Do you ever wonder how the numbers in market estimates come about? And how can these numbers be so “precise”? For example, I just saw a market size estimate of USD 25.73 billion for the global shampoo market. Why not 21.92? Or 4.67? Or even 98.32? How could you check the plausibility of these numbers?
Enrico Fermi, a master estimator
A while ago, I read a book about the great physicist Enrico Fermi. The book is “The pope of physics”, by Gino Segrè and Bettina Hoerlin. Highly recommended reading.
Apparently, Fermi liked making back-of-the-envelope calculations. He was notoriously good at getting approximate answers to complex questions. In doing so, he only needed very little data. Here are two examples of such questions:
- “How many piano tuners are there in Chicago?”
- “What is the mass of all the automobiles scrapped in North America this month?”
These “Fermi questions” are very similar in kind to market estimate questions: in order to answer them precisely, you would have to put in massive efforts. Or… you could try to estimate, just like Fermi did. Let’s take a look at some examples, and see how one could do this for market estimates. By the way, we think that “fermiing” should be in the methods toolbox of any good innovation analyst.
How to Fermi yourself toward a reasonable market estimate
Here I use three market estimates from very different industries or tech fields to illustrate the method: robotic lawn mowers, shampoo, and fiber optics. Also, I will cheat a bit. I will use data from the web, mostly from Wikipedia. Just in general, Wikipedia is a fantastic resource for many tech and business discovery endeavors.
Related: How to use Wikipedia to boost your discovery skills.
In this article, we will use two estimation methods:
- Estimate based on number of units to be sold. If you know what a unit is and how much one unit costs, you know how many of the units would have to be sold in order to meet a certain market size.
- Estimate based on important players’ revenues. If it is too hard to specify a unit, you can check the revenues of important players in a field. This should give you a reasonable upper bound for a market size.
Let’s get started, and apply the number of units to be sold estimation method to robotic lawn mowers.
Global robotic lawn mower market estimate
I used Mergeflow’s market estimate extractor tool, which I already discussed in this blog in a different context. With the tool I found a recent estimate for the size of the global robotic lawn mower market. The report said that for 2020, this market would be USD 1.3 billion.
How can we get a reasonable approximation to see if this makes sense, and not spend more than 20 minutes on our estimation?
First, let’s define a “unit to be sold” and its price. Then, if we divide the market size by unit price, we get the number of units that would need to be sold. Here we are lucky, because the unit is one robotic lawn mower. Of course, this means that we exclude things like battery packs for the lawnmowers, maintenance, lawn mowing service companies, and probably some other things as well.
I checked Amazon, and one robotic lawn mower sells for ca. USD 600. Therefore,
Number of units to be sold = USD 1,300,000,000 / USD 600 = 2,166,667
So, for the estimated market size to work out, ca. 2 mio robotic lawn mowers would have to be sold globally in 2020. Is this possible?
Let’s focus on households for now. This means we need to address the following questions:
- How many households are there globally?
- How many of these households have a garden? (if you have a garden, you probably have lawn to mow)
- How many households with gardens (and lawns) use a robotic lawn mower, as opposed to manual or non-robotic lawn mowers?
How many households are there globally?
In order to estimate the number of households worldwide, we can use Wikipedia. Fantastically, Wikipedia has a list of countries by number of households.
This table is an extremely useful resource. You can use it for doing plausibility checks of all kinds of market estimates where households play a role, not just for robotic lawn mowers.
How to import Wikipedia data into Google Sheets
If you use Google Sheets, you can import the data from this Wikipedia table into your sheet, like this:
=IMPORTHTML( "https://en.wikipedia.org/wiki/List_of_countries_by_number_of_households", "table", 2)
The result looks like this (click on the image to see a larger version):
I learned this from a blog article by Justin Pot at Zapier, “6 Google Sheets functions that do more than math”. Thank you, Justin!
If you want more details on this, and see how you can import data into Excel, please see the following article in our technical knowledge base:
How to get Wikipedia data that helps you estimate market sizes
Market size plausibility check, based on number of units to be sold
From the “number of households” table, we learn that…
Number of households worldwide = 1,600,000,000
How many households have a garden?
I did not find data on how many households have a garden. So we have to estimate. How about 5-10%? Let’s say 5%. If 5% of all households worldwide have a garden, then there are…
Number of worldwide households with a garden = 5% x 1,600,000,000 = 80,000,000
How many households with a garden and lawn might use a robotic lawn mower?
Again, we have to estimate. How about 10%? Assuming this, we get…
Number of households with a garden that might use a robot lawn mower = 10% x 80,000,000 = 8,000,000
So, based on the assumptions we’ve made, 8,000,000 households worldwide might use a robotic lawn mower. Above, we estimated that the number of units to be sold in 2020 is ca. 2,000,000, if the market size estimate of USD 1.3 billion were to work out. So does it work out?
The data and estimates we have suggest that of 8,000,000 households using a robotic lawn mower, 2,000,000, or 25%, would have to buy one in 2020. This seems a bit high to me. On the other hand, if we were to assume that not 10% but 15% of lawn-having households might use a robotic lawn mower, we’d have 12,000,000 households. Then, the 2,000,000 new lawn mowers would be 17%, not 25%. This seems a bit more attainable.
We can now play with our data a bit more. For example, we could stick to the 8,000,000 households that use robotic lawn mowers. Then we could assume that many of these households already have a lawn mower, and that only 10% buy a new one in 2020. This would be 800,000 households. And 800,000 x USD 600 (our unit price) = USD 480,000,000. So, perhaps a very conservative lower bound for our market estimate could be USD 480 million.
But let’s leave robot lawn mowers now, and move on to a different topic. Shampoo.
Global shampoo market estimate
I already mentioned this in the introduction: Recently I saw a market estimate for the global shampoo market. This estimate put the market size at USD 25.73 billion in 2019.
The plausibility check for the shampoo market estimate
Shampoo and robotic lawn mowers are really different, of course. But we can use more or less the same method for our plausibility check of the market estimate.
Unit to be sold this time is a shampoo bottle. Unit price for a 250ml shampoo bottle is somewhere around USD 3 probably. Yes, I know, there is much more expensive shampoo as well. But let’s assume that we are talking about the mass market version of shampoo here. So, we have…
Number of units to be sold = 25,730,000,000 / USD 3 = 8,576,666,667
Let’s approximate this to 8,500,000,000. This is the number of shampoo bottles that would have to be sold in order to reach the market size of USD 25.73 billion.
OK, next question: How many 250ml shampoo bottles does one household use per year?
Of course, this depends on many things. Household size, how often each household member washes their hair, how much shampoo they are using per wash, etc.. So we could research ourselves into an abyss, or we could just estimate. Let’s estimate, and say that one 250ml shampoo bottle lasts two months in a household. So 6 shampoo bottles per year per household. Based on this assumption, and the household data table from Wikipedia (see above), we get…
1,600,000,000 households x 6 shampoo bottles = 9,600,000,000 bottles of shampoo per year
Then we pluck in our unit price, USD 3, in order to get the market price (which is number of units sold x unit price):
9,600,000,000 x USD 3 = USD 28,800,000,000
I’d say that this is remarkably close to the USD 25.73 billion size from the published market estimate.
Does this mean that our estimate, or the published estimate, is “correct”?
Well, I would be more careful. I would say that we have built a scenario in which our estimate and the published estimate are a plausible outcome.
How the shampoo market estimate is probably easier to do than the lawn mower market estimate
There is one thing that makes the shampoo market estimate a lot easier than the lawn mower estimate: with shampoos, we can reasonably estimate how long one bottle lasts. With robotic lawn mowers, this was a lot more difficult. How often do people buy new robotic lawn mowers? Do they go to other lawn mowing methods when their robot is broken? So I think that the margin of error for the lawn mower estimate is probably higher than for the shampoo estimate.
OK, now that we’ve mowed the lawn and got all nice and clean, let’s move on to our last market estimate example, fiber optics.
Global fiber optics market estimate
Another market estimate I discovered via Mergeflow’s market extractor tool put the global fiber optics market size at USD 4.48 billion in 2019.
Now, this is a tough one. What’s the unit here? With lawn mowers and shampoo, this was a lot easier. But here… one fiber optic cable? One fiber? Let’s say we could agree on one unit, and a price for it. But even then, how could we possibly find out how many of them are needed every year? And what for? Because they are used in so many applications, from transcontinental data cables to internal “wirings” of devices.
So let’s take a different route for our plausibility check instead:
- Find the most important fiber-optics-making companies.
- For each company, get their annual revenue.
- Estimate the percentage of their revenue that’s from fiber optics. This is important if the company list includes companies that make many other things as well.
- Sum up the weighted revenues (= the fiber optics revenues) estimate. Use this as a reasonable upper bound for the market size.
The most important fiber-optics-making companies
Finding a list of important companies is relatively easy for most tech fields. This is true for fiber optics too. A web search for “fiber optics companies”, for example, produces a number of such lists. Here is a list that I found within 5 minutes of searching:
- Adtell Integration
- ADVA Optical Networking
- Ciena Corporation
- Cisco Systems
- Finisar Corporation
- Furukawa Electric
- Hamamatsu Photonics
- Huawei Technologies
Get revenues for companies
Again, this is easy and done relatively quickly. Within 10 minutes, just web searching “COMPANY_NAME revenue” got me this information.
Estimate the revenue percentage that’s from fiber optics
This is a bit more tricky. Here is how I did it: I made the following buckets, and assigned percentages to them:
- Company is all about fiber optics: 90% (not 100% because I think no company does “just this one thing”).
- Fiber optics is an absolutely essential part of the business: 50%.
- Fiber optics is an important business unit: 10%
- A company does many, many other things, and fiber optics are just “one more thing”: 1%
Sum up the weighted revenues
OK, here is the table I made, with all the data and estimates:
All in all, it took me about 20 minutes to make this “plausibility check table”. So we will be slightly over our time budget of 20 minutes per market estimate.
Based on my data and estimates, I would probably be surprised if I saw a market estimate that is orders of magnitude off. For example, USD 100 billion would surprise me. But the USD 4.48 billion I mentioned above probably seem reasonable.
But you are just making assumptions!
Yes. I agree. But now the methods and the assumptions are out in the open. It’s easy now to put them into a spreadsheet, and see what happens if I change one of my assumptions. Also, now I can discuss any plausibility check I did above with others. And to me, this is one of the greatest benefits of the Fermi method. It is a lot less about being right than it is about triggering useful discussions.