Ask for a sample quote and the first three questions back are always the same. What is the incidence? How long is the survey? Who is the audience? The answers set the cost per interview, and getting the first two wrong is the most common reason a project runs over budget or stalls in field.
Incidence rate
Incidence is the share of people who qualify among those who start the survey. If 100 people enter and 25 pass the screener, incidence is 25%.
ResearchDart calculates live incidence as completes divided by completes plus terminates. Over quotas and quality terminates are left out: the first group qualified, and the second was removed for reasons that have nothing to do with targeting.
Why it drives price
Every screen-out still costs the supplier. They invited the person, the person gave up their time, and many panels reward screen-outs. At 50% incidence a supplier needs about two entries per complete. At 5% they need about twenty.
What the benchmark shows
In the Insights Association and GDQ Data Quality Benchmarking Wave 1 report, incidence sold to suppliers averaged 50.8% and actual incidence averaged 41.1%. For research agencies it was 61.6% sold and 55.2% actual. Healthcare provider studies were sold at 65.7% and came in at 42.5%.
In other words, the industry quotes easier studies than it fields. Our reading of the benchmark goes through what that means for feasibility.
Estimating it without optimism
Start from the most specific qualifying condition and work back to the population the supplier will invite. Watch for stacked conditions: if 40% own a car and 30% of those plan to buy one this year, incidence is 12%, not 30%. Then soft launch and compare. If actual incidence is materially below the quote, re-quote before full field, not after.
Length of interview
LOI is how long a qualified respondent takes, normally the median in minutes. The benchmark put median LOI for qualified completes at 12 minutes for general B2C studies, 17 for general B2B, and 34 for healthcare provider studies.
Research teams' own test timings are nearly always too short, because they already know the questions. Use the median from the first real completes.
LOI also sets a quality control: a minimum LOI below which a complete is treated as a speeder. Around a third of the median is a common starting point. Surveys with heavy skip logic produce genuinely short completes, so check the distribution before applying a hard rule.
Cost per interview
CPI is the price of one complete. It moves mainly with:
- Incidence: lower incidence, higher price.
- LOI: longer survey, higher price, more drop-off. In the benchmark, 12.6% to 13.9% of respondents dropped out globally.
- Audience: B2B, medical and hard-to-reach groups cost more.
- Country: panel availability and incentive norms differ by market.
- Deadline: tight field time narrows the suppliers who can deliver.
On an exchange, the buyer's CPI and each supplier's payout are separate numbers. On ResearchDart the project carries the buyer CPI and each allocation carries its own supplier payout.
The cost nobody puts in the quote
Post-survey cleaning. The same benchmark found 6.6% to 7.4% of qualified completes removed after the survey globally, and 18.9% for general B2B. A quote that assumes every complete survives cleaning is a quote that will be renegotiated. Build the overage in at the start, and reverse bad completes with GDQ codes so the conversation with the supplier is about causes rather than blame.
Reading live numbers
| You see | It usually means | What to do |
|---|---|---|
| Incidence well below quote | Targeting broader than described, or a stricter screener | Re-quote or add targeted suppliers |
| LOI well above quote | Survey longer than estimated | Shorten or re-quote |
| Many over quotas | Cells filling unevenly | Tell suppliers which cells are closed |
| Many quality terminates from one supplier | Speeders or fraud from a source | Check screening flags for that allocation |