Research Contact Policy: Frequency Caps, Suppression Lists and Who Owns the Customer Inbox

A research contact policy is a written, enforced rule set that governs how often any one customer can be contacted for research, by which team, and with what cooling-off period in between — backed by a single suppression list that every team checks before it sends anything. Without one, each team stays comfortably inside its own limit while the customer receives the sum of everyone else's.

This is the operational counterpart to survey fatigue. Fatigue is the symptom you observe in your response rates; a contact policy is the mechanism that prevents it. Most teams try to fix the symptom by rewriting subject lines. The fix is almost always upstream, in governance.

The arithmetic nobody runs

Every team believes it is being restrained. Here is what restraint looks like in aggregate at a mid-sized B2B company:

Nobody in that table is behaving badly. Each cadence is defensible in isolation. The customer, who does not see your org chart, experiences a feedback request roughly every eleven days from a company they pay to use software.

The first job of a contact policy is to make that total number visible to somebody who is accountable for it. In most organisations, no such person currently exists.

What over-contacting actually costs you

The obvious cost is the falling response rate. The Pew Research Center's telephone survey response rate fell from 36% in 1997 to 9% in 2016, then 7% in 2017 and 6% in 2018 — one of the best-documented declines in the discipline, driven partly by a robocall volume that reached an estimated 3.4 billion calls per month.

But the rate is not the real damage, and Pew's own methodological work makes this point sharply: across studies in 1997, 2003, 2012 and 2016 they found little relationship between response rate and accuracy. A low response rate is not automatically a biased one.

The real damage is compositional. Over-contacting does not shed respondents at random. It sheds the busy, the senior, the sceptical and the merely indifferent first, and retains the people who enjoy answering. After eighteen months of unmanaged contact, your "customer voice" is the voice of a self-selected minority who answer everything — and their opinions have been shaped by the twenty surveys they already completed. You end up with a healthy-looking sample size and a systematically unrepresentative one. That is a far more expensive problem than a low response rate, because it is invisible in your dashboards. Correcting it after the fact means survey weighting, which can only adjust for imbalances you measured.

The seven components of a contact policy

Setting the global cap with arithmetic, not opinion

Do not pick a number because it feels reasonable. Derive it from your contact budget:

Annual contact budget = eligible population × cap per person

Contacts required = number of planned studies × sample size needed per study ÷ expected response rate

If those two numbers do not reconcile, something has to give: a larger eligible population, fewer studies, smaller samples, or a higher yield per contact.

Worked example. You have 8,000 eligible contacts and set a cap of 4 per person per year, giving a budget of 32,000 contacts. You plan 14 studies needing 400 responses each; at an 8% response rate that requires 5,000 invitations per study, or 70,000 contacts. You are over budget by more than double. The honest options are to cut to six studies, accept smaller samples — and read statistical power and minimum detectable effect before you do, because a smaller sample may not detect the change you care about — or raise the yield per contact. That last option is the one most teams never consider, and it is usually the largest lever available.

The cool-down window

The cap controls annual volume; the cool-down controls clustering. A cap of six per year still permits six contacts in one fortnight. Set a floor of 30 days between any two research contacts to the same person, and 45–60 days after a long commitment such as a moderated session or a diary study.

Pre-emption: who wins

When two teams want the same account in the same window, you need a rule decided in advance, not a negotiation each time. A workable priority order:

1. Legal or regulatory obligations (accessibility conformance testing, regulated-sector requirements)
2. Triggered transactional feedback the customer expects (post-support CSAT)
3. Decision-blocking research with a named decision and a date
4. Tracking programmes on a fixed wave schedule
5. Exploratory and opportunistic research

Anything at level 5 yields to anything above it. Publish the order; the arguments stop.

The suppression list

One list, checked by every team, before every send. Industry codes treat this as mandatory rather than optional: the EphMRA Code of Conduct, which governs pharmaceutical market research in Europe, states under 4.9 Do Not Contact Status that research subjects who have opted out of being contacted for market research must be excluded. The principle generalises to every sector.

What belongs on it:

Two failure modes to design out. First, a per-tool opt-out is not an opt-out — if unsubscribing from the product team's survey tool leaves the CS team's tool untouched, you have not honoured the request and, depending on jurisdiction, may not have met your legal obligation either. See the law on research recruitment outreach for where those obligations actually bite. Second, suppression must be checked at send time, not at list-build time; lists built a fortnight before launch go stale precisely when it matters.

Sample rotation: protecting your best respondents

Left alone, every sampling process converges on the people who answer. Randomly sample 500 from 8,000 each quarter and, over two years, your reliable responders will have been invited far more often than the arithmetic suggests, because they are the ones who convert.

Rotate deliberately. Sort the eligible pool by last-contacted date and draw from the least recently contacted first, then randomise within that stratum to preserve representativeness. Track a simple health metric: the share of your responses in the last 12 months that came from people who responded more than three times. If that number climbs above roughly a third, you are running a de facto panel — which is a legitimate thing to run, but it should be built and managed as a panel with the recruitment and refresh discipline that implies, not arrived at by accident.

Exemptions, stated explicitly

Some contacts should sit outside the cap, but the list must be closed and written down or it will expand until the policy is meaningless. Defensible exemptions: post-transaction feedback the customer initiated or expects; a single reminder for a study they already opened; contacts to people who explicitly asked to be involved in future research; and legally required testing. Everything else counts against the cap.

Raising the yield instead of raising the cap

Here is the part most contact policies miss. The cap exists because each contact is expensive and low-yield. A survey invitation costs a unit of goodwill and returns, on average, a fraction of an answer — mostly closed-ended, with no ability to follow up on anything surprising.

That trade-off is a property of the instrument, not a law of nature. If a single contact returns substantially more understanding, you need fewer contacts to answer the same questions, and the arithmetic above stops being a crisis.

This is where AI-moderated research changes the governance problem rather than just the tooling. With a platform like Koji, one conversational interview does the work that previously took a survey plus a follow-up interview request:

• AI follow-up questions probe the interesting answers in the moment. A traditional survey returns a 3-out-of-5 and no explanation; a Koji interview asks why, and then asks about the reason behind the reason. That is the difference between spending a second contact and not needing one.
• Voice and text in the same study means you are not running a separate study to reach people who will not do one or the other. Koji's AI moderates both, and both flow into the same analysis.
• Structured questions — six types: open_ended, scale, single_choice, multiple_choice, ranking and yes_no — let one conversation carry your trackable metrics and the qualitative reasoning behind them. That collapses the classic "survey for the number, interview for the story" pattern into a single contact, which is a direct halving of the contact cost of a tracking programme.
• Real-time reports show themes and quotes as interviews complete, so you can stop at saturation rather than running the full planned sample out of habit. Contacts not spent stay in the budget.
• The quality gate means only conversations scoring 3 or above consume credits, so a contact that yields nothing useful does not cost you a credit — though it does still cost you a contact, which is the scarcer resource.

The honest version of this argument: a contact policy is still necessary with better tools. What changes is that you can hold a cap of four per year and still run an ambitious research programme, instead of choosing between the two.

Rolling it out in 30 days

Week 1 — Measure. Pull every research-adjacent send from the last 12 months across every tool. Compute contacts per person, the distribution (not the mean), and the share of responses from repeat responders. The distribution is the number that creates urgency: the mean will look fine while your top decile is being contacted twenty times.

Week 2 — Draft. Set the cap, the cool-down, the priority order and the exemption list. Keep it to one page. A policy nobody can recall is not enforced.

Week 3 — Consolidate. Build the single suppression list and give every sending team read access at send time. This is usually the only real engineering work.

Week 4 — Publish and calendar. Name the owner, put a shared research calendar in place, and make the cap visible in the request intake form so teams see the constraint at the moment they plan.

Review quarterly against three metrics: contacts per person (distribution), repeat-responder share, and response rate by contact recency. That third one is the most useful diagnostic you will have — if response rate drops steeply for people contacted within 60 days, your cool-down is too short.

Frequently asked questions

What is a reasonable global frequency cap? Four to six research contacts per person per year suits most B2B companies, with a 30–45 day cool-down between any two. Consumer businesses with large populations can afford tighter per-person caps because the eligible pool is bigger. The right answer is whatever reconciles your contact budget with your research demand — derive it, do not copy it.

Does transactional feedback like post-support CSAT count against the cap? Usually no, if it is triggered by an interaction the customer initiated and arrives immediately after it. That is feedback on an event they just experienced, and customers read it differently from an unsolicited research invitation. But cap it separately (one per month is common) or it becomes the largest source of contact volume in the company.

Who should own the contact policy? ResearchOps, where that function exists; otherwise whoever owns the research repository and intake process. The owner needs authority to arbitrate between teams, which means the role must sit outside any one of the competing teams. Ownership by committee reliably fails.

How is this different from just reducing survey fatigue? Survey fatigue is the outcome you observe — falling response rates, thinner answers, more abandonment. A contact policy is a governance instrument that prevents it, by making total contact volume visible and capped across teams. Rewriting your surveys addresses the symptom; capping and rotating addresses the cause.

Do we need a separate suppression list if our email tool already handles unsubscribes? Yes. An email tool's unsubscribe list covers that tool only, and research contacts frequently go out through several systems — a survey platform, a CRM sequence, an in-product prompt and a research platform. A person who opted out in one and keeps hearing from the others has not, in any meaningful or legal sense, been allowed to opt out.

How do we keep a cap without shrinking our research programme? Raise the yield per contact. A conversational AI interview that asks intelligent follow-ups and captures both structured metrics and open-ended reasoning in one sitting returns considerably more per contact than a survey does, which means fewer contacts for the same learning. Stopping at saturation instead of running the full planned sample recovers more budget still.

Related Resources

• Structured Questions in AI Interviews — the six question types that let one conversation carry both your metrics and the reasoning behind them
• Survey Fatigue: Why It Is Getting Worse — the symptom this policy is designed to prevent
• Is It Legal to Email or Text Someone for Research? — CAN-SPAM, TCPA, CASL and PECR rules that constrain outreach
• How to Build a Research Participant Panel — what to do when repeat responders become a deliberate asset
• Survey Weighting: How to Correct a Skewed Sample — repairing the compositional damage after the fact
• ResearchOps: The Complete Guide — where contact governance sits in a wider operations practice
• How to Increase Survey Response Rates — tactics that work once your contact volume is under control

---

Ready to get more from every customer contact? Start free with 10 credits and run your first AI-moderated interview — voice or text, with structured questions and automatic analysis.