Overview
What a tone experiment does
- The tone of a payment reminder changes whether people pay — a tone experiment A/B tests two tones against real collection traffic so you stop guessing.
- It does not send anything on its own. It rides on your existing dunning workflows: the reminders they already send become the experiment's traffic.
- Each reminder is delivered in one of two tones, payments score which tone won, and you keep the voice that collects more.
Before you start
What you need first
- At least two tone profiles defined (the two voices you want to compare).
- Running dunning workflows that send overdue reminders — the experiment only accrues data while those workflows fire.
- Open or overdue invoices for those workflows to act on.
Step 1 — Define your tones
- Go to Receivables Controls → Tone Profiles.
- Create the voices you want to test, for example Neutral, Firm, and Empathetic.
- Each profile is an AI tone the engine applies to your reminder templates; you can also set a team default.
Step 2 — Create the experiment
- Go to Receivables Controls → Tone Experiments → New Experiment.
- Name it, then choose a Variant A profile and a Variant B profile (they must be different).
- Set the traffic split — for example 50% / 50% — to decide how reminders are divided between the two tones.
- It saves as a Draft until you start it.
Step 3 — Start it
- Open the experiment with View details, then choose Start experiment.
- Only one experiment can run at a time. If another is already running, pause it first — you'll be told which one is blocking.
Step 4 — Let it run
- From here it's hands-off. Each time a dunning workflow sends an overdue reminder, the customer is placed in Variant A or B by your traffic split.
- Assignment is sticky: a given customer always stays in the same variant, so their experience stays consistent.
- That customer's reminder is rewritten in the assigned variant's tone before it goes out.
How scoring works
Step 5 — Conversions attribute automatically
- When a customer pays, the payment is credited as a conversion for whichever variant they were in — no manual tracking.
- Each variant's conversion rate is its conversions divided by its assignments.
Step 6 — Read the results
- View details shows Assignments, Conversions, and Conversion rate per variant, plus a variant-performance chart.
- The variant with the higher conversion rate is the tone that collects better.
Step 7 — Roll out the winner
- Pause or Mark completed when a winner is clear.
- Set the winning profile as your default in Tone Profiles so future reminders use it.
Good to know
Tips and limits
- Results accumulate only as fast as reminders send and customers pay — treat it as a background test that bakes over time, not an instant readout.
- Run one experiment at a time and give it enough volume before calling a winner.
- Because assignment is sticky, a customer only ever experiences one of the two tones.