Project setup is a three-step wizard that creates the project and its audience together, so there’s no separate step for making a project first: name the project, describe the audience, then review everything, including what your own words were read as, before you start. It takes about three minutes of your time, then Candor takes 25-35 minutes in the background to generate your audience and its participants. Here’s what each field controls and how to fill it in.
You don’t need an existing project. Naming the project is the first field of this same setup. What you do need is a clear sense of who you want to understand and roughly what you want to learn. The description doesn’t have to be perfect, since the audience generation phase refines it, but vague inputs produce generic participants, so spend a moment on this before you start.
If you have prior research (interview transcripts, market reports, survey results, competitor analyses), have those documents ready to upload. Candor will extract signals from them and weight them higher than web evidence in the audience.
A short label you’ll recognise later. Example: Bookkeeping research. This shows up in your dashboard, in shared links, and in emails to teammates. It has no effect on who gets generated.
One sentence naming who you want to speak with. Example: SMB owners evaluating bookkeeping tools. This one does the work: it’s the sharpest signal Candor has for narrowing the audience, so a specific phrase beats a bare label. “Home owners” casts far wider than “home owners who use their backyard for hosting and family meals”.
B2B and B2C use different attribute models, personality distributions, and bias profiles. Pick the one that matches the actual decision being made. If a small-business owner is buying for their own business, that’s B2B. If a freelancer is buying personal software, that’s B2C.
Region pinning matters more than people expect. Personality distributions, cultural defaults, regulatory context, and purchasing behavior all shift by region. Pick the most specific region that actually represents the audience. If your audience is global, pick “Global” explicitly rather than leaving it blank.
When you click Continue, Candor briefly reviews what you’ve entered and may suggest sharper phrasing. You stay on the step you are already on, and Candor moves the page down to the recommendations so you can see what came back. For example, an audience defined by something people own (“homeowners with decks”) tends to surface product content rather than evidence about the people themselves, so Candor may suggest reframing it around who they are and how they use the space. Apply a suggestion with one click, edit the field yourself, or continue as-is; a recommendation that is only an observation offers the jump to the field rather than a one-click apply. The same quick review runs on the detail fields when you click Generate audience, including the fields where you describe motivations, frustrations, goals, and tools. If an entry looks too thin or unclear to steer good research, Candor flags it so you can sharpen it. Recommendations are optional and never block you.
Two to four sentences going deeper than the one-line description on step 1. Cover who they are, what context they’re in, and what kind of decision they’re making. Avoid corporate filler. Specifics beat adjectives. Age and income (B2C), or role, company size and tools (B2B), sit under it, along with what you already know and what you’re unsure about. Write company size as a number of employees, such as “100 to 500” or “fewer than 50”: the line under the box shows the range Candor will use. What you’re unsure about plans your interviews. It does not shape the audience or its web research.
Upload any research documents you want grounding the audience. Candor parses each file, chunks it, embeds it for semantic search, and extracts signals from it asynchronously. Supported formats include PDF, DOCX, and CSV. There’s a per-file size limit shown in the upload panel.
You can submit a project with no documents. Candor will fall back entirely to web evidence. Documents you upload always weight higher than web evidence in the resulting audience.
Uploaded files are private to the project and automatically deleted 30 days after upload. Candor keeps the signals it extracts from them, so your audience and participants stay intact after the original file is gone. See data and privacy in Candor for the full retention policy.
The last step lists every input back to you with an Edit button on each. It also shows something you didn’t type: the participant details your own words already decide.
Those details sit in one list headed Every participant will match this list. It also holds the requirements no field can hold, such as “eats eggs for breakfast most days” (labelled Must), every line of the exclusions box (labelled Never include), and any community or gender your words confine the study to. A line goes on the list only when your words say it of everyone you want to talk to. A view about some or most of them, or anything about your own product or service, stays off, because that is what the research tests.
Under each line you see the sentence it came from, as Because you wrote: “…”. Drop a line that isn’t true of everyone, reword it, or type your own into Add something every participant must match and choose Must or Never include. A plain list of groups, such as “homeowners and renters”, becomes one line saying every participant is one of them, and the research decides the mix.
Every participant Candor builds is checked against every line after it is written. For one whose profile contradicts a line, the fields that contradict are corrected, up to twice, and the participant is left out only if it still contradicts after that. Nothing else in “Tell us more about this audience” or “What do you already know” shapes who the participants are, though your interview guide still uses all of it. The exclusion lines you keep or add become the study’s exclusions.
Write “I want to speak with Black women” and the list shows Only these communities with a tick beside each heritage background Candor can build (African-American and Caribbean-American, for that example), plus Only women. Untick a community you didn’t mean, or drop a limit if your words don’t set one. A community Candor has no background for yet is named as such, and those participants get names and backgrounds written by the model instead.
If a box you filled in disagrees with what you wrote about the audience, the review marks it under that row. Say the age range box reads 18-25 but your description says Gen Z, which is about 18 to 29 today: the age range row says your words ask for a different range. The same check covers role and region. When two answers disagree, the box is used, so change whichever one is wrong. The note is only advice and never stops you starting.
Steps 2 and 3 both carry a Back button at the bottom left, beside the button that moves you on, so you can step back without losing what you have typed. It disappears once generation has started, because nothing typed after that point reaches it.
Write “Gen Z coffee drinkers” and Candor reads an age band out of “Gen Z”, then builds every participant inside it. Write “nurses in rural settings” and it reads where they live. These are fair readings of what you wrote, but they are still readings, and a wide one can be wrong for what you are actually studying. Gen Z reaches down to about 13, which is not who buys premium coffee.
So each one gets a row you can change before the 25-35 minutes start. Edit the value to what you meant, or drop the reading and let the evidence decide instead. A row you leave alone is taken as read.
Only readings from your free text show up here. Anything you set in a box already has its own row further up, so it isn’t repeated. And when one of these rows covers something you left blank in the setup fields, changing it here is enough. You don’t need to go back and fill that field in too.
Changing a value counts as you telling us, so that field reads Grounded on the participant profile with your correction named on it. Dropping a reading means we never claim you told us anything about that field, and it reads Derived instead. Both of those are settled after generation rather than during it, so they hold whatever the research turns up. See what evidence-grounded means.
What you type is saved in this browser as you go, so you can reload, close the tab or step away and pick up later on the same screen, with your answers and any recommendations still there. It lives in this browser only, so a different computer or browser starts with an empty form. To clear it and begin again, use Start over at the top of the page.
Documents are the exception. A browser can’t hold the files themselves, so when you come back Candor lists the ones you had attached at the top of the page and asks you to attach them again. If you no longer need them, choose I don’t need these.
If you leave after pressing Generate audience and participants, the project already exists. Coming back to project setup within the hour takes you to its progress screen instead of an empty form, so you don’t create it twice.
The slot isn’t claimed until you press Generate audience and participants, so on a shared workspace a teammate can take the last one while you’re filling this in. Candor stops you there and says that’s what happened, rather than leaving you wondering why you’re being asked for a slot you thought you had. You can add one from that message, or archive a project you’ve finished with. See Projects, slots, and archiving.
Adding a slot can take you to a payment page. Everything you typed is saved and will be here when you come back, but any documents you attached will need attaching again.
Candor kicks off generation in the background. You can close the tab. The dashboard shows progress and a notification email arrives when your audience and its participants are both ready. It runs as one job: the audience first, then the participants, with no step in between for you to action. Generation typically takes 25-35 minutes; longer if you uploaded a lot of evidence.
Start with a free project, or walk through Candor with us first.