year 4. We have found the levers that we need to pull to maintain 1mil recurring rev. Most things are getting harder, and we need to keep getting better at each thing to still get a good return.

    This will really help my team and I to ask each day, month, etc: out of all the things I could focus on – what is the one thing that has the best chance of being 10x better than the other things.

    Then, using the 80/20 rule to temper our energy on the other end. So I do one thing – hard – until we get 80% value. Then pick another 10x thing. Rinse, wash, repeat.

    It sounds so simple….. ;)

  4. I love the famous story of how Fermi used his estimation technique to estimate the yield of the first atomic weapon.

    He dropped some pieces of paper, looked at how far they were blown by the blast wind from the bomb and used a ‘back of the envelope’ calculation to work back from there to the bomb yield. He was within a factor of 2. That is what I call a physicist.
    http://www.dannen.com/decision/fermi.html

    • It’s no longer a startup, but I work for moraware.com. It reached a million before having 4 total people on the team. Now we have 6 full-time people and more than 2 million annual revenue.

      Some things to note about it so it can be emulated:
      1. Really tight niche – not “manufacturing” but “countertop fabricators”

      2. Our customers have clear pains that they’re willing to spend money to overcome. Our lowest price offering is $50/month and $300/month is very common.

      3. SaaS is awesome – but it took 5 years or so before the founders started making real money. Now it’s great.

      I love my company – it’s the most successful “lifestyle” business I’ve ever come across. I wish more people would strive for businesses like this instead of trying for the next facebook.

  5. I just found this article in my Feedly reading list and it comes at the perfect time. I was just working on our plan for this year and struggling to make kinda precise predictions.
    Using Fermi estimation seems like the best bet for us right now. It doesn’t make much sense for us to make precise plans at this point as we don’t have enough data to back up our estimations. This is just perfect!

    Knowing we’re also in the middle of a second round of funding this type of estimation process could become really useful :)

  6. Thank you for your post. In my opinion, to start a business, we must have MONEY, money can control anything, although we have the best idea. If we do not have must money, we can start a small business by building a small website buying something. Ex, a website buying shoes, http://topwalkingshoesformen.com Then we have big total, we start a bigger business, and more…

  7. Best Chandigarh Escorts is a very popular Chandigarh independent escorts which provides the at most excellent Escorts in Chandigarh, beautiful call girls, call girls services in Chandigarh for your fellow feeling at any occasion.Setup a Romantic date with one of our attractive Chandigarh escorts services now. http://jannatmodels.com/

  8. Ms Aashi Rana and setup a sexy date with one of our beautiful Goa escorts now. Search some of our glamour girls here and then just give us a call to set up. Escorts in Goa – Goa Escorts Service – Goa Independent Escorts – Goa Call Girls and Goa Escort. http://aashirana.in/

  9. This is marvelous blog. your blog is exceptionally intriguing a debt of gratitude is in order for sharing this valueable blog Delhi is a standout amongst the most happening place in India for the sake of entertainment, happiness and administrations, administrations like which gives you delight a considerable measure of joy You will get all your blessing from heaven while messing around with our Escorts in Delhi.
    Delhi Escorts

  10. Such a nice and Superb article, we have been looking for this information and here we are really like to thank for your yours superb really its much helpful and appreciate your work. Thanks & Regards Painter in dubai

Sign up to receive 1-2 articles per month: