What Is a Good Survey Response Rate?

Because of this, a "good" response rate for an employee survey sent by leadership is a very different number from a good one for a cold survey of the public. The most useful benchmark is usually your own previous surveys to the same audience using the same method.

Reference points from published research

These figures come from published sources. They are not general benchmarks for your survey, but they show how wide the range is.

Two lessons from this table:

1. Rates for cold outreach can be very low and still support well-regarded research. Pew notes that its studies have found little relationship between response rates and accuracy.
2. Rates for people who've already agreed to take part are much higher. Pew's panel surveys are completed by about 80% or more of selected panelists. That is partly because the hard work of recruiting happened earlier, at a much lower rate.

What about opt-in online panels?

If you buy responses from an opt-in online panel, a classic response rate usually isn't meaningful. AAPOR's Standard Definitions note that response rate calculations make little sense for non-probability samples because there's no well-defined denominator. For those studies, focus on how the sample was sourced, quotas, and quality checks instead.

How to calculate response rate (simplified AAPOR approach)

AAPOR's Standard Definitions sort every case into outcomes and give six formal response rates (RR1 to RR6). Here is a simplified version for an email or web survey.

Group everyone you invited into:

• I: completed the survey
• P: partially completed (answered enough to count as a partial)
• R: eligible but refused or broke off early
• NC/O: eligible but never responded or couldn't respond for another reason
• U: unknown eligibility (for example, you can't tell if the person belongs in your target group)
• Not eligible: bounced invitations to invalid addresses, or people screened out as outside your target. Leave these out of the denominator entirely

RR1 (most conservative): completes ÷ (all eligible + all unknowns) RR1 = I ÷ (I + P + R + NC + O + U)

RR3 (estimates how many unknowns are eligible): AAPOR's RR3 multiplies the unknown cases by e, an estimate of the share of unknowns that are actually eligible. AAPOR says e should be based on the best available evidence, not chosen to make the rate look better. RR3 = I ÷ (I + P + R + NC + O + e × U)

Worked example

Illustrative example (made-up numbers): you email a satisfaction survey to 2,000 customers.

• 100 emails bounce (not eligible)
• 300 complete (I = 300)
• 40 partially complete (P = 40)
• 60 start and quit early (R = 60)
• 1,400 never respond (NC = 1,400)
• 100 are unknown eligibility (U = 100). For example, you can't confirm they're still active customers

RR1 = 300 ÷ (300 + 40 + 60 + 1,400 + 100) = 300 ÷ 1,900 ≈ 15.8%

If you estimate that 60% of the unknowns are eligible (e = 0.6):

RR3 = 300 ÷ (300 + 40 + 60 + 1,400 + 60) = 300 ÷ 1,860 ≈ 16.1%

The completion rate among those who started is a different number: 300 ÷ (300 + 40 + 60) = 75%.

Report which formula you used. "Response rate: 16%" means little unless readers know what was counted.

Response rate is not the same as data quality

The real concern is nonresponse bias: when the people who respond differ from those who don't, on the things you're measuring.

A National Academies report on survey nonresponse summarizes this research: bias depends on both how many people don't respond and how different they are from responders. It cites a meta-analysis by Groves and Peytcheva (2008) of 59 studies that found very little correlation between nonresponse rates and their measures of bias. Pew Research Center similarly calls response rates an unreliable metric of accuracy. Pew also cautions that low rates still signal higher risk, especially when the topic is related to who chooses to respond. Its example is volunteering, where people who volunteer are also more likely to answer surveys.

Illustrative example (made-up numbers):

• Survey A has a 30% response rate. Responders are 80% satisfied, but non-responders are only 60% satisfied. The true figure across everyone is (0.3 × 80%) + (0.7 × 60%) = 66%, but the survey reports 80%, which is 14 points too high.
• Survey B has a 10% response rate, but responders and non-responders are equally satisfied. The survey reports the true figure.

Survey B is lower-response but more accurate. In practice you never see the non-responders' answers, so you look for clues instead:

• Compare respondents with what you know about everyone invited. Check plan type, region, tenure and purchase history from your records. If respondents skew heavily toward long-time customers, be careful about generalizing.
• Compare early and late responders. People who answered only after reminders are sometimes more like non-responders.
• Weight cautiously if you have reliable data on the full population. Weighting helps only for the characteristics you weight on.
• Watch for topic-driven response. A survey about a recent outage may draw the angriest customers.

AAPOR's overview of response rates makes the same practical point: report your response rate, but also look at other quality indicators, especially when rates are low.

How to improve your response rate

These steps raise response without biasing who responds.

Keep it short and say so. Pew Research Center caps its online panel surveys at 15 minutes and says the longer a survey is, the fewer people are willing to complete it. State an honest time estimate in the invitation.

Make it relevant. Tell people why you're asking and what you'll do with the answers. "We're deciding whether to extend weekend support hours" is more motivating than "We value your feedback."

Offer an appropriate incentive. A GESIS review of the research concludes that incentives increase response rates in all survey modes. Prepaid incentives generally outperform promised ones, and cash generally outperforms gifts. Where prepaying isn't practical, a guaranteed small reward on completion is usually better than an uncertain prize draw.

Send reminders. One or two polite reminders to non-responders, spaced a few days apart, are standard practice. Don't remind people who've already completed.

Time it well. Send transactional surveys soon after the experience. Avoid sending to employees during peak workload or to consumers during holidays unless the topic is seasonal.

Make it easy on mobile. Use short questions, avoid wide grids, put one question per screen, and show a progress bar.

Build trust. Use a recognizable sender name, explain how data will be used, and offer anonymity where appropriate, especially for employee surveys.

Don't over-survey. Contacting the same people too often wears out goodwill. Coordinate across teams so customers don't get three surveys in a week.

Plan for response, not just a rate

Instead of aiming for an arbitrary percentage, work backward from the number of completes you need:

Invitations needed = completes needed ÷ expected response rate

Illustrative example: you need 400 completes, and your last three surveys to this list had response rates around 12%. 400 ÷ 0.12 ≈ 3,334 invitations. If your list is only 2,000 people, you'll need to accept a smaller sample, raise response, or add another source.

Use the sample size calculator to find how many completes you need. If your own list isn't large enough, or you need people outside it, you can buy survey responses from a targeted audience. With SurveyHood you get an instant quote per approved complete, and you pay only for responses that pass automated and human quality review.

Key takeaways

• There's no single good response rate. It depends on mode, audience, topic and effort.
• Compare against your own past surveys to the same audience.
• Calculate response rate consistently, and say which formula you used.
• Nonresponse bias, not the rate itself, is what threatens accuracy.
• Improve response with shorter, relevant, mobile-friendly surveys, sensible incentives and reminders.