What Predictive Hero Images Does

Today, every subscriber on your list sees the same hero image. You might have a hunch that different audiences respond to different visuals — a lifestyle shot for younger subscribers, a product-focused image for older ones — but manually creating and assigning those variants is too time-consuming, so most teams don't do it.

Predictive Hero Images solves this. You provide 2–10 hero images, toggle on the feature, and Backstroke handles the rest. Our models analyze each image, scores it against your audience segments based on demographic profiles, and assigns the best-performing hero to each group. When the email sends through Klaviyo, each subscriber sees the hero predicted to drive the highest engagement for their profile.

The result: better click-through rates without any additional manual work beyond uploading your creative.

How It Works

Predictive Hero Images uses a proprietary machine learning model trained on real email engagement data. Here’s the high-level flow:

Image Analysis — Each hero image you upload is analyzed by an AI vision model, which converts the image into a numerical "fingerprint" that captures what's actually in it — scene type, colors, mood, product placement, text overlays, and more. This is far more nuanced than simple categories like "lifestyle" or "product shot."

Audience Matching — The model combines its understanding of each image with what it knows about your audience segments to predict which image is most likely to drive a click for each group.

Smart Assignment — For each audience segment, the model ranks all your hero images and assigns the winner. This means different segments can receive different heroes — all within a single campaign send.

Dynamic Delivery — Backstroke generates dynamic logic that gets embedded in your email. When Klaviyo sends the email, it evaluates each subscriber’s profile and serves the correct hero image at send time. No subscriber ever sees a broken or missing image.