Candor: pay per job, with a subscription mixed in

Real businesses rarely fit one model cleanly. Here's how Candor works, and where it isn't pure pay per job.

Candor is a pay-per-job business. Customers pay for each interview in a study. But a subscription sits on top, and that changes the diagram and the numbers I watch.

Candor is the synthetic user research product we're building at Highline Beta. You describe an audience. Candor researches it, creates synthetic participants grounded in that research, interviews them and writes a report.

I drew Candor's business model as a systems diagram in Focused Chaos. This page walks through it, so you can see where a business fits a model and where it doesn't.

Start with the three questions

The three questions on the overview place a business: who pays, what they pay for and who does the work. Here's how Candor answers them.

Question	Candor's answer	Matching model
Who pays	Product managers, consultants and founders who need research	Pay per job
What they pay for	Each automated interview: $10, or $7 on a $75-a-month plan	Pay per job, with a subscription mixed in
Who does the work	Candor's software	Pay per job

The core: pay per job

A study is the job. Customers know the price before the work starts, and they pay whether or not the study tells them what they hoped. That puts all the pressure on quality. If the first study isn't useful, there's no second one.

One twist: I price the interview, not the study. A study with more interviews costs more, so the customer decides how big each job is. Revenue is customers × studies × interviews per study × price per interview.

Every new workspace gets a trial project with one real study. That's standard for pay per job, because nobody sees the product work without a real job. But every free study costs me compute, so I count it as part of what it costs to win a customer.

Customers also come and go with their projects. A founder might run a study before a big decision and not need another for two months. That founder isn't lost, just between jobs. That's why reuse, a second paid study, is the metric that matters.

Where it isn't pure pay per job

Customers can stay on the free plan and pay $10 for each automated interview. Or they can pay $75 a month for the Standard plan. (Interviews you run by hand yourself are free on both.)

The Standard plan does two things. It lowers the interview price from $10 to $7. And it gives you five project slots instead of two, where each project is one audience. Extra slots cost $15 a month on the free plan, or $10 on Standard.

The slots feel like traditional SaaS: a monthly fee for capacity, whether or not you run research that month. They're also a little like Subscription for usage credits, where a monthly fee buys a set allowance. But a slot holds a project. It doesn't get used up the way a credit does.

The lower interview price is already part of Pay per job, which includes subscribing for a lower price per job. It still means Candor isn't pure pay as you go. Users are unlimited on every plan, because I charge for research, not headcount.

What the subscription changes in the diagram

In pure pay as you go, a customer leaves by going quiet. With a subscription in the mix, Candor's diagram has two ways to leave, and customers can move between plans.

Dormant, then churned. A pay-as-you-go customer never cancels. They just stop running studies. Days since the last study is the signal, and a quiet customer may only be between jobs.
Account cancelled. A subscriber can cancel their account. That's a clear event with its own paid churn rate, the way it works in SaaS.
Subscribe and unsubscribe. Repeat customers move between the two plans. Every study is still paid either way. The plan sets the price per interview.

The scorecard

The key metric comes straight from pay per job: do people come back and pay for a second study? The guardrails stop me from improving it by breaking something else. The subscription shows up in the diagnostics, because the plan people pick after the trial tells me whether price is the problem.

Role	Question	What I'm measuring
Key metric	Do trial workspaces come back and pay?	Share of workspaces that run a second, paid study within 30 days. Target: 20–30%.
Guardrail: quality	Was the first study useful?	Qualitative feedback, plus a one-click survey on reports to assess the value.
Guardrail: economics	Does usage support the business?	Gross profit per study, including model calls and retries.
Guardrail: economics	Can I afford the free trial?	Free-study cost per paying workspace. At 25% conversion, each paying customer covers about four free studies.
Diagnostic	Is the silence surprising?	How often each customer expects to run research, against days since their last study.
Diagnostic	Who comes back?	Reuse by role (product managers, consultants, founders) and use case.
Diagnostic	Is price the barrier?	Plan chosen after the trial, and price objections in replies.

What I don't know yet

My 20–30% conversion target from free trial to paid might be high against early benchmarks I've seen around AI-tool conversion. I'm also not sure that conversion rate pays for the free trials. Candor is in the Stickiness stage (in Lean Analytics terms), so that's the question to answer first.

If I miss the target, the scorecard tells me where to look. Poor feedback points at the product. Good feedback but no second study points at the wrong customer, or a need that doesn't come up often enough. Price objections point at the plans.

Map your own business

Find your core. Answer the three questions. The model that matches the most answers is your core, and its diagram is your starting point.
Identify changes to your core model. Most businesses are a blend of multiple models. For Candor, it's the subscription.
Draw the changes. Each extra piece adds paths to the diagram, like a second way for customers to leave.
Keep one key metric from the core. Use the extra pieces for guardrails and diagnostics.

Go further

Systems Thinking Needs a Scorecard, the Focused Chaos post this case study comes from
Candor's pricing page

This diagram first appeared in Focused Chaos, where I called it Lean Analytics Model Seven. On this site, its core is model 4, pay per job.