Estimating market size (part 3)

Top Down, Bottom Up

This week, we're taking a look at a few different ways to estimate the market size for a web 2.0 start-up.

Although you don't want to say you can sell your product to everyone (that would just be too easy), you can start there and then break the market into smaller segments to determine the number of potential buyers. So how many of those people are out there?

One way to look at Swivel Business's potential market is to estimate how many people have Microsoft Office and are currently employed.  Now, let's break down a broad market, such as geographic area, into smaller segments:

• How many people are in the US? The US Census estimates 304,059,724 people.
• How many of those people were employed in 2008? A look at some data from the Bureau of Labor Statistics yielded the number 137,248,000 for non-farm workers.
• How many people in the US use Microsoft Office? A few blogs cited 550 million people use Office worldwide, which means about 249 million people in the US have the software.
• We can estimate that about 112 million employed people in the US have Microsoft Office. 
• Now let's hit our target market. If 1 in every 50 of those people use Excel in their jobs, then that's a market size of 2.24 million.

After you reach your target market segment, you can do some sensitivity analysis. What if every 1 in 20 used Excel at work? 1 in 10?

The next step is to look at a different angle and build the market size from the bottom up. For example, I used data from the BLS Occupational Outlook Handbook to estimate how many people had occupations where using data and creating reports was in the job description.

I still have a lot to learn about marketing, but I thought I would share my strategies with you. I'd love to hear marketing tips and stories from you too!

Comments

[Miramon]

This is well and good, but where did you get the idea that 1 in 50 people use Excel? Unless that's reasonably good data, your approach is all but meaningless.

Sensitivity analysis with conservative or optimistic numbers based on the 1/50 basis may look good on some chart or presentation, but since all those values are equally baseless, you still haven't really gotten anywhere. Even a delphi poll requires some degree of knowledge and expertise from the participants; just pulling the key number in this process out of thin air doesn't do any good.

Of course if you have good numbers that 1/50 really do use Excel then the market segmentation makes some sense, but from your discussion, it appears that that is just an imaginary value.

[Natalie Glatzel]

The 1 in 50 number is an estimate used to find a range of the possible market size. I used the methods found at the following sources to complete the exercise. Both suggest using estimates and logic to narrow down a range.

http://www.marketingprofs.com/Tutorials/marketsize1.asp

www2.latech.edu/~jpratt/IVR/Market%20Size%20Training%20Module.ppt

If you have a different method, I'd love to hear it. There's always more to learn. :)

[Miramon]

I don't have any inherently superior suggestion, unfortunately. I am actually in the same boat as an analyst for an even fuzzier and worse defined market, and it's horrible to try to do projections under such conditions.

My point is that when you are multiplying a series of factors together, your confidence in the result is really no greater than the lowest confidence of any of the factors -- in fact it's worse than that, but just for simplification's sake, let's assume all your other numbers are precise and totally accurate.

Since (I gather) we have very little confidence in the 1/50 ratio, and since we no doubt would have to impose a very high variance on that figure to bound it with any confidence, that means all the other figures, with their relatively higher confidence levels, are all but irrelevant, and the final market size estimate is very weak.

Imagining myself as a VC looking at a business plan based on this formula (which I'm not) I would have to say, well, you may have a great idea and great code and all that, but this analysis just isn't very helpful.

[Miramon]

By the way, as an aside, I think that writing down a figure like 2.24 million is very misleading in this situation, when the error is so huge. As you probably know, when someone sees decimal places, unless told otherwise, they will have the reasonable expectation that any error is associated with the final digit -- otherwise there would be no validity to the decimals. So this figure implies a great deal of precision and certainty about a market estimate which probably doesn't really exist.

As an analyst, when I see clearly implausible decimal places, I think of Disraeli's supposed remark about lies, damn lies, and statistics, and so I am immediately biased against the source of the figures, which may be unfair, but could be avoided with a little circumspection.

Since in fact there is hardly a strong basis for the first digit, much less the other two, I think it would be much better to write down something like "$2-3 million" or whatever the plausible range of values may be within some reasonable expectation of accuracy.

[Natalie Glatzel]

I completely agree with what you've posted in your comments. Perhaps, I didn't emphasize my goal of this exercise enough in the blog post. Rather than attempting to find a specific target market, we were trying to find a broader possible (emphasis on possible) market size.

The part posted here is only one step of the full exercise, which looked at the market from several different angles. Our final estimate is a range of 0.5 to 5 million people who could possible use the product, which is a pretty wide range. And as you say, they are not figures set in stone, but what I believe to be a reasonable estimate.

[John]

Which could be the steps to estimate the size of a potential market ? We are thinking in a mix: qualitative and quantitative methodology.

===========================

John

New Cars

[Natalie Glatzel]

Hi John,

I agree with the compromise you mention between quantitative and qualitative analysis. The sources I consulted (links above) suggest using a combination of logic and hard data to estimate the market size.