How to Fix NPS Bias: Using Anchors to Overcome Self‑Assessment Heterogeneity

What is NPS?

Net Promoter Score (NPS) measures the likelihood that customers will recommend a brand or service to others, using a single 0‑10 question. Scores 9‑10 are "promoters", 7‑8 are "passives", and 0‑6 are "detractors". NPS is calculated as the percentage of promoters minus the percentage of detractors.

Companies Using NPS

Many leading enterprises embed NPS in their operations and KPIs, including Apple, Dell, HSBC, KPMG, China Ping An, Tmall, Didi, Ant Financial, and China Mobile.

Problems with NPS

The underlying assumptions of NPS may be flawed.

Potential misclassification of respondents.

A trade‑off exists between the method’s simplicity and its usefulness.

Self‑Assessment Heterogeneity

Research shows that respondents often differ in how they interpret the rating scale. For example, two users both give a score of 7; one considers any score ≥6 a recommendation, while the other requires ≥8. This heterogeneity can bias analysis if not controlled.

Similar heterogeneity appears in other subjective measures such as self‑rated health, happiness, and job satisfaction.

Anchoring Method to Reduce Heterogeneity

Academics address the issue by adding an “anchor” question that asks respondents to specify the score they consider a recommendation.

Question 1: “How likely are you to recommend our product/company to friends?” (0‑10).

Question 2: “From what score onward would you consider it a recommendation?” (0‑10).

Both questions share the same measurement origin, allowing the responses to be standardized and comparable across individuals.

Empirical Case: Bank A NPS Study

Data were collected via the bank’s app banner and WeChat official account, yielding 1,046 valid questionnaires (≈81 % effective rate) with an average completion time of six minutes.

Findings on self‑assessment heterogeneity:

The average personal recommendation threshold was 8.26, higher than the conventional pass‑line of 6 and the “excellent” line of 8.

Women set a higher threshold (8.36) than men (8.16), a statistically significant difference (t‑test, p = 0.04).

Education level also mattered: junior‑high respondents averaged 8.53, while those with a master’s degree or higher averaged 7.67; ANOVA showed a significant difference (p = 0.002).

Proportional Standardization

The core idea is to adjust each respondent’s original NPS score proportionally to a common benchmark (set to 9). The personal “recommendation threshold” is treated as a scaling factor.

Formula: (original score / personal threshold) = (adjusted score / 9).

Conclusion

While NPS has proven valuable across many business domains, its limitations—especially the heterogeneity of self‑assessment—must be acknowledged. Introducing anchor‑based standardization improves comparability and reliability, and the same principle applies to other subjective metrics.

References

Idler E L, Benyamini Y. Self‑rated health and mortality: a review of twenty‑seven community studies. J Health Soc Behav. 1997;21‑37.

Tandon A, Murray C J L, Salomon J A, et al. Statistical models for enhancing cross‑population comparability. Health‑systems performance assessment: debates, methods and empiricism. WHO, 2003:727‑746.

王广州, 王军. 中国家庭幸福感测量[D]. 2013.