Run a multi-call sequence

Pick one of the built-in sequences, set the timing and memory options, split people into an A/B test if you want one, then launch.

A multi-call sequence campaign runs a scheduled series of calls to each person you choose. The caller remembers each conversation and builds on it, from a friendly first call to the real ask. This page covers the options that only a sequence has. What a sequence is, and the 21 built-in ones, are on Multi-call sequences explained.

The later calls of one sequence are never spaced out by your cooling-off rule, so a sequence is never cut in half. See Fair testing rules.

​

Before you start

• An attacker with a voice. See Create an attacker.
• The people to call, with consent recorded. See Add the people you will call.
• Enough minutes for every call of the sequence. A three-call sequence to ten people is thirty calls. See How minutes work.

​

Steps

1. Choose Multi-call sequence

   Open Red Team, Campaigns, New Campaign. Under Step 1 · Choose how to run it, pick Multi-call sequence.

   Under Pick a sequence, click one card. Each shows the name, the level, the number of calls and the spread. Browse the sequence library opens the full pages. A copy saved in your library opens here too, from its Launch button.

2. Read the plan

   A box titled Here is what ”…” will do shows every step, its phase and its job. Full details opens the sequence page. Below the plan, set the options.

3. Set the timing

   Option	What it does	Default
   Demo mode	”Call your test contact back-to-back, about 30 seconds between calls, ignoring the call window, so you can watch the whole sequence run in minutes instead of days. Best with a single test contact.”	Off
   Custom gap, in minutes	Shown when demo mode is off. The calls run as designed at the sequence’s own spread, or at the gap you type, for example 60	As designed

4. Choose memory and A/B options

   Option	What it does	Default
   Remember targets across campaigns	Resume from what an earlier campaign already learned about each person, so the attacker picks up the same backstory and rapport weeks later. Leave off to test each person fresh	Off
   A/B test two attack angles	Splits people evenly between Authority & urgency and Rapport & helpfulness. The report shows which one your people fell for more	Off
   A/B test three language models	Splits people across GPT-4.1 (control), Claude Haiku 4.5 and GPT-6 Luna (fast tier), same script and voice. The report shows each model’s reply time and how many people it convinced	Off

   Only one A/B test can be on at a time. Ticking one unticks the other. A person stays in the same arm on every call of the sequence.

5. Finish the wizard

   The four steps are the same as a single call. Under Who calls, the block The script for each call (N calls) lets you edit each step’s Opening line and Script for this call. A text step shows Message instead. Under Authorise and launch, a Call plan · N calls card lists every step. See Build a campaign and Authorise and launch.

​

What happens next

The first call to each person goes out on the schedule you set. Each later call waits the gap, then fires, inside your call window unless demo mode is on.

The results page shows every call, one row per call, not one row per person. See Read your results.

​

Problems

The second call never came

The gap has not passed, or the next call fell outside your call window and is waiting. Check the campaign page for the next planned time. Demo mode ignores the window.

Demo mode called a whole team back-to-back

Demo mode calls every selected person at the 30-second gap. Use it with one test contact.

I cannot tick both A/B tests

One at a time. Pick the question you want answered first: the angle, or the model.