A survey can break even when every question was worded well, at the stage where you count the answers. If you count answers that say the same thing separately, or merge them into a word nobody said, you end up with a winning answer that does not exist.
The source behind the booklet has a name for the count, tabulation, and a set sequence of steps for it. The booklet lays that sequence out step by step. Here we deal with the two mistakes that make a count worthless, and with a sign that something went wrong before you started counting.
The counting itself does not take long. Suppose an installer counts twenty answers on a single sheet of paper and has a clear picture within twenty minutes. That time pays off only if you count correctly.
The breakfast example
The source shows a survey in which 100 people were asked for their favourite breakfast food. The answers scattered: croissants, eggs, scrambled eggs, fried egg, and a few more. According to the booklet, this is the example that teaches the whole chapter.
Count each answer separately, and croissants win, with just under a quarter of the answers. The source says that without further categorising, a wide spread would have come out, and the leading answer, which the source calls the button, would have looked like croissants.
Three of those answers are the same thing, though. Eggs, scrambled eggs and fried egg are all eggs. Put together, eggs reach close to half of the answers, and croissants drop to second place.
The customers did not change between the two counts, and neither did the answers. All that changed was how the person counting read them.
In a small business it looks like this. When customers say personal service, you answer quickly and you are reachable, they are giving one answer in three phrasings. Counted separately, they leave no winner. Price, expensive and budget belong to one category too.
The booklet also gives the opposite example. Suppose a studio owner does not count at all, and simply reads the answers and gets a sense of them. In the end she remembers the odd answers and forgets what most people said.
The trap in the other direction
You can overdo it the other way. If you categorise too coarsely, then in this example you could arrive at a category such as carbohydrates, which collects nearly half the answers and wins.
The source rules that out sharply. Nobody answered carbohydrates, so it is not an answer you can build on. According to the source, that tabulation misleads, because it makes a category invented at counting time look like the answer that came back most.
One rule comes out of it, and it is easy to remember: “A category nobody said is not an answer.”
Suppose an owner files every answer under professionalism and gets a clear majority. Not one of his customers used that word.
Now suppose the same owner leaves the customers' words alone. The sentence I don't have to chase you turns up in many answers, and it goes into his advertising exactly as it was said.
The winning word has to be a word people said. If it was born with you at counting time, the result was born with you too.
When there is no clear winner
What happens when the top answer gets a low percentage? According to the source, that signals something wrong with the survey. An answer that reaches a quarter of the people will appeal to only a quarter of the public, so your potential market shrinks by three quarters. In a case like that, you may not have asked the right questions.
A scattered result usually means the question was not accurate, and needs rewording and asking again. It is easy to read it as a trait of the public instead, and that is the mistake.
Suppose an owner gets eight different answers from ten people. His question, what matters to you, was too broad for a shared answer. When he asks instead what is most irritating about the process today, six out of ten say the same thing.
Why the count matters so much
All the work before the count ends up in one winning answer: breaking the list into publics, wording the questions, and the conversations themselves. If the count is crooked, the message built on it is crooked too, because it rests on an answer the customers never gave. That is why the booklet gives this stage a whole chapter, and no mere line about how to count.
The booklet has, for this chapter, the steps of the count itself, one exercise in doing and one in checking, and self-check questions. The next chapter deals with what you do with the winning answer, and why you must not improve its wording.
The count is where it is easiest to change a result without noticing. The winning word has to be one your own customers said.