We work on a lot of employee communication surveys here at Davis & Company, so we’re sensitive to the nuances of effective survey design.

And the fact is, creating a survey requires precision and discipline. Every detail matters—from how you word a question, to how the survey is organized, to what scale you use for closed-ended questions. Each decision you make affects how employees respond.

Which leads me to today’s example: A rating scale so flawed that it encourages participants not to respond positively.

I’ll explain. In a recent blog, I mentioned that after I take my car for service, the manufacturer sends me an electronic survey. At the time, I complained that the survey’s 10-point scale was weighted toward the positive.

The scale had 1 on the left and 10 on the right. Above it, not exactly aligned to the numbers, were the descriptive terms: Unacceptable (roughly above numbers 1 to 2), Average (3 to 5), Outstanding (6 to 8) and Exceptional (9 and 10). If you consider average to be neutral to positive, the numbers 5 through 10 are all affirmative. That meant the carmaker was getting too many high scores.

But when I took the latest survey I noticed the scale had been changed. The choices are all the same except for one: Instead of Exceptional (9 and 10), the heading now reads Truly Exceptional.

Think about that for a moment. What have you experienced in any aspect of your life that was “truly exceptional”? A five-star hotel in Paris, perhaps, or a fabulous Caribbean resort. That romantic dinner where you proposed (or your sweetie proposed to you). The time you visited a natural wonder (The Grand Canyon, perhaps, or The Great Barrier Reef) and were overwhelmed with how beautiful it was.

This feeling—Wow!—is unlikely to occur while getting your car serviced. So the effect Truly Exceptional had on me as a survey respondent was to push me further down the scale. I had a pleasant service experience, yet I didn’t rate any aspect more than an 8 (which, as you recall, is Outstanding—pretty darn good, by any measure).

As a result, even my average score was probably 7.5, it’s likely that the manufacturer is going to be disappointed. But the problem isn’t the service—it’s the scale. The lesson, of course, is that any detail can skew a survey. In survey design, precision counts.