Surveys

Four of six LunchLoop interviewees said their plans change in the morning, after the 9 p.m. cutoff. Is that a quirk of four people, or the story of half the customer base? Interviews cannot tell you; they are too few. A survey can, as long as it is built to count rather than to explain.

The rule

Interviews find the why; surveys count the how many.

A survey works best after qualitative research, to measure a pattern you already understand. Good survey questions follow a few rules:

• One idea per question. "Was delivery fast and the food good?" asks two things, so a single answer cannot be read.
• Neutral wording. No "Don't you agree…" and no loaded words.
• Options that cover every case and never overlap. "0, 1–2, 3–5, 6 or more" works. "1–2, 2–5, 5+" does not: where does 2 go, and where does zero go?
• Recent, concrete behaviour. "In the last 4 weeks, how many tiffins did you not eat?" beats "Do you often waste food?"
• Short. Every extra question costs you finished responses.
• The right people. Asking only active subscribers leaves out the ones who already cancelled, perhaps over this very problem.

Here is the counting question LunchLoop used:

In the last 4 weeks, how many of your LunchLoop tiffins did you not eat? ○ None ○ 1–2 ○ 3–5 ○ 6 or more ○ Not sure

The shortcut

Before sending, point at each question and ask: "Which decision changes depending on the answer?" If you cannot name one, cut the question.

A survey tells you 70% of people found out in the morning. It does not tell you why their mornings change. That is what the interviews were for; the two methods work as a pair.

Drills

Answer fast — trust the rule.

Remember

• Interviews find the why; surveys count the how many.
• Shortcut: for every question, name the decision its answer changes, or cut it.
• Trap: double questions, overlapping options, and surveying only the people who stayed.