Data: Deeper Dive

The Content Insights section of Portal is populated once we receive results on how our Variants have performed in market (whether your are using Predictive Content and/or Full Experiments). Keep in mind that you must send multiple Variants per deployment (at least 2 + your control message) in order for insights to populate. We recommend deploying 3 to 4 Variants + your control message when you are getting started with Predictive Content so our machine can begin gathering insights about your audience as quickly as possible.

Every Persado Exploration involves the fitting of a logistic regression to obtain metrics to explain and interpret how different elements of your message are performing (we sometimes call these the "gene-level results"). Element contribution, or how particular elements of your message contribute to overall success, is one of those metrics. This is calculated by taking the deviance explained by a gene (GD) and dividing that by the model's null deviance (ND) minus the residual deviance (RD), i.e., GD / (ND - RD). This percentage is then published in the Element Contribution and Language Performance chart in Portal.

There are a number of factors that impact business performance, such as the economy, merchandising decisions, seasonality, competitors entering the market, quality of leads, etc.

Some of these external factors may cause a downturn in metrics like click-through rates. For example, if a customer recently doubled their email audience file because there was a big promotion on site, the overall click-through rates may decline over time (even if the total business revenues are on the rise).

While Persado cannot account for all of these external factors, Persado can still optimize against your baseline/control, providing a positive uplift and helping mitigate risk of underperformance.

“Incremental revenue” is the additional revenue generated due to Persado’s optimized messages. We calculate this as: (Persado revenue per impression - control revenue per impression) x total Persado impressions.

If Persado Variants perform worse than the control and have a negative revenue lift, incremental revenue will be negative. This is a normal occurrence when testing into new areas where we don’t know what to expect or what will resonate with consumers. Sometimes, a few rounds of testing are required before finding Variants that perform well.

For content that has been rolled out to 100% of the audience, with no control audience, incremental revenue is calculated using an assumed lift. The assumed lift is taken from observed lifts of the same or similar campaigns run in the past.

Elasticity helps us answer the question: “do words matter here?” in our tests. High elasticity means that the elements we tested drove a material difference on customer behavior. Put another way, high elasticity means that there’s a big difference between the losing and winning message.

The more elasticity, the better! Keeping an open mind about what elements your team is comfortable testing may help improve elasticity and ultimately campaign performance.

For example, let’s say a promo bar test had a Variant with the lowest CTR of 5% and Variant with the highest CTR of 15%. The elasticity is (15%/5%-1)*100 = 200% elasticity.

To re-test a campaign, you are welcome to submit:

Persado strongly recommends against using a previous Persado Variant (whether it's best predicted, best observed, etc.) as your control in the Broadcast phase. Not only will this result in redundant learning, as the message has already been explored in market, but you won’t be able to see the true distinction in uplift between your original control and your ultimate champion message.

Instead, we recommend finding equally well-performing values from a new Exploration. This means we are expanding your brand-specific knowledge.

We have found no evidence that Persado's Variants are at a higher risk of encountering deliverability issues such as being marked as spam. In an in-depth analysis for one customer, for example, we found no significant difference in average aggregate delivery rate (this customer experienced a 98.87% delivery rate vs. Persado having a 98.86% delivery).

If you are encountering concerns over email deliverability, please reach out to your Persado CSM to see how they can assist. Factors such as your sender reputation and engagement levels are more likely to be the culprits of deliverability issues, and in instances where Persado content is overperforming your BAU creative, using Persado on a greater percentage of your campaigns may help address these problems.

A correlation of 100% is perfectly positive (as length increases, the rate goes up), a correlation of -100% is perfectly negative (as length decreases, the rate goes up), and a correlation of 0% indicates no relationship. Correlations above 50% or -50% should be considered "strong" and correlations between 30% and -30% should be considered "weak."

Put simply: