A trial user and a long-time customer can still get the same campaign.
which customer behaviors or differences should genuinely change the next message.
You’re testing subject lines. The real questions are who gets what, and when.
We look at what customers do in the product, find the differences that should change what they get, and test changes to the path itself, like a branch, the timing or the offer. Then we check whether customers did more of what you wanted.
Optimize
Use evidence to decide what is worth changing
BaselineJourney works
ObservedBehavior shows an opportunity
ChangeRevise the path for the right group
MeasureKeep, change, or remove
The basics work. The next improvement is harder to see.
Optimize begins with a working baseline. We are looking for an opportunity, not diagnosing a broken system.
The stable flows have only been tested at the surface.
a structural change that could teach us something useful about customer behavior.
The product can see a high-value action, but nothing happens when someone takes it.
whether that action should change the message, route someone to a person or move them into a different path.
The reporting says the program drove revenue, but nobody knows what would have happened anyway.
whether there are enough customers for a useful holdout or comparison, and what the best valid option is if there aren’t.
Observe, decide, measure.
The answer isn’t always another email.
Pick the behavior worth improving
We start with the customer action that matters, not a backlog of campaign ideas.
Find a difference worth acting on
We look for customer behavior, value, where someone is in the journey or what they’ve told you they want that should genuinely change what happens next.
Design a meaningful change
That may be a different branch, timing, message, channel, offer, suppression or human follow-up.
Decide how we will judge it
Before launch, we define the outcome, the comparison and what we’ll watch to make sure the change isn’t hurting anything, such as unsubscribes, complaints or support volume.
Keep, change or stop
If the evidence doesn’t support the extra complexity, we don’t keep it just because we already built it.
Not every change needs an A/B test, and not every audience has enough volume to support one.
When volume is too low for a decision-grade test, we do not pretend statistical certainty exists. We use the strongest evidence the situation can actually support.
If opens go up and the customer doesn’t move, the test didn’t do its job.
We measure as close as practical to the customer behavior the program exists to change. Opens and clicks can help explain what happened.
The system still needs to see the behavior we want to act on.
If feature use, teammate invites, plan limits or other useful product signals never reach the CRM or email platform, we’ll identify what the automation needs and work with the technical owner on the requirement.
What this work can produce
- A prioritized map of where a test is worth running
- Behavior-based audience or journey logic
- A written test plan
- A comparison or holdout design where valid
- The next recommendation based on what the test taught us
Measurement plan
Connect the change to customer behavior
- Behavior
- Define the customer action we want to influence
- Signal
- Choose the evidence that tells us the state changed
- Intervention
- Change one useful part of the lifecycle path
- Downstream check
- Measure whether customer behavior moved
- Decision
- Keep, change, or remove
If you don’t trust what is already running, start with Fix. See Fix.
Turn a vague audience idea into an operational rule.
See how behavior-based segmentation turns customer state into explicit inclusion, exclusion, and downstream logic.
Know it could be doing more, but not sure what to test next?
Tell us what is working and what you want customers to do more often.