Customer Interview Cadence: How Often Should You Talk to Users? (2026)

The right customer interview cadence is at least one interview per week per product team — and most teams should be running three to five. Teresa Torres' continuous discovery research is unambiguous on the lower bound: if you go a full month without talking to a customer, you spend that month making decisions on stale assumptions. The upper bound is set by what you can synthesize: there is no point collecting more conversations than you can read. With AI moderation eliminating the scheduling cost, the practical sweet spot for a product trio in 2026 is 3–10 customer interviews per week, run async, with synthesis happening in real time as each interview completes.

This guide gives you cadence benchmarks by team size and stage, a decision framework for picking yours, and the operational playbook for sustaining whatever cadence you choose.

TL;DR cadence benchmarks

The minimums are the floor below which decisions outpace insight. The "healthy" column is where most well-run teams live. The "high-performing" column is where AI moderation pays for itself many times over — it is essentially impossible to sustain manually.

Why weekly is the canonical floor

The modern consensus on cadence comes from Teresa Torres' Continuous Discovery Habits and the Product Talk research that preceded it. Her position: each product trio should have at least one weekly touch-point with customers, conducted by the team building the product, in pursuit of a current product outcome.

The argument for weekly is structural, not aesthetic. A monthly cadence means you spend three weeks of every four making decisions on assumptions you haven't tested. A weekly cadence keeps your model of the customer current with your roadmap. Below weekly, decisions outpace evidence. Above weekly, the marginal interview produces less marginal insight — until you are running a deeper study where the right sample size becomes 10–20.

The inverse is also true: teams that try to "save up" research into one big quarterly study almost always discover that the questions they wrote 12 weeks ago are the wrong questions now. Cadence is not just about volume — it is about staying in sync.

How to pick your cadence

Three variables drive the answer: stage, team structure, and synthesis capacity.

Stage

• Pre-PMF or new market. You have more open questions than time to answer them. Run as many interviews as your synthesis capacity allows — 10+ per week is normal for an early founding team.
• Active product development. Each squad should hit at least 1/week, ideally 3–5/week, to keep the build cycle informed.
• Mature product, optimizing. 1–2 per week is usually enough, supplemented by occasional deeper studies.
• Enterprise or regulated. Cadence is gated by access to participants, not time. 1/week is aspirational; 1/month is the realistic floor.

Team structure

A product trio (PM + designer + engineer) should run cadence per trio, not per company. Five trios at one interview per week is five interviews per week of company throughput, not one. This is where the math becomes painful for teams using traditional moderated interviews — five trios scheduling separate calls means five different recruitment ops and five sets of calendar coordination per week.

Synthesis capacity

The rule that nobody likes hearing: if you are not synthesizing each interview within 48 hours, more interviews will hurt more than they help. Unsynthesized interviews don't change decisions; they pile up as guilt. Pick a cadence you can actually metabolize. With Koji, synthesis runs in real time as each interview completes — every conversation generates structured themes, quotes, and quality scores automatically — which is what lets teams sustain cadences that would otherwise be unmanageable.

The four cadence patterns

Most successful cadences fall into one of four patterns:

1. Weekly Discovery (the Teresa Torres baseline)

One interview per trio per week. Synthesis at the end of each week. New assumptions tested next week. The lowest viable cadence for a team that wants to stay current.

2. Always-On Discovery

A standing study runs continuously, recruiting from in-product traffic or your CRM. Interviews come in throughout the week without the team initiating each one. This is where async AI interviews shine — the published study runs on its own and the team reviews completed interviews in their normal workflow.

3. Sprint-Aligned Discovery

Research is timed to product cycles: 5–10 interviews in week 1 of a sprint to scope the work, then 3–5 in week 2 to validate decisions. Useful for teams already running rigorous sprint cadences.

4. Continuous + Deep

The most mature pattern: continuous always-on discovery (1–3/week) supplemented by periodic deep studies (15–30 interviews) when a major decision approaches. The continuous stream catches drift; the deep studies answer specific strategic questions.

Operational playbook for sustained cadence

Most teams pick a cadence and miss it within four weeks. The blockers are predictable, and they are operational, not strategic.

Eliminate scheduling

Moderated calls need a calendar invite, a meeting link, a reminder, a no-show buffer, and a follow-up. At three interviews per week, that's 3–6 hours of operational overhead before any insights arrive. Async AI interviews remove this entire layer — see Asynchronous User Interviews for how. With Koji, a participant clicks a link, the AI moderator runs the conversation in voice or text, and the transcript is in your dashboard 30 seconds after they finish.

Recruit continuously, not per-study

The second-biggest cadence killer is per-study recruitment. Teams running weekly cadence build a continuous recruitment funnel — usually a research widget embedded in their product or a recurring CRM segment that drips invitations. Invitations should go out automatically, not manually.

Standardize the question set

Writing a fresh discussion guide every week is unsustainable. Maintain a "core" guide (the questions you ask in every interview to track over time) and a small "spotlight" section that rotates with the current research question. Koji's structured questions make the core comparable across interviews — every NPS rating, feature priority, or willingness-to-pay number ladders into the same chart automatically.

Synthesize as interviews complete

The synthesis-batching tax is real: if you wait until end-of-month to read 20 interviews, you get 20 interviews of recall lag and a deeply boring afternoon. Real-time theme detection means each interview produces a usable summary the moment it finishes. By Friday you have a week's worth of digested evidence, not an undifferentiated pile.

Make the loop visible

The cadence dies fast when the team can't see the impact of the conversations they're running. Pipe finished-interview notifications into Slack (Slack Research Insights Integration) so the trio sees insights land in real time. Every two weeks, link a roadmap decision back to the specific interview that informed it. Visibility creates the gravity that sustains the habit.

How Koji enables high-cadence research

The single biggest reason teams fail their cadence target is operational, not motivational. Koji is built around the cadence problem:

• Always-on async interviews mean a published study collects participants 24/7 without a moderator. See Always-On User Interviews.
• Three-layer AI probing ensures every interview produces real depth, not just answers — so you're not running more interviews to make up for shallow ones. (AI Probing Guide)
• Real-time research insights mean themes, quotes, and quality scores surface as interviews complete, not weeks later. (Real-Time Research Insights)
• Quality gating ensures that only interviews scoring 3 or higher consume credits — junk interviews never enter your dataset and never cost you anything.
• Six structured question types let you compare quantitative signals (NPS, ranking, scale) across every week's interviews automatically.
• Participant management at scale. Personalized links, CRM imports, and embed widgets all run from one workspace (Managing Research Participants).

With these in place, the marginal cost of moving from one interview a week to ten is essentially zero — which is what lets a small team operate at the cadence of a full research function.

Related Resources

• Structured Questions in AI Interviews — the six question types that make weekly comparisons possible
• Continuous Discovery: Weekly Customer Interviews — Teresa Torres' methodology in depth
• Always-On User Interviews — the 24/7 standing study pattern
• Asynchronous User Interviews — why async beats moderated for cadence
• Real-Time Research Insights — synthesis as interviews complete
• In-Product Research Recruiting — the recruitment funnel that sustains cadence
• Slack Research Insights Integration — pipe insights to where decisions get made