Actions

The Actions tab is a run’s ranked playbook, and the quickest way to understand what to do about your use case. Each action is a strategic play that answers the goal and KPIs set in the use case definition, and the page is organized around the questions you’d actually ask before funding one: what’s the play, what’s it worth, who’s in it, what do I do, and why should I believe you. Beneath every action sit the segment tactics it composes — each contributing one segment’s slice of the audience and the per-segment evidence behind the play. Where the Discovery tab answers who your customers are, Actions answers what to do next.

Browsing actions

The tab is a master–detail layout under a Ranked Playbook header. The left rail lists every action in rank order, each as a compact row: its rank number, title, a value line (the action’s headline figure, such as ~$36.7M, with a short caption saying what the figure is), and a reach bar with the audience count and its share of the dataset. The rest of the screen is the reading pane, which always shows one action in full; the top-ranked action is selected when you land.

• Click any rail row to read that action.
• Use your keyboard’s left and right arrow keys to step through in rank order (the header shows the ← → switch actions hint).

Because the rail never leaves the screen, you can compare the whole playbook — and what each play is worth — at a glance while reading any one play. The tab is enabled in the use case sidebar once the run has actions to show.

The value headline

Each action opens with its value headline: a single large figure for what the play could be worth, over a caption sentence saying exactly what the figure is, with a see the math ↓ link into the calculation below. Depending on the play, the headline is expressed in dollars, customers, a rate move (a measured baseline and its after-play value, such as 27% → 29.4%), or units of the KPI it targets. Up to three supporting stats sit alongside it — the measured facts the headline is anchored to.

The headline is not a sentence the AI wrote: it is computed from your run’s measured data (real audience counts, stored feature statistics, lift columns) combined with a stated response assumption. Two rules keep it honest:

• Grounded vs. estimate. When a figure rests on measured data — a real per-order value from your lift columns, a counted audience — it stands on its own. When it rests on an assumption instead, it carries a visible estimate badge.
• One denominator. Every percentage on the surface is a share of the same total: all dataset rows, including unclassified customers. The rail, the hero, and the audience card never disagree about what 35.8% means.

Treat the headline as potential to rank by, not a forecast to plan against — the math card below shows precisely which parts are measured and which are assumed.

Tier chips

Under the headline, a row of chips summarizes the action’s standing in plain language instead of raw model scores:

Chip	What it tells you
Impact	How large the likely KPI movement is, relative to the other actions in this run.
Effort	How demanding the play is to execute — channels, budget, operational lift. Low effort is good news.
Confidence	How strongly the evidence supports the play’s reasoning.
Margin	Whether the play’s mechanics protect or dilute margin: Protective, Neutral, or Dilutive.

Each chip carries a one-line reason — hover (or focus) it to read why the tier was assigned. Impact and Effort tiers are relative to the run’s other actions, so a “High” means high within this playbook. There is no chip for strategic fit: the rank order itself is the fit judgment.

The Play

The Play card is the strategy in operational form: a one-sentence standfirst stating the play, over four fixed steps —

Step	What it answers
Who gets it	The audience in one line.
The offer	What you put in front of them, offer terms included.
The nudge	The trigger or message mechanic that carries the offer.
Who to exclude	Customers to suppress, so the play doesn’t discount people who’d act anyway.

If you brief a channel team on one card, it’s this one.

Where the number comes from

The math card shows the value headline’s arithmetic as an explicit equation — audience × assumed response rate × per-unit value — with each term in its own box and a provenance line stating where each term came from (measured, or assumed). The audience term is the play’s real, counted audience; the per-unit term names its source (for example, an average order value from your lift columns).

The assumed response rate is the one term the AI chose, and it’s adjustable: drag the what-if slider and the headline recomputes from the real terms at the rate you set. If you believe 3% instead of 8%, the card shows you what the play is worth at 3% — the claim is explorable, not take-it-or-leave-it.

Why we believe it

The evidence card argues the play in claims, not statistics. Each row is a short claim about your customers, a plain-language explanation grounded in your data, and — where the claim rests on a measured comparison — a ratio chip such as 3.2× vs dataset average. A small position marker shows each claim’s signal strength relative to the action’s strongest claim, which is labeled Strongest signal.

For analysts who want the underlying detail, an analyst toggle on the card reveals the raw feature-level rows — feature names and attribution values, grouped by the contributing segment tactic — without putting them in the marketer’s way.

Who’s in the audience

The audience card is where the action’s composition shows. Its heading carries the totals — the exact audience count and how many segments it spans — and each row beneath is one contributing segment tactic: the segment’s name, a one-sentence angle saying what this segment’s version of the play looks like and why it differs from its siblings, and that slice’s measured reach.

Select a row and it unfolds inline:

• How this segment’s version differs — the concrete variant steps for this slice.
• Why this segment responds — the segment-level evidence behind the variant.
• What this could look like — where available for your organization, an example creative rendered in its channel’s native form (an email, an SMS, a landing page). Offer values in examples are tactic recommendations, not approved offers.
• Explore in Discovery → — a link to the segment’s detail view when you want the persona and insights behind the group.

Reach counts are real audience sizes computed from the run’s labeled data, so you can sanity-check a play’s scale — and each segment’s share of it — before committing a team to it.

How you’ll know it worked

Every action closes with its measurement plan: the experiment that would verify the play worked, stated before you run it.

Element	What it tells you
Design	The experimental setup — typically a holdout percentage of the audience.
Success metric	The KPI the play is accountable to.
Readout	When to read the result, as a measurement window.

Committing to the measurement up front is what turns the value headline from a claim into a testable expectation.

Note. Actions are AI-generated guidance grounded in your data and use case definition. The computed figures keep the arithmetic honest, but your team decides what ships; review the play, its assumptions, and its evidence before committing budget.

Tip. A practical activation loop: pick the play, check its evidence and each segment’s angle, adjust the what-if slider to your own response assumption, then pull the audience via CSV export or warehouse delivery using the segment labels.

Next steps

• Understand the per-segment layer: Segment tactics.
• Get the audiences out of Neuralift: Exporting overview.