How to Dissect a Message for Better Testing

By BRADY EVAN WALKER

FACT: A Message is the Sum of its Parts

This makes A/B-style testing, which compares messages as wholes, just slightly better than useless.

At Persado, we look at marketing messages the way one would a math problem. Each word or phrase is something that can be measured and scored. In other words, we’ve codified language in a way that a computer can read it as data.

Unlike your standard A/B test, we don’t just say which message performs better — we also tell you why.  This insight has far more short- and long-term benefit because it opens avenues for improved optimization.

For example, any common mode of testing would look at the two messages below as immutable wholes.

You’re one lucky duck! Storewide savings until Friday at midnight!

Good job, valued customer! Click for your exclusive storewide savings!

At Persado, we see a collection of parts that can be tested, moved, and modified. Without this perspective, you won’t know if “You’re one lucky duck, valued customer!” or “Click for savings until Friday at midnight” would garner more opens, clicks, or conversions.

To make a mathematical analogy, an A/B test gives you simple sums without the underlying equation.

Your A/B test might show that:

A = 10

B = 14

But that offers no long-term benefit and no way of knowing what to write to get to 15, 20, 30, or beyond.

At Persado, the sums are less interesting than the equation. With our methodology, we see:

A = 2 x 4 + 2

B = 30 / 2 – 1

Knowing the equation, we can understand exactly what needs to change to move the sum in a positive, predictable direction. “Oh, the strong opener was divided by half by the second phrase. What if we moved this ‘times 4’ from message A to the same spot in B? What would happen then? And why is this a -1? Can we find a +5 in the database to replace it?”

You can’t manage what you can’t measure. Having only a vague understanding of message design means you’ll never know how to make it better. There’s no way of knowing if an A/B test is only telling you the difference between a crappy message and a mediocre one.

THE BREAKDOWN

Persado’s question is simple:

How does the wording of a marketing message affect its performance in a specific audience and context?

To get to the answer, we organized the variables by categorizing grammatical and syntactical writing components. We break them down into “genes,” something our platform can read as data.

These genes are:

Emotive Genes

Language that is emotionally motivating, broken down into 15 distinct emotive genes representing the key marketing emotions

Persado was born from the idea that we could systematize emotional language to make it testable and readable as data. We pared the most common marketing emotional motivators to five broad categories encompassing 15 distinct emotions.

We map specific words, phrases, and images to these core motivators. We then analyze campaign performance to discover the elements that most inspire action in each audience. Persado’s machine-learning algorithms can then incorporate greater emotional intelligence for new campaigns. (Come by next week when we describe our approach to emotion in greater detail.)

In the context of what Persado tests, “Click here for a 50% discount sitewide!” would be considered emotive-free because it doesn’t include the type of language we consider to be emotionally motivating. You can argue that a 50% discount has an emotional valence. You’d probably be correct. But by isolating components of marketing grammar, we can test with greater precision. The above message is actually a combo of two other genes types — more on that below.

As an example of the yield clients can get by including emotives in their subject lines, look at the predicted lifts from adding an intro eliciting Achievement:

subject lines 4

To test emotionally motivating language for yourself, click here for free tools to get you started.

Descriptive Genes

Language that describes the product or offer

The descriptive gene is the meat and potatoes of any given marketing message. It’s usually provided by the client and looks a lot like the second part of our emotive-free example above. With the offer in hand, our platform and team are able to construct testable variations to find the exact wording to get the greatest lift in opens, clicks, and conversions.

Functional Genes

Language that describes the action desired from the audience, i.e.., CTA

The functional gene is the almighty Call to Action. This gene is, of course, incredibly important for Facebook campaigns, landing pages, email bodies, and banner ads. After emotionally resonant language, the CTA is the second-most important aspect of any marketing message.

Subject lines use the functional genes less often than other types of digital marketing. Subject lines are themselves big buttons to be clicked, so telling your audience to click can be redundant, though sometimes it’s a good emphasis.

message_genes_3

Formatting Genes

This includes capitalization, punctuation, bold, italics, and symbols

By testing formatting, we can inquire into the effect of everything from ALL CAPS, exclamation points versus periods versus no punctuation, and all manner of emojis and special characters. These create their own distinct emotional experience that our system can track over time.

We’ve found that, generally, including formatting genes at the beginning of a message will have a positive effect on lifts.

message_gene_4

Positioning Genes

The order in which the language genes are presented

Another reason we break messages into constituent elements is to learn how structure affects performance. Just as in a math problem, the order of operations can change the outcome.

So the below messages are likely to perform in market differently:

message_gene_5

RESULTS

After nearly 7,000 experiments, we’ve calculated the percentage by which each gene contributes to the response rate of any given message:

gene_table

Marketers love to obsess over the CTA, and for good reason. But they’re missing a big opportunity when they overlook optimizing the emotional experience, cutting as much as 61% from the optimal response rate.

By creating this marketing grammar, we’ve turned short digital ad copy into a math problem and scientific experiment, the ultimate complement to creativity.