Research tools do not really come in two pricing models. They come in six metering units, and the unit — not the sticker price — decides what your research costs.
The short answer: per-seat pricing fails by waste (Zylo's 2026 SaaS Management Index puts average license utilization at 54% — you buy roughly two seats to get one active user — costing the average organization $19.8M a year in unused licenses). Usage-based pricing fails by surprise (78% of IT leaders reported unexpected charges tied to consumption-based or AI features in the same index, and 61% of organizations cut projects or initiatives because of unplanned SaaS cost increases). Neither model is cheaper in the abstract. Pick the one whose failure mode your team can absorb: seats if research demand is steady and the roster is small and fixed; usage if demand is bursty and the audience for insight is wide.
The six metering units, with real 2026 numbers
Most "pricing model" articles stop at per-seat vs usage-based. In research tooling that misses four of the six units actually in market. Here is what each vendor is really counting, verified from their own pricing pages:
| Metering unit | Who prices this way | The 2026 number | What it punishes |
|---|---|---|---|
| Per seat | Great Question, SurveyMonkey, UserTesting | $129/seat/mo, or $1,290/seat/yr, capped at 5 paid seats on self-serve | Wide, occasional access |
| Per launched study | Optimal Workshop | $199/mo billed annually ($2,388/yr) for 5 launched studies per year = $477.60 a study | Running many small studies |
| Per session | Lookback | Session packs of 10 at $299 / $178 / $137 → $29.90 / $17.82 / $13.74 per session | Bursty demand on annual-only plans |
| Per participant-minute | Lyssna | 1 credit per minute per participant, ~$1/credit, +2 credits for a screener, +2 for a recording | Depth — longer sessions cost linearly more |
| Percentage of participant pay | Prolific | 42.8% platform fee for corporate customers, 33.3% for academic/non-profit | Paying participants properly |
| Per credit (action-metered) | Koji | From €1 per qualified interview, €3 per qualified voice interview | Nothing much, until volume is very high |
Each of these is a different bet about what correlates with the value you get. Seats bet on headcount. Studies bet on project count. Credits bet on conversations. When the bet is wrong for your team, you overpay — and the sticker price never tells you which.
We have broken down each of these vendors individually: Great Question pricing, Optimal Workshop pricing, Lookback pricing, Lyssna pricing and Prolific pricing.
What per-seat pricing actually costs
Per-seat is popular because it is legible. Finance understands headcount. The problem is that headcount is a terrible proxy for research activity.
Zylo's 2026 index tracked license utilization rising from 47% in 2024 to 54% — a 13% improvement that still left $19.8M of average annual license waste (down from $20.9M). Read that the other way round: after a year of active optimization, nearly half of every seat-based subscription is still dead weight.
Research tools make this worse than average, for three structural reasons:
- Research is bursty by nature. A designer who runs three studies in Q1 and none in Q2 pays the same twelve months of seat fees either way.
- The people who need the insight are not the people who run the study. A PM who reads one report a quarter cannot justify a $1,290 seat, so the insight stays locked behind the researcher — the "single point of contact" failure that research democratization is supposed to fix.
- Seat caps force a plan jump, not a seat purchase. Great Question's self-serve tier stops at 5 paid seats and Enterprise starts at a 5-seat minimum. There is no gentle upgrade path: the sixth person triggers a custom annual contract, and the security review that comes with it gates SSO behind the same door.
That last point is the one to watch. In this category a "seat" price is often really a plan price wearing a seat costume.
What usage-based pricing actually costs
Usage-based pricing solves the waste problem and creates a forecasting problem. The 2026 index found 77% of IT leaders hit unexpected costs that surfaced after a contract was signed, and 78% specifically reported unexpected charges tied to consumption-based or AI features. Spending on AI-native SaaS applications rose 108% year over year, to an average of $1.2M per organization — a lot of new meters spinning in a lot of places nobody had budgeted for.
The consumption traps specific to research platforms:
- Undefined units. Optimal Workshop meters "launched studies" and — as we found when we read both the pricing page and the help-centre plan comparison — never defines what a study is. Their own documentation confirms chained studies stay separate in the workspace, so a card sort, a tree test and a survey on one project consume three of your five annual launches.
- Meters that punish the thing you want more of. Lyssna charges per participant minute. A 20-minute session costs twice a 10-minute one, and adding a recording both costs 2 credits and lengthens the estimated duration you are billed against.
- Percentage fees that scale with fairness. Prolific's fee is a percentage of participant reward, so paying people the recommended £9.00/$12.00 an hour rather than the £6.00/$8.00 minimum raises your platform fee by the same 50%.
- Overage priced above list. Not universal — Lookback's extra session packs price at parity with the plan rate — but common enough that it belongs on the diligence checklist.
The break-even arithmetic
Here is the calculation nobody runs before signing.
A seat is worth buying when the annual seat cost is less than what you would spend metering that person's actual work. Formally:
Seats win when: annual seat price ÷ price per metered unit < units that person will actually consume in a year.
Put 2026 numbers in it. A self-serve research-ops seat at $1,290/yr, against a voice interview at as low as €3 per qualified voice interview, breaks even at roughly 430 interviews per person per year: about 1.7 interviews every working day, every day, for a year. Below that volume, the seat is the more expensive way to buy the same research.
Now the same comparison at realistic volume. Take a four-person product team running 8 interview-based studies a year at 12 participants each — 96 conversations:
| Model | Annual platform cost | Cost per conversation | Notes |
|---|---|---|---|
| Per-seat self-serve (4 seats @ $1,290) | $5,160 | $53.75 | Before incentives or recruiting fees; 100-participant cap included |
| Per launched study ($2,388, 5 studies/yr) | $2,388 at the cap | $477.60 per study | 8 studies does not fit — this plan forces an Enterprise quote |
| Credit-metered (Koji Interviews plan, billed yearly) | €790 | ≈ €8.23 | 948 credits a year against the 288 that 96 voice interviews use; 3 seats included, a fourth builder adds €150 a year, and readers join as free viewers |
The gap is not a rounding difference. It is the difference between paying for capacity and paying for work.
Two honest caveats. First, none of these figures include participant incentives or recruiting fees, which are frequently the largest line in the budget — our user research cost calculator handles the full stack. Second, a research-ops platform's coordination work is genuinely real; the fair criticism of a per-seat ops tool is not its price but the researcher-hour it does not remove, because a human still sits in all 96 sessions. Add a loaded researcher hour at $50–75 and a $50–100 incentive to a $53.75 platform cost and the true per-interview figure lands somewhere between $150 and $230. See UX researcher salary and team cost benchmarks for the loaded-cost inputs.
Match the model to your team shape
| Your team | Best-fit model | Why |
|---|---|---|
| 1–2 dedicated researchers, continuous cadence | Per seat | High utilization is exactly the case seats are priced for |
| Product team where 8 people occasionally need insight | Usage/credit | Seats for occasional users are the definition of shelfware |
| Agency or consultancy with lumpy client work | Usage/credit or per session | Cost tracks billable projects instead of stranded capacity |
| Enterprise with steady programme + procurement | Per seat, negotiated | Zylo found 12-month contracts averaged 16.4% savings, vs 14% at 24 months and 13% at 36 — the longer lock does not buy a better rate |
That last row deserves emphasis, because it inverts the standard procurement instinct. Multi-year deals grew from 23% to 38% of SaaS agreements, yet in Zylo's data the shortest contracts negotiated the deepest discounts. Length is being traded for something other than price.
The six questions to ask before you sign
- Define the unit in writing. What exactly is a "study", a "session", a "response", a "seat"? Get it in the order form, not the marketing page.
- Do unused units roll over? Monthly-reset credits are a use-it-or-lose-it tax on bursty teams.
- What is the overage rate versus the in-plan rate? Parity is fair; a markup is a penalty for succeeding.
- Who counts as a seat? Ask specifically whether viewers, observers and stakeholders are free.
- What is gated by plan rather than by usage? SSO, SCIM, data residency and audit logs sitting behind an enterprise wall means your security review, not your volume, sets your price. See enterprise security for AI research platforms.
- Where does participant cost sit? Inside the meter, passed through at cost, or marked up? Check the vendor DPA at the same time — the commercial and legal reviews share most of the same questions.
Where Koji sits
Koji meters conversations, and keeps seats for the people who build studies. Readers never need one.
- Credits pay for the research. One credit runs a text interview and three run an AI-moderated voice interview. On paid plans extra credits cost €1 each, so a text interview works out at about €1 and a voice interview at about €3.
- Seats are for builders. Each plan includes seats per workspace: one on Free and Insights, three on Interviews, ten on Enterprise. Interviews adds extra seats at €15 per seat per month (€150 a year on annual billing), and Enterprise adds them on request.
- Viewers are free on every plan, Free included. A viewer reads studies, reports and results and never takes a seat. So the PM who reads one report a quarter, the founder, the client and the sales lead all get the evidence without a licence.
- A quality gate protects the meter. A conversation that scores below 3 out of 5 is free, so a participant who drops out after two words does not cost you a credit.
- Prices are published. The plans and credit costs are on the pricing page. There is no sales call to find out what an interview costs.
- 10 free credits on signup, no card needed.
- Top-ups are in your control. Credit packs cost €1 a credit and last 12 months. Auto-refill and pay-as-you-go overage are optional, with spending caps you set. Plan credits expire 30 days after they are added, so size the plan to your usual month and top up for the busy ones.
- API and MCP connector come with the Interviews and Enterprise plans.
- Participants can come from Koji. You do not need an audience of your own: describe who you want and Koji returns a per-respondent quote that you approve before anything launches.
Every interview is a real AI-moderated conversation that asks its own follow-ups, transcribes itself and lands in an automatic thematic analysis, so the researcher-hour that per-seat ops tools leave on your payroll is not in the model. Koji also supports six structured question types (open ended, scale, single choice, multiple choice, ranking and yes/no), so one study can carry the numbers and the reasons without a separate survey tool.
The practical upshot for team shape: two people who build studies fit inside the three seats on Interviews, and everyone else who needs the findings joins as a viewer at no cost. Your bill grows with the number of conversations, not with the number of people who read the report.
For a full cross-vendor cost comparison, see what UserTesting, Qualtrics, Dovetail, dscout and Maze actually cost, or how to do user research on a budget if the constraint is spend rather than model.
Frequently Asked Questions
Is usage-based pricing always cheaper than per-seat?
No. Usage-based is cheaper when utilization is low or uneven, which is the common case — Zylo's 2026 index put average license utilization at 54%. But a heavy, continuous user can exceed the break-even point. The rough test: divide the annual seat price by the per-unit price. At $1,290 a seat against a €3 qualified voice interview, one seat only pays for itself past roughly 430 interviews a year.
Why do research tools meter studies instead of seats?
Because a study is a better proxy for value delivered than headcount — in theory. In practice it creates a definition problem. Optimal Workshop meters launched studies but does not publish a definition of "study", and its own documentation confirms chained studies remain separate, so one research project can consume several of your annual launches.
What is the biggest hidden cost in per-seat research tools?
Not the seats you buy — the people you exclude. When a seat costs $1,290 a year, stakeholders never get one, so insight bottlenecks behind whoever holds the licence. The measurable cost is shelfware: roughly half of seat spend goes unused, at an average of $19.8M a year per organization in Zylo's data.
Do longer contracts get better research tool pricing?
Not according to the 2026 data. Zylo found average savings of 16.4% on 12-month contracts, 14% on 24-month and 13% on 36-month — the shortest terms negotiated the deepest discounts, even as multi-year deals grew from 23% to 38% of agreements. Long terms buy stability, not price.
How should I forecast a usage-based research budget?
Forecast in units of work, not currency: studies per quarter × participants per study × credits per participant. Then add a 20–30% buffer, since 78% of IT leaders reported unexpected consumption or AI-related charges last year. Check whether unused units roll over and whether overage is priced at or above the in-plan rate.
Does Koji charge per seat?
Partly. Seats are for the people who build studies: each plan includes some (one on Free and Insights, three on Interviews, ten on Enterprise), and Interviews adds extra seats at €15 per seat per month. Viewers are free on every plan, including Free, and never take a seat, so people who only read reports cost nothing. The research itself is metered by credits: one per text interview, three per voice interview, €1 per extra credit on paid plans, and a conversation that scores below 3 out of 5 is free.
Run the comparison on your own numbers
Take the last four studies you ran. Count the conversations. Multiply by €3. Compare that against what a seat-based plan would have charged you for the same quarter — including the seats held by people who logged in twice.
If the metered number is smaller, the model is the problem, not the vendor.
Start with 10 free credits — enough for three full AI-moderated voice interviews, no card and no sales call required. Invite the people who only read findings as viewers: they are free on every plan.
