Pulse is where you set up and launch a study. You decide what you're testing, who you're testing it on, and at what scale — then run it and watch the results come in.
Pick the mode that matches your question:
Each persona attribute shows how many people in the panel currently hold what you picked. Worth watching, because aiming at one of your own attributes never creates new people — a value only exists for someone already on the panel. If your choice reaches fewer people than your sample size asks for, the Pulse runs with the people it found rather than inventing the rest, and Pulse tells you before you launch.
Persona attributes belong to the brand that declared them. Switch brands and the attribute pickers clear and reload, so you are only ever offered the attributes of the brand you are working in.
Use Sampling distribution beside the sample-size control to choose how a Pulse selects its audience. Census weighted is the default: eligible census records are selected in proportion to their population weights, so more populous categories have a greater chance of appearing.
Equal distribution aims for equal representation of the available categories within each demographic dimension: region, age band, gender, race/ethnicity, income, education, and household type. For example, an unrestricted 100-person sample aims for about 25 people in each of the four census regions. With targeting, balance applies only within the eligible audience. Each respondent has equal influence in equal-mode aggregate scores and shares, including generated respondents. This is a category-comparison sample; it does not represent population shares.
Balance is approximate. A small sample, correlated demographic combinations, missing categories, persona-generation failures, or a limited custom-attribute audience can prevent exact equality. Equal distribution balances realized persona demographics, including education and household values enriched beyond the census seed. It balances each dimension, rather than giving every census record or every demographic combination the same share. Brand custom attributes still restrict eligibility; they are not additional balancing dimensions. National generation can use all four regions, while state geographies map to their Census region. Original seed geography stays the eligibility filter.
Choose Census weighted or Equal distribution before running any Pulse study type.
Panel shows the saved sampling distribution with each Pulse. Study timelines and frozen-versus-varied comparisons track changes to the saved mode. A rerun inherits its source's mode. Choose Resample audience to switch modes; Reuse source personas keeps the original respondents and distribution. Competitor comparisons and automatic iterations retain the mode too. Existing Pulses without a saved mode are shown as Census weighted.
{
"sampling_mode": "equal_distribution"
}
// Omit sampling_mode for the census_weighted default on new Pulses.
// On a resampling rerun, omit it to inherit the source Pulse's mode.Under Demographics in persona context, uncheck a built-in demographic or an attribute under Custom attributes to omit it from this Pulse's persona narrative and follow-up interviews. All start enabled. Custom controls include numeric attributes and attributes that cannot be targeted; analysis-only attributes already stay out of model context.
Both built-in and custom attributes can be omitted from one Pulse without changing the audience or brand schema.
See Brand schema overrides for custom attribute setup, rerun controls, and API/MCP request examples, or Plan and run Pulses with chat for the supported attribute keys and planning guidance.
When you run a study, Pulse hands you off to Panel, where results stream in and you can explore them as they arrive. Anything you set up here — context, competitors, audience — is reflected in the results.
The planning agent beside the form on wide screens — and above the direct controls on smaller screens — can turn a goal into a complete Pulse configuration, look up relevant brand and capacity context, and explain its assumptions. Each message continues an authenticated planning thread for the active brand. The most recent thread returns when you reload or revisit Pulse, and it remains scoped to the signed-in user and active brand. Agent revisions update the form automatically; use Revert to restore the immediately previous reviewed draft. Its recommendations and tool actions use the same versioned guidance references published in Plan and run Pulses with chat.
The authenticated brand thread and direct controls collaborate on one reviewed Pulse draft.
At 1,050 pixels and below, planning comes first and the direct configuration controls follow it.
The agent cannot silently launch a study. Creation always pauses for your approval of the exact draft revision. Once handled, the card becomes read-only Approved or Rejected history.
The current approval stays beside the message composer while the planning transcript scrolls independently above it.
The message box and Send button fit below the page heading. The conversation scrolls within the chat panel.
On smaller screens chat appears above the form and fits the visible page height.
This page is safe to send to a colleague who has not signed up. Signed out, /pulse is a public page: instead of a sign-in prompt it plays the loop walkthrough — one study going from configuration to verdict, then a demographic filter, a persona interview, and a rerun — so they can see what a Pulse produces before deciding anything.
What a signed-out visitor sees at /pulse. Run your own pulse in less than a minute. opens account creation in place; once they finish, the same URL becomes the configuration form described above.