Stop Flying Blind - A book by Michael Mace

14. How to segment the market for a new product

Last time I talked about the need to segment the market if you’re designing a new type of product. If you design a product to please everyone, chances are you’ll end up with inoffensive pablum that excites no one. That works pretty well in politics, where voters have only a couple of choices. But in new product design, where consumers can choose from an almost infinite range of new products, unexciting is usually deadly. So you should optimize the product to make one segment of customers deleriously happy, and not worry about the rest.
Unfortunately, segmenting the market for a new type of product is a lot harder than you might expect.

There’s a huge amount of accepted wisdom on how you’re supposed to use research to identify market segments. But most of it is designed to refine the segments in an existing market — for example, what’s the under-age-12 market for tennis shoes like? When you apply those same processes to defining the market for a new type of product, something nasty happens: you can’t find any segments. The reality is that market segments for a new category of product don’t exist until that product is delivered. Segments gradually coalesce from a feedback loop between the desires of customers and the products that companies offer to them.

The process is a little like the way that astronomers say the solar system was formed. You start with a big cloud of gas and dust. Small lumps and thick areas in the cloud slowly draw together under the influence of gravity. Wait long enough, and stars and planets will eventually emerge.

When you do research on potential new markets, you’re searching around in the cloud for thick spots. The evidence will be vague and contradictory, and you can easily miss it if you’re not careful. The trick is to look not for segments themselves, but for groups of people who share desires or other characteristics that you can mold into a new segment.

For example, there was no real market for sports utility vehicles in the United States until some clever folks at the auto companies called it into being. Early civilian jeeps were sold as farm tools, believe it or not. But after the disappearance of the station wagon, there was a need for cars with a lot of carrying capacity and with a less domesticated image than minivans. Virtually no car buyers would have thought to ask for an SUV, but when offered a car that could haul a lot of stuff and also had a buff image, people jumped all over it.


How to find the lumps

As I mentioned in Chapter Nine, a lot of research companies are happy to sell you ready-made market segmentation schemes that they have derived from demographic data. These segmentations are built around age, income, and other basic characteristics of the population, and usually split a country into about a dozen groups, each with around 6-12% of the population.

In some cases these segmentations can be useful, especially if you’re selling a product that shoots for very large generic markets, or is closely tied to age or income (TV shows and blockbuster movies come to mind). But for new product categories, especially in high tech, I’ve found that generic segmentations are close to useless. They’re backward-looking, telling you how people have behaved in the past, not what they’re going to do in the future. And the segments aren’t cut the right way.

For example, a generic segmentation will tell you that young people tend to be early adopters, something that’s often true but is also meaningless if you happen to be designing a product for older people. The best customers for the RIM Blackberry mobile e-mail device have been business professionals over age 40 working on Wall Street. A generic national segmentation wouldn’t have ever pointed you at that segment, because it’s too small to show up in a mass national survey. And most generic segmentations point new tech products not at over-40 adults, but at 20-something technology lovers. There are some companies that have tried to make e-mail devices for the under-20 crowd, most notably the Danger Hiptop. But they have achieved only a tiny fraction of RIM’s success.

So you need to do your own segmentation. But even then things are still tricky. Traditional research methodologies were designed in the consumer goods industry, in which it’s pretty easy to tell who your customers will be. If you’re making a new laundry soap, you talk to people who do laundry. The traditional research approach says you start with some very general focus groups with those laundry-users. Those groups help you think of some guesses — hypotheses, if you want to sound professional — about what the customers might want (“I know, let’s make it lemon scented!”). You then use quantitative studies to test those guesses and gather precise information on how the market will react (“women under 40 prefer citrus cents, but those over 40 prefer floral scents”). The process culminates when you have enough information that you can tell the product development team exactly what to create. They build prototype products (“Blammo, the zesty new laundry soap for a new generation”), you test them, and when you’re ready you launch the product.

That process falls apart rapidly when defining a market for a dramatically different product. The first problem is that in order to recruit people for the focus group, you have to know what your market is. If the market’s not yet defined, you are almost 100% certain to recruit the wrong people for the groups. The focus groups are the foundation of the whole research process, so if you start with the wrong people, it will invalidate everything else you do. As we say in the computer industry: Garbage in, garbage out.

There’s also a practical reason for not starting with focus groups. I’ve also found that it’s almost impossible to make an engineering team wait for you to complete the three-step research plan. Because of long product development lead times, they need to get started on their work very early. So they’ll come to the focus groups and start development based on whatever they happen to hear there. In most cases, they’ll lock onto whichever customer comments match their preconceptions, and ignore the rest. By the time you get the quantitative research done, they’ll be halfway finished building the product.

When you’re defining a new market I think you need to turn the traditional research process on its head. Your goal isn’t to gradually build up a fine understanding of the market, it’s to get a general idea of the opportunities as quickly as possible so the engineers can start work. Then you refine your understanding of the market as they refine their product.

Do the quantitative study first. That means the first step isn’t focus groups, it’s quantitative research to try to get a feel for the structure of the market. One of the best ways I’ve found to do this is with a feature survey. Start with a broad cross-section of the public. Give them a very general description of the type of product you’re working on (for example, a mobile phone with advanced features). Then give them a long list of features, and get them to rate each feature for its relative importance if they were buying that product. Again using the phone example, you could list things like long battery life, low weight, color screen, ability to play MP3 music files, and so on. You don’t have to list every possible feature, but make sure you’re covering all the possible categories — entertainment-related features, communication-related features, etc. Compiling this list is a great place to have the engineers help you brainstorm.

You should also capture demographic information on everyone in the study — income, education, job, age, etc.

When you get the data, your temptation will be to look up which features “scored the best.” Go ahead and look if you want, but afterwards you should ignore that information. What you want to look for is not high scores, but lumps in the cloud. By that I mean features that tend to cluster together for some groups of customers — for example, if someone favors long battery life, do they also favor color screens? There’s a type of research called discrete-choice modeling that’s great for this sort of work. A good research company can help you do this analysis. The clusters you identify are your potential market segments. For each of them, dig into their demographics and other information until you understand what makes those users different from the rest.

The segments will seem depressingly vague when you look at the data. The researchers may tell you that correlations between features are weak, or there may be only a few features that form a particular cluster. That’s fine. Remember, you’re looking for tendencies that you can turn into a segment, not full-blown pre-existing segments. If the market were fully formed, there would already be 12 companies selling to it. Even a major market like SUVs didn’t just materialize in one year, it developed gradually as car makers noticed an interest in more rugged transportation, and consumers responded to the first such cars targeted at that idea.

Next, do a quick and dirty product concept test. Once you’ve identified the clusters, the next step is to have the engineers create some product concepts for them. These do not have to be full prototypes. In fact, few high tech companies have time to test-market working prototype products these days. It’s enough to have a nice sketch, along with a one or two paragraph description of the product, and basic price, size, and weight estimates. This is where you want to get the most creative thinking from your product development people. You have to make sure the concepts are well enough described that people can understand them and picture how they’d be used. Then do a quantitative test of the concepts, contacting the segments you identified in step one (something you can do since you gathered good demographics on them), and seeing how they react to the product descriptions. Do they like the ideas? How interested are they in buying? How much would they pay? This is the time to gather as much information as you can on preferred buying channels, price points, and so on.

You should test all the product concepts on all the segments, even if you think that one product will really only appeal to only one group. So, for example, if you were testing phone ideas, you’d ask the communication-focused users to rate the entertainment-featured phone, even though you think they won’t like it. If you’re correct about the segments, the right people will want the right products. If you’re wrong about the segments, your results will be all over the map and you’ll need to rethink the market.

Once again, the correlations here may be frighteningly mild. For example, 60% of your target market might want the product designed for them, while 40% want something they weren’t supposed to want. That’s fine; people are always more complex in reality than they are in a segmentation model. As long as you have a little angle to work with, good marketing and product design can solidify it into a full-blown segment.

Focus groups come last. The concept test should give you the information you need to complete product development. The final step is to do focus groups with the target customers, when the product is nearly finished. You don’t do the groups to gather information; you do them to gather ammunition. Record video of the groups, and edit the video down to about ten minutes of the customers describing themselves and reacting enthusiastically to the product. If you’re in an established company, you can use the tape in-house to help explain your product and get people to support its launch. If you’re in a startup, you use the tape to help you raise money for the launch. And in either case, the tape helps you educate the press and analysts about your product.

Researchers are sometimes uncomfortable with using focus groups this way. They feel that research should always be a search for the objective truth, and that selectively editing the focus group findings is a sort of crime against nature. Don’t let them throw a guilt trip on you, baby. Focus groups aren’t statistically valid anyway. Besides, you’re not pursuing abstract truth, you are building a new market out of (almost) thin air. You need ammunition to bring that market to life, and the focus groups are your source of ammunition. Use them unashamedly.


To summarize, traditional research on a market works like this:

1. Focus groups to brainstorm. Do focus groups to get a feel for the customers, get some ideas about what they want, and create some hypotheses of user segments to test.

2. Quantitative research to test. Do a quantitative study to validate the hypotheses you formed in step one, and size the segments.

3. Product concepts to refine. Conduct product concept tests to validate the designs produced by your engineers.


The new market definition process works like this:

1. Quantitative research to find lumps. Do rapid quantitative research testing many feature possibilities. Analyze results to identify customer clusters (potential market segments).

2. Quick product idea test to latent segments. Conduct quantitative product concept tests, with very sketchy product descriptions, to validate the segments and give the engineers enough information to go to work.

3. Focus groups for marketing ammunition. At the end of the process, conduct focus groups with people reacting to nearly finished products, to collect video verification of the segments and help you prepare for launch.


Next time I’ll talk about a totally different type of market research — advertising proof studies.

7 Responses to “14. How to segment the market for a new product”

  1. 14. How to segment the market for a new product :: Newstack said on December 2nd, 2006 at 12:56 pm

    [...] Read more: here [...]

  2. Jose Antonio Morales said on February 25th, 2007 at 11:46 am

    Hello Michael.
    1. you could give a check on Yahoo Answers. That can be also a reference for some kind of market research. Microsoft have a service named Q&A.
    2. I’m a small business owner in Slovenia. We are a team of 4 full time employed persons and some 6 external members. We are growing in a rate of 40% per year (revenue) but we are still under the 300K EUR. We can’t make a market research or a very long process to test our new products or services. Our strategy is focused on partners (Microsoft, HP) and with their support we reach potential customers. That means that we adapt our activities to what our partners need and we generate models for increasing revenue after we perceive the needs of them. Like a sample: MSFT needs to get in touch with small business, then we create a portal for small business sponsored by them and we offer our consulting services to them co-finaced by MSFT budget. Living our 4th year we have new challenges and we found how the market perceive our company. Now we are ready to target better, even though we are still offering a very wide set of services. That meand that we never segmented the market for a new service or product (too expensive for us) but we waited for the market to recognize in what we are more useful for them. Doing that we discovered skills, values and social needs that match with our organizational culture. Now the doors of better business are opening because of our passion and love for what we want to achieve more than any other thing.
    I see that business oriented books are ignoring this kind of experiences from small companies… or very small companies.
    I wanted to share this with you. Maybe some good idea would come from it. For sure if you have time to answer the post I will have the chance to learn something new.

  3. Mike Mace said on February 25th, 2007 at 9:20 pm

    Hi, Jose.

    You make a good point. The smaller a company is, the less it can afford to spend on market research. If you’re planning a hardware product (which typically requires a big up-front investment in development and inventory), I think careful segmentation is essential. Same thing if you’re creating an online service that requires a large engineering and marketing investment.

    On the other hand, the approach used by a lot of tiny Web 2.0 companies is to just put the web app on the market and see who responds. There’s so little up-front investment required for a web app that you can just throw away your work if you get it wrong. But if you take this approach, you need to make sure you listen carefully to your users, and make any changes they want. This can be tricky, since the enthusiasts who talk the most may not represent your average customer.

    Just as the Web has made software development cheaper, it has also made research cheaper. I think most companies could benefit from doing at least a little research, even if they’re small. But you make a good point that trial and error has its place too.

    Thanks again for the interesting comment.

  4. anothr user said on April 2nd, 2007 at 6:27 am

    Anothr feed track -Stop Flying Blind

    One new subscriber from Anothr Alerts:

  5. Khurram Qureshi said on March 30th, 2008 at 6:51 am

    Amazing stuff! I really liked the articles they are really addictive and informative. Talking about cluster of information in a quant study, what if you dont find a cluster or if its not a featured base survery then how do you group toghter the study?


  6. Chuck said on November 5th, 2008 at 4:31 pm

    Michael – it has been a long time since the last post. Will you continue with your book, or should I stop coming back here to check every week? What you have written so far has been extremely valuable and I hope you continue.


  7. Mike Mace said on November 7th, 2008 at 6:39 pm

    Thanks for asking, Chuck. You should not be checking every week. I have a lot more material that I’d like to post to this blog, but this format hasn’t been working — while traffic to my other blog (Mobile Opportunity) took off, traffic to this one never ramped at all. Since I didn’t have much time available, I have been focusing on the blog that gets more traffic.

    But I do intend to come back to this one once I figure out how to make it reach a broader audience. Longtail managed to make it work, so I don’t intend to give up. The topic of the book is something that I care about a lot, but I want to make sure I’m reaching a reasonable audience.

    I am very open to suggestions on approaches I should use.

    As for how to track this blog, here’s what I suggest — if you use an RSS reader, subscribe to this blog. Then when I figure things out you’ll see the posts start back up (or I will post here saying where the content has moved, if I move it).

    Thanks again for your feedback.

Leave a Reply