A research plan helps define what you want to learn, the approach and the interview guide. You can have multiple research plans per project. The interview guide is the script Candor generates from a plan, and it lives inside one. This article walks the four screens that build a plan and turn it into a guide.
A research plan defines what you want to learn, the approach, the material you want reactions to, and the goals. An interview guide is the structured question list Candor generates from a plan, and it lives inside one. Inside any project, Start research takes you to your Research Plans, or straight into building one if you have none yet. Each card shows the plan’s name and the date it was created, plus an Archived badge once you archive it. A plan’s name starts as its version and approach, like v3 · Concept testing. Open the plan (the page titled Research plan:) and click the pencil beside the title to rename it so it says what the research is about; a blank name goes back to the default.
Four screens. You answer one question, Candor suggests an approach, you add the material, and you review the plan before anything is generated.
One free-text box: what you want to learn or decide with this research. Write it the way you would say it out loud. A short list above the box says what Candor can research: problems and pain points, solutions and concepts, value propositions, names, messaging, and pricing. This is the most important thing you type. It decides which approach Candor recommends, it becomes the lens the learning goals are written through, and for problem discovery it is the required input that stops the interview falling back to generic questions. A one-line answer produces a thinner plan than a few sentences.
Candor cannot test designs or screens yet. Participants read and respond in words, so they cannot look at a mockup, prototype, page layout, logo or packaging. If your whole question is about how something looks or works on screen, you stay on this screen: a note under the box says so and suggests describing what the design offers or says instead. If only part of it is, you move on and the plan covers the rest and says so. Wording that lives on a page, like which headline works, can be tested. An answer with nothing to find out also keeps you on this screen, with a note under the box, until you reword it.
Candor reads your answer, plus the audience you built earlier (including what you said you were unsure about), and works out which of the four approaches your question needs, with a short explanation of why. Most questions need one. Some need two or three, and when yours does you get all of them, numbered, in the order to run them, each with a line on why it sits where it does. A note can also appear above the plan, naming research the plan does not contain and saying why it would help. Ask what to charge for one idea, for instance, and you get a price test, with a note suggesting you test the concept first if nobody in this study has been interviewed on a concept yet. That note is advice, not a step: to act on it, go back and reword your answer so it asks about the concepts. Ask what to charge for several ideas before this study has two tested concepts and the concept test is a step rather than a note, because comparing prices only works on concepts participants have already seen.
The plan is a suggestion, not a decision. Candor sets up the first step. The later steps are there to show you what to run next; they are not clickable, and each one is set up later as its own research plan. If the plan misreads what you meant, go back and reword your answer to get a different one.
If the suggestion cannot be worked out (a rare service hiccup), the screen says so and opens on your study’s own approach with no explanation. Going back and continuing again tries once more.
The four, each shaping how the guide gets generated and how synthesis later interprets the data:
What this screen asks for depends on the approach. Concept testing asks What should participants react to? Price testing asks Which solution(s) do you want to test pricing for? and gives each solution an optional Target price point. Problem validation asks What problem(s) do you want to test? Problem discovery asks What do you already believe to be true? and the answer is optional. Every approach also gets one optional box, Is there anything else you want to explore? Use it for specific assumptions, concerns, or topics you want to touch on. It sits at the top of the screen for problem discovery, and below the required input for the other three. It is optional for every approach, problem discovery included, because step 1 already asked what you want to learn.
Participants read each concept description word for word as the thing they react to. If a description reads like a research brief, because it talks about what you want to learn, the study, or the participants instead of the thing being tested, a yellow note appears under it as you type. Describe what you’re testing in that box, and put what you want to learn in Is there anything else you want to explore? instead. The note shows on concepts, price-test solutions, base concepts, and value proposition and message options. It skips a name test’s “Why this name” note, because participants only ever hear the name. It does not block you from continuing.
Your answer, the approach, and the material you entered, shown back read-only, with the learning goals underneath. The goals are the only editable thing here, and they are what the guide gets built from. Going back to change your inputs and coming forward again works the goals out from scratch, which replaces any edits you made to them. Candor asks before doing that.
Generate interview guide writes the guide and approves it in one step, then takes you to the guide, ready to launch interviews. Approved guides are locked, so read the plan before you click it.
Each learning goal is one specific thing you want to understand from this set of interviews. Click Add learning goal and type the goal as a short, concrete sentence. The order you list them in becomes the priority. Three to 6 goals is the range, and 6 is a hard maximum: the interview has a fixed number of turns, so more goals means thinner coverage on each. At the maximum the Add learning goal button switches off — fold the extra idea into an existing goal, or delete one you care less about.
For problem validation, you’ll also enter explicit problem hypotheses. These are statements the report gives a verdict on, based on the interview evidence.
With four or fewer problem statements, most of the guide is spent on the statements themselves. So when a learning goal isn’t answered by any of your problem statements, the guide adds one short section, asked before any statement is introduced, with a question for each such goal (up to five). Candor works out which goals those are when it writes the guide, so a goal none of your statements covers still gets asked about in its own terms, even when you have as many statements as goals.
When you move forward from the input step, Candor reviews what you’ve entered and pauses once if something looks like it would undermine the study: concepts that contradict themselves or aren’t comparable, problem statements too thin to test, an answer to “Is there anything else you want to explore?” written as instructions (question scripts, answer scales, or directions to the participants), or inputs that ask about things your audience wouldn’t plausibly know. Candor designs its own questions from your goals, so instructions written in that box aren’t followed as written. These are suggestions, not errors: fix what you want, or click the same button again to continue as-is. One exception: placeholder problem statements (like “Problem Statement 1”) block problem-validation studies until you replace them with a real statement, since they’re shown to your synthetic participants word for word.
Click Generate interview guide. Candor takes a few minutes to write a complete guide: sections with intent, question variants, interviewer notes, re-anchor variants for participants you’ve already interviewed, and stopping criteria per section. You’ll be redirected to the guide detail page when it’s ready.
You don’t have to wait on the screen. Closing the tab or going elsewhere in Candor doesn’t stop the run. Come back to the plan builder while it’s going and it shows the run still going rather than offering to start another. Come back after it finishes, in the same browser you started it in, and you’re taken to the plan it made or told why it didn’t make one. Only one guide generates per study at a time.
Each concept can appear only once per guide. In price testing, the Reuse tab pulls concepts from prior approved guides in this study. Pricing more than one concept needs every one of them from Reuse, so a second concept can only be added once the first came from there, and from then on the Text, File and URL options are switched off. A concept already added to the guide is marked “already added” and can’t be picked again. A reused concept’s name and description are locked. The participants have memory of the concept, and changing it would reduce the benefit of that memory and require them to re-learn the concept while pricing it. You can set a price to test. To A/B test two variants of a concept, add each variant as its own concept in a concept test, then price-test them together by reusing both.
The plan page is titled Research plan: followed by the plan’s name. Beside the title are two icons: a pencil to rename the plan (leave the name blank to go back to the default) and an archive box to archive it, which asks you to confirm first. Below the launch controls:
The Interview guide: how the interview flows section below those is open by default (it folds on a click) and says how many sections the guide has. Read through it before launching interviews:
When you have five or more problem hypotheses, the generator switches to a card-sort methodology. Participants sort the hypotheses into four pain buckets (very challenging, somewhat challenging, not challenging, not applicable), then the interviewer probes the top-ranked ones in depth. The guide detail page surfaces a banner explaining when card-sort is active. A card-sort guide has no extra section for learning goals your statements don’t cover; that section only appears with four or fewer statements.
When you set up a follow-on concept test — one where you test candidate names, value propositions, or messages for a single concept — Candor checks each option as you enter it, at two levels. If an option is clearly the wrong kind of thing for a value-proposition or message test — a bare product name, or a whole product description, where a benefit or a headline belongs — Candor stops guide generation until you replace it, because that option would waste the run. Softer issues — a vague benefit, a feature stated instead of a benefit, a price, or an option that just reads oddly — show a yellow note that does not block, often with a one-click rewrite of your own wording. Name tests only ever warn, never block. Editing a flagged option clears its note right away.
A guide is approved the moment it is generated; there is no separate approval step and no draft to come back to. A generated guide cannot be edited, which keeps the interview agent consistent across every interview run against it. To do different research with participants, create a new research plan. Each gets a version number (the old one stays available for reference). Plans can be archived, which hides them from the default list, but never deleted.
Start with a free project, or walk through Candor with us first.