Top-box score

Top-box score is the percentage of respondents who chose the highest, most favorable option on a rating scale, such as "5" on a 1–5 scale or "very satisfied". Top-2-box adds the second-highest option ("4 + 5"). Both turn a rating question into one easy-to-read share, so teams use them to report satisfaction, purchase intent or agreement and to compare stores, products or periods.

Formula

Top-box score = respondents who chose the top option ÷ all respondents who answered the question × 100%

Top-2-box score = respondents who chose either of the two top options ÷ all respondents who answered × 100%

The base is everyone who answered that question. Decide in advance whether "Not applicable" or "Don't know" answers are excluded from the base, and keep that rule in every wave.

Worked example

200 customers rate their delivery on a 1–5 scale:

• Top-box = 84 ÷ 200 = 42%
• Top-2-box = (84 + 66) ÷ 200 = 75%
• Bottom-box (1) = 5%, bottom-2-box (1 + 2) = 12%
• Mean = (5×84 + 4×66 + 3×26 + 2×14 + 1×10) ÷ 200 = 800 ÷ 200 = 4.0

You can run the same numbers in the top-box calculator, which also gives top-2 and top-3 box.

Top box, top-2-box or top-3-box?

• Top box shows enthusiasm: how many gave the best possible answer. It suits goals like "delight" and premium service.
• Top-2-box is the usual choice for 5-point scales. CSAT is often reported this way, as the share of 4s and 5s.
• Top-2 or top-3-box fits longer scales (7, 10 or 11 points), where the single top point is rarely chosen. On the 0–10 NPS scale, promoters (9–10) are a top-2-box.

Net top box is top-box (or top-2-box) minus bottom-box (or bottom-2-box). In the example, net top-2-box = 75% − 12% = 63 points. NPS works on a similar principle, but with its own fixed cut-offs (9–10 minus 0–6).

Top box vs the mean

The mean mixes all answers into one number and hides the shape of the distribution. Two stores can both average 4.0: one where everyone gave 4, and one where half gave 5 and half gave 3. Their top-box scores are 0% and 50%. Report the top box (or the full distribution) next to the mean when the difference matters.

How to interpret a top-box score

• Compare only like with like: the same question wording, scale length, labels and type of respondent. A top box on a 5-point scale and on a 10-point scale are different measures.
• Watch the trend more than a single value. External "good levels" are only meaningful if they come from a study using exactly your question and scale.
• Account for the margin of error. A top box is a proportion, so its 95% margin of error is about 1.96 × √(p × (1 − p) ÷ n). For 42% from 200 answers that is roughly ±7 points, so a move from 42% to 46% next month may be noise. The margin of error calculator does this for you.
• Read top and bottom together. A rising top box with a rising bottom box means opinions are polarizing.

Common mistakes

• Changing the definition between reports, for example top box in one quarter and top-2-box in the next.
• Changing the scale or labels and comparing with old results.
• Reporting a share without n. 60% of 15 answers and 60% of 1,500 answers are very different findings.
• Ignoring scale direction. On scales where 1 is the best answer, the "top box" is 1.
• Forgetting the middle. If many respondents cluster on the midpoint (central tendency bias), a low top box may reflect answering style more than experience.

How to measure it with a survey

1. Use a balanced rating scale with labeled ends, such as 1 = very dissatisfied, 5 = very satisfied, and keep it the same in every wave.
2. Decide which boxes you report (top, top-2, net) before fieldwork.
3. Export the answer counts per option or read them from the question's distribution in the report, then divide by the number of answers.
4. Report the score with n and the margin of error, and break it down by segment (store, product, period) where the sample allows.