Let’s consider a common situation where you want to report on a numeric finding that has at least two components. For example, you want to report a benchmark for the ratio of practicing lawyers of a law department or law firm to all other staff. On your survey questionnaire, should you ask each element of the ratio separately and do the math yourself, or should you ask the respondent to provide the ratio directly after you give them clear instructions for how to do the calculation? Even that hypothetical contains a second example of the decision: you could further break down a series of questions about numbers of staff – paralegals, secretarial assistants, administrative managers and others. In the analysis phase an analyst could add those components to arrive at “staff”. On the one hand you want to make answering easy for respondents, but on the other, you may eventually want a calculated ratio using two or more of the components.

I favor asking for the components separately. Here are my reasons to ask for them:

• ease of response for the person taking the survey – they can simply plug in the numbers rather than struggle with any mathematics (no matter how elementary the math some people will make mistakes or omit the answer);

• once you have the components you can be more accurate; other than by spotting outlier values, you have no way of knowing whether they divided the right number by the right number or added correctly;

• you can add parameter limitations to the response block and thereby more easily check the facial validity of the values entered. Suppose you want the ratio of lawyers per office for law firms with fewer than 50 lawyers. The questionnaire would ask for the number of office locations, with a parameterized maximum of 10 offices (assume no firm will have more than that number) and a minimum of one;

• you can refine and improve any of the component questions independently, a step you can’t take as well if you ask for a calculated answer; and

• you may want to use the components for other purposes, but you can’t salvage the eggs from a ratio or summed omelet. You might make use of the individual values in a way that you did not anticipate when you launched the questionnaire. For instance, asking for equity partners and for all other partners enables more nuanced analyses later.

What you don’t want to do is stump respondents who might know a total or a ratio, but not its constituent parts. Think of asking for a law firm’s effective billing rate as compared to asking for the component figures that go into calculating that number. The latter request will thwart many respondents. Or think of asking for the total number of practicing lawyers as compared to asking for practicing lawyers by level or by years since graduation from law school. The second approach, asking for a breakdown, requires research – and research increases guesses, drop out, and skipped answers.

The piece-wise approach does add one or more questions to the survey, but respondents generally don’t mind simple questions as they can breeze through them. The analyst also has to code the calculations that are needed thereafter.

If you choose to ask for a ratio or a calculation, explain what you are seeking, give step-by-step instructions and include a worked example. If you ask for revenue per lawyer, you might write “Take the firm’s revenue in the last fiscal year and divide that figure by the number of practicing lawyers on Dec. 31st of that year, e.g., $50,000,000 revenue divided by 100 practicing lawyers equals $500,000 in revenue for each lawyer.”