Feed Experimentation
Test your product data before you scale

Productsup Feed Experimentation enables you to run controlled split tests on product titles, descriptions, images, and more directly inside your feed workflow. Gain confidence in your product content’s performance before scaling to boost ROAS.

Hero image Feed experiments (1)

Trusted by industry leaders worldwide

  • puma-black
  • sephora.png
  • sika.svg
  • loreal.svg
  • HP.svg
  • Eileenfisherlogo
  • thule.svg
  • henkel-black
  • hellyhansen.png
  • beiersdorf-black.svg

The impact of Feed Experimentation for feed teams

Brands spend time and budget rewriting titles, testing AI content, and swapping images with no reliable way to know what actually moved ROAS. Feed Experimentation changes that.
  • icon - bars graph.png

    Make smarter, data-backed feed decisions

    Split any product subset 50/50, test two content variants, and measure the impact in your analytics or advertising platform before committing catalog-wide. Every change becomes a decision, not a gamble.

  • icon-magic-hand

    Eliminate risk from AI-generated content

    AI-generated descriptions, highlights, and use-case tags are powerful, but rolling them out catalog-wide without evidence is a business risk. Test content on a controlled SKU subset first, then scale what wins.

  • icon-screen-shopping-cart.svg

    Gain autonomy from IT

    As Feed Experimentation is embedded inside the Productsup rule box interface, setting up a test takes minutes. There’s no need for development tickets, duplicated exports, or context-switching.

How Feed Experimentation works in practice

Set up your first test in 5 easy steps

Data View with Rule box conditions highlighted

Where to start

Open Data View and click Rule box conditions in the top ribbon. Select the Export stage to scope your experiment to a specific channel.

Step 1 — Define your product subset

Create a new condition inside Rulebox Conditions. Set attribute filters — category contains Shoes, price greater than 50, brand contains Nike — to define exactly which products enter the experiment.

Enabling the Feed Experimentation toggle

Step 2 — Enable the Feed Experimentation toggle

Activate the Feed Experimentation toggle. The platform splits matching products 50/50 into Variant A and Variant B using deterministic hashing — the same SKU always lands in the same group, every refresh.

Applying different rule logic to each variant

Step 3 — Apply different rule logic

Drop rule boxes into each variant. Variant A uses Set Column if not Empty to apply AI-generated titles. Variant B is left empty as the control group, keeping the original title unchanged.

Configuring UTM tracking

Step 3b — Configure UTM tracking

Open Tracking Configuration. Distinguish each variant either via UTM parameters on the product URL (the platform auto-appends _a or _b) or via a custom label tag — then track performance in GA4, Google Ads, or Meta.

Scope and preview the test

Step 4 — Scope and preview the test

Add the Rulebox Condition to the attribute where you want the variant logic to run — for example, the title, description, or image_link attribute. And switch to Full View and trigger a Refresh to preview the experiment output safely, without re-importing data or exporting to any channel.

Measuring results and scaling the winner

Step 5 — Measure and scale the winner

Variant A shows +55% CTR and a 4.2x ROAS. Once you have a clear winner, roll it out to the full catalog in a single click.

1 / 7 2 / 7 3 / 7 4 / 7 5 / 7 6 / 7 7 / 7
← Previous
Next →

What types of tests can Feed Experimentation run?

Any attribute in a product feed is fair game, but here are the four tests that drive the most impact.

1 Title format variations

Compare brand-name-prepended titles against descriptive titles in a specific category. Measure the CTR delta in Google Shopping before committing to a full catalog rewrite.

2 AI-generated content validation

Run AI-enriched descriptions on Variant A and the original copy on Variant B. Collect four weeks of conversion data on a high-traffic category before any catalog-wide AI deployment.

3 Image type testing

Test lifestyle imagery against plain product shots on a defined product subset. Let performance data decide instead of guesswork.

4 Custom label and attribution strategies

Test different custom label values or UTM structures across a product tier or price band. Measure the downstream impact on bidding and ROAS without touching the full catalog.

What makes Productsup Feed Experimentation truly unique

  • Embedded in workflows

    The feature lives inside an interface that most Productsup platform users use daily (Rulebox Conditions). No need to learn a new tool or context-switch.

  • Stable SKU assignment

    With deterministic hashing, the same product stays in the same variant group across every feed refresh. Teams don’t need to worry about measurement noise over multi-week tests.

  • Native AI content validation

    Feed Experimentation is integrated with Productsup AI Enrich, so there’s a complete AI-to-performance loop. Test AI-enriched attributes and mappings inside a controlled experiment before catalog-wide rollout.

Ready to see more return from your ad?

See Feed Experimentation in action with a live walkthrough of the Productsup platform tailored to your business.

Find more details in the Productsup Help Center →

Book a personalized demo

By requesting a free demo, you agree to Productsup’s Privacy Policy.