Measure Customer Trust: Surveys & Behavior
This guide shows you how to measure customer trust by combining two complementary sources: survey responses (what customers say) and behavioral data (what customers do). You’ll learn to design a short trust survey, identify key behavioral signals, and merge both into a single trust score that you can track over time. No special software is required beyond a spreadsheet and your existing analytics tools.
What You’ll Need
- A survey tool (e.g., Google Forms, Typeform, SurveyMonkey) or a way to email a questionnaire.
- Access to your website/app analytics (e.g., Google Analytics, Mixpanel, or your own logs).
- A customer database or CRM with purchase history and support interactions.
- A spreadsheet program (Excel, Google Sheets) for combining and scoring data.
- At least 100 customers to survey for statistically meaningful results (fewer is okay for qualitative insights).
Step-by-Step
- Define what trust means for your business. Trust isn’t one thing. For an e-commerce site, it might be delivery reliability; for a SaaS, it might be data security. Write down 3–5 specific behaviors that would prove a customer trusts you (e.g., repeat purchases, referring a friend, upgrading a plan).
- Design a short trust survey (5–7 questions). Use a mix of Likert scale (1–5) and open-ended questions. Recommended items:
- “How likely are you to recommend us to a friend?� (0–10, NPS style)
- “How much do you trust us to keep your personal data safe?� (1–5)
- “How confident are you that we will deliver on our promises?� (1–5)
- “Have you ever hesitated to purchase because you weren’t sure about us?� (Yes/No)
- “What would increase your trust in us?� (open text)
- Distribute the survey to a representative sample. Send it to a random subset of your active customers (not just your most loyal). Aim for at least 100 responses. Keep it anonymous to encourage honesty.
- Score the survey responses. Convert each Likert answer to a number. For the NPS question, classify promoters (9–10), passives (7–8), detractors (0–6). Calculate the average trust score: sum all Likert scores and divide by the number of questions. For NPS, subtract the percentage of detractors from the percentage of promoters.
- Identify behavioral signals from your analytics. Pick 3–5 metrics that reflect trust. Common ones:
- Repeat purchase rate (percentage of customers who buy again within 90 days).
- Churn rate (percentage of customers who cancel or don’t renew).
- Average order value (higher values often indicate greater trust).
- Support ticket sentiment (positive vs. negative language).
- Referral clicks or shares (if you have a referral program).
- Normalize the behavioral metrics. Since these metrics have different units (percentages, dollar amounts, counts), convert each to a 0–100 scale. For example, if your repeat purchase rate is 40%, that becomes 40. For churn, invert it: if churn is 5%, the trust score is 95 (100 – 5).
- Combine survey and behavioral scores. Decide on weights. A common approach is 50% survey, 50% behavior, but you can adjust based on your confidence in each. For instance, if you have few survey responses, weight behavior higher. The final trust score = (survey average × 0.5) + (average of behavioral scores × 0.5).
- Track the score over time. Re-run the survey quarterly and compute behavioral metrics monthly. Plot the trend in a line chart. A rising score indicates growing trust; a falling score signals a problem.
- Segment the data. Break down the trust score by customer segment (new vs. long-term, high vs. low spend, different product lines). This reveals which groups trust you less and where to focus improvements.
- Act on the findings. If survey comments mention shipping delays, fix your logistics. If behavioral data shows high churn after the first month, improve onboarding. Trust measurement is only useful if it drives action.
Common Mistakes
- Asking too many questions. Long surveys get lower response rates and rushed answers. Keep it under 10 questions.
- Ignoring non-respondents. People who don’t answer may be less trusting. Compare the behavior of respondents vs. non-respondents to check for bias.
- Using only one data source. Surveys alone are subjective; behavior alone can be misleading (e.g., high repeat purchases might be due to lack of alternatives). Always combine both.
- Not normalizing metrics. Mixing a 0–10 scale with a 0–100% scale without scaling will skew the score.
- Overlooking small sample sizes. With fewer than 30 responses, the survey average is unreliable. Treat it as directional, not definitive.
Tips
- Keep the survey consistent each time you run it so you can compare results over time.
- Use a 5-point Likert scale instead of 7 or 10 for simplicity—most people can’t reliably distinguish more than 5 levels.
- Include one open-ended question to get qualitative context behind the numbers.
- Automate the behavioral data collection using your analytics tool’s API or a simple export.
- Share the trust score with your team monthly so it becomes a key performance indicator.
- Consider a control group: survey a small group of non-customers to see how your trust score compares to the general public.
Troubleshooting
- Low survey response rate. Send a reminder after a week, offer a small incentive (e.g., a discount code), and keep the survey short. If you’re still below 50 responses, extend the collection period.
- Behavioral data seems contradictory to survey responses. For example, customers say they trust you but churn is high. Investigate the timing—maybe trust is high at the moment of survey but drops later due to a specific event (e.g., a billing issue). Look at the data in segments and check for recent changes.
- Trust score is flat even after improvements. Trust changes slowly. Give it at least two quarters before expecting a significant shift. Also, check if your survey questions are sensitive enough—maybe they’re all 4s and 5s, leaving no room to increase.
- You don’t have access to all behavioral metrics. Start with what you have (e.g., repeat purchase rate) and add more later. Even one behavioral metric combined with a survey is better than nothing.
- Survey results are all positive but customers complain on social media. This may indicate that your survey is biased—perhaps you only sent it to loyal customers. Re-sample a random group, and consider adding a question about negative experiences.
Frequently Asked Questions
How often should I measure customer trust?
Run the survey quarterly and compute behavioral metrics monthly. Trust changes gradually, so monthly tracking of behavior helps catch early shifts, while quarterly surveys provide deeper context.
What is a good trust score?
There’s no universal benchmark. A score above 70 out of 100 is generally considered good, but it varies by industry. Compare your score to your own past performance and to competitors if you can find public data.
Can I measure trust without surveys?
Yes, but it’s incomplete. Behavioral data like repeat purchases and low churn are strong signals, but they don’t tell you why trust exists or how to improve it. Surveys add the “why� and catch early signs before behavior changes.
How do I handle customers who don’t respond to the survey?
Compare the behavior of respondents vs. non-respondents. If non-respondents have lower repeat purchase rates, your survey results may be overly positive. Consider weighting your survey score downward accordingly.
What if my sample size is too small?
With fewer than 30 responses, treat the survey as exploratory. Combine it with behavioral data and look for qualitative themes in open-ended answers. You can also run a shorter pulse survey more frequently to gather more responses over time.
Should I use a single question or multiple questions to measure trust?
Multiple questions are more reliable because trust is multi-dimensional. A single question like “Do you trust us?� is too vague. Use 3–5 questions covering different aspects (reliability, honesty, competence) and average them.
Related
Method: the standard formula for this calculation, rounded as each result states. See how the calculators work or report an error.