How to Use Predictive Demographic Attributes to Personalize Your Marketing [New Feature]
When it comes to marketing your products, wouldn’t it be great to have as much data as possible about your target customers? It would certainly help you take greater ownership of your marketing and send more relevant messages to your target audience.
But how can it help you answer tricky questions, like: “Are some people more likely than others to prefer certain types or categories of products?” It’s hard to identify such traits across all of your individual customers. That’s where predictive attributes come in.
Predictive attributes use customer data to learn and then predict characteristics and behaviors. Gender, for example, is a predictive attribute and predicted gender is now available within Klaviyo.
The gender prediction attribute does exactly what it implies: it predicts the gender of your customers.
You can use the predictive gender feature to target gender-specific products and content in more relevant ways to your audience. You can also use it to build segments, and filter and split your email automations accordingly.
How does Klaviyo’s gender prediction algorithm work?
Our gender prediction algorithm uses a customer’s first name along with census data to predict whether the name is likely male, female, or uncertain.
To do this, we used lists of first names where individuals had self-identified their gender and compared those directly to verified census data. We found many common names to be in line with expectations. For example, the name Steven was 99.6 percent male on the census and self-identified as male 99.9 percent of the time. Susan was female 99.8 percent of the time on the census and self-identified 99.3 percent of the time. But, names with uncertain genders generated larger errors between census and self-identified data. For those names, we erred on the side of caution and labeled them as uncertain.
Gender targeting isn’t foolproof because using a first name to predict someone’s gender isn’t a perfect science. Some people may not identify with the gender that their name traditionally implies. Others won’t be interested in the things their gender typically indicates.
Since gender prediction and gender targeting aren’t perfect, it’s important to make sure your messaging and your content don’t completely assume a customer is definitely male or female. For example, you can create a gender-targeted campaign that simply reorders content based on gender. The campaign can still show both product lines, so if you’ve somehow mistargeted your customer, they’ll still receive content that’s likely to be of interest.
You can also collect your customer’s gender with a sign up form and store it as a custom property so you can keep that data and can use it. Klaviyo’s predicted gender is a Predictive Analytics property and can coexist with any custom properties you’ve already set. It’s always better to use the first-hand information your customers give you when possible, but the gender prediction feature can be a handy tool to help you fill in the gaps.
Let’s take a look at how one brand has used the new gender prediction feature to personalize its marketing.
How James Black uses gender attributes to personalize its marketing
James Black sells clothing for both kids and adults. The brand previously sent their back-to-school promotion highlighting their new kid’s clothing, and the campaign was successful. It outperformed last year’s results and attracted many first-time customers. This year, James Black is trying something new. They’re doing a cross-sell campaign that promotes the adult styles to the same audience using the “Treat Yourself” message.
Since the brand carries both men’s and women’s clothing, they want to use the gender attribute to promote the right content to the right people. For women, they want to highlight women’s products first and men’s second and for men vice versa. With this strategy, customers will first see products they’re more likely to prefer and eventually purchase, which will ultimately lead to a better conversion rate and more revenue.
Because no data science algorithm or method is perfect, James Black still has a number of email recipients whose gender is unpredictable and labeled as uncertain. They know they have more men’s styles than women’s and they generally have more male customers, so James Black decided to send their customers who have an uncertain gender prediction the men’s variation email that highlights men’s products first.
Sending your customers emails with the most relevant content possible helps them to stay engaged with your brand and increases the likelihood that they’ll make a purchase. By using the gender predictor and reordering your content so your customers see products or images that they’re more likely to prefer, you add an additional layer of personalization to your messaging that helps build longer-lasting relationships with your customers.Back to Blog Home