Читать книгу Validating Product Ideas - Tomer Sharon - Страница 25

Begin analysis when you’ve collected about five percent of the expected number of responses. This will help you fine-tune the rest of the analysis. Look at each of the answers you collected, and one by one, classify them into the categories you have predefined. If you work in a team, do the first chunk together. This way you’ll understand better how to classify answers in a consistent manner. For example, here’s an answer you might get:

A slow cashier combined with an elderly person who was in front of me in the line caused me to be late to pick up my son from school.

This answer could be classified as follows:

1 Location: at the store

2 Close people: N/A

3 People at the store: other customers, cashier

4 Issue: long lines

Get the idea? This way, you disassemble each and every answer you collected into components you can count later.

As you make progress with classification, you will realize that some categories need to change, split, merge, or be removed, and that new ones should be added. For example, you might find that under Location, it doesn’t make sense to have both in car and way to store, because they are redundant. Or you might find there are many answers that would benefit from creating a separate Cost category with values such as item, expensive, cheap, compare to other store, and so on.

Eliminate answers that are just incomprehensible or irrelevant. Don’t assume you understand what they mean. Only refer to the text in the answer. If there’s doubt, there’s no doubt. For example, if your question was “What frustrated you the last time you went grocery shopping?” and you get the answer “Dogs,” don’t guess what the reason was or assume any type of category for classification. This is an incomprehensible answer you should ignore.

If you classify data as a team, work together while you are all in the same room so that you can discuss things as you make progress. If this is not possible, catch up with team members once in a couple of hours (or at least daily) to do the following:

• Spot-test how team members are classifying so that you ensure the whole team is being consistent. (It’s a pain to undo later with thousands of data points.)

• Highlight and discuss any data that is hard to classify with existing categories.

• Create new categories and then share them with the whole team as soon as possible. If you don’t do this, then team members start to invent their own inconsistent categories or just use existing categories that are inappropriate, which is harder to consolidate later.

Work on one spreadsheet when you classify. This spreadsheet should include all individual answers in one column and categories in columns to the right. This process allows you to filter easily and sort by category later. Sorting by a particular category makes it super easy to eyeball the data for that issue and understand what happened. Set up the spreadsheet to automatically tally up frequency counts as you analyze (see Figure 1.3). The resource page for experience sampling on the book’s companion website at leanresearch.co includes a template spreadsheet for you to use during the analysis step.

FIGURE 1.3 An experience sampling spreadsheet set for tallying up categories.