Campaigns, A/B tests & holdouts

A campaign is a one-off send to an audience — a newsletter, a sale announcement, a product launch. Where journeys react to individual behaviour over time, campaigns go out to everyone at once (or on a schedule).

Creating a campaign

In the dashboard: Campaigns → New campaign. Pick the channel, the audience (a list or a live segment), the content (write it inline or start from a template), and send now or schedule. Via API: POST /v1/campaigns.

A/B testing, properly

Add up to several variants — different subject lines, different content, even different send times. You choose:

• Split size — e.g. test on 20% of the audience (10% see A, 10% see B).
• Winning metric — opens, clicks, or conversions (a tracked goal event).
• Decision window — how long to wait before declaring a winner.

After the window closes, the platform sends the winning variant to the remaining 80% automatically. No spreadsheet, no manual second send.

Holdout groups

A holdout is a random slice of the audience that deliberately receives nothing. Why would you pay for an audience and not message them? Because it is the only honest way to measure incremental effect: if the messaged group converts at 4.1% and the holdout at 3.9%, your campaign added 0.2 points — not 4.1. Configure a holdout percentage on any campaign and the ROI report shows messaged vs. holdout conversion side by side.

Guard rails

Campaign sends respect frequency caps and quiet hours, suppression lists, and consent per channel — a contact who unsubscribed from email but not SMS is excluded from the email audience automatically. You can cancel an in-flight campaign with POST /v1/campaigns/:id/cancel; already-queued messages are recalled where the channel allows it.