This article covers tips on how to provide actionable feedback for revising Variants while maintaining each message’s data-driven potential. These best practices apply to Variants you receive from Persado for Predictive Content, Full Experiments, and Dynamic Email campaigns.
✅ Prioritize Performance over Preference
Variants are built using data-driven components (Emotions, Narratives, Formatting) predicted to drive engagement. Avoid subjective edits (e.g., "I prefer 'Get' over 'Receive'"). Instead, rely on the data-driven components unless a strict Brand or Legal violation exists.
To learn more about the data components we test in our Variants, check our comprehensive Ontologies Guide—and reach out to your Persado team if you get any ideas for specific things you’d like to test!
✅ Provide Specifics and Contexts
Our AI is rule-based, so the more specific and contextual your feedback, the better results we can produce. Provide context about why and when certain content should be avoided. Even though we use AI, our team is heavily involved in ensuring content quality and catches things that fall outside of the AI’s capabilities.
Examples of helpful feedback:
“Open ASAP” is too aggressive for this campaign type – can we ban “ASAP” for any campaign other than “last chance” promos?
This headline is a claim. We need to avoid mentioning A, B or C for compliance reasons, but can say X, Y or Z.
We refer to our customers as clients. Please replace “customer” with “client” in Variants 4, 8, 12, 16.
Examples of less helpful feedback:
This feels off brand.
Too loud.
Down market.
Replace ‘last chance.’
Provide an ALT.
The more specific you can be about why an element of the Variant isn’t working, the higher the chance we can find a more effective solution and avoid similar language in the future.
Lastly, consider how feedback may apply to the control. The control message should be held to the same standard as Variants. If feedback applies to the control, the Persado team may ask for an updated version of the control.
✅ Maintain Variety of Language
Be sure to test a wide variety of language in the Variants. After editing, you want to make sure they don’t all sound too similar to one another. Testing a wider array of language and emotional elements gives the test more elasticity. More elastic tests means a higher chance of reaching statistical significance and opportunities to gain insights. If all the language sounds too similar, it’s harder to learn what worked and what didn’t.
The takeaway? Avoid "over-editing" or homogenizing the tone. A wider array of emotional and narrative elements allows us to better learn about your customer base and (if applicable) convert different customer segments.
✅ Aggregate Feedback
You can make edits to Variants yourself in Portal. However, if you need the assistance of the Persado team, please submit feedback in a single request to help reduce the number of rounds of revisions.
Tips for Experiments (Not Applicable to Predictive Content)
Approval in Exploration
If you’re conducting an Experiment, it’s important to only approve content in Exploration that you’d feel comfortable deploying to a broader audience in Broadcast. Language tested in Exploration is often used to create high-performing content in Broadcast.
Updates to Variants
Because we implement experimental design in our creative process for Experiments, updates made to a single Variant often affect 4+ other Variants. For instance, in the example below, if we were to change “Congrats {Full name}!” to “For you {Full name}!” in Variant 6, we would need to make the same change in Variant 7 to ensure clean testing data.