Skip to main content
Case Study · Shopping AI

Lumen AI hit $8.20 effective RPU by matching ad supply to commercial intent

Lumen AI runs on pure shopping intent — users come to ask "what should I buy" questions. That makes CPMs high, but only if the ad layer matches the query tightly. Lumen’s integration of Surfacedd produced $8.20 effective monthly RPU, roughly six times the cross-vertical average Surfacedd publishes.

The problem

Shopping assistants have a ceiling problem: if the product recommendations are sponsored by default, users distrust the advice and churn. If they are purely organic, the app has no revenue.

Existing affiliate networks paid poorly and required aggressive prompt-engineering to trigger eligible links, which biased the recommendations.

Commission-based monetization was not stackable across multiple product categories because of conflicting affiliate policies.

What they did

Integrated Surfacedd’s LlamaIndex wrapper to separate retrieved product context from sponsored recommendations at the query engine layer.

UI renders organic recommendations inline with the AI response; sponsored items render in a distinct, labeled block below — never mixed.

Enabled vertical-specific CPM scenarios: high scenario for fashion and electronics, mid for home goods, low for consumables. The SDK routes queries to the right pool automatically.

Kept affiliate revenue from organic links running in parallel with Surfacedd sponsored surfaces — both monetization layers compose cleanly.

Results

$8.20
Effective RPU
per MAU per month; ~6x Surfacedd network mid-scenario
$164,000
Monthly ad revenue
20K MAU × $8.20
9.3%
Click-through rate
vs 1.8% category baseline for display ads in AI chat
+4 points
Recommendation trust score
quarterly user survey; separation of organic and sponsored drove the lift
+$42K/mo
Stackable affiliate revenue
organic affiliate layer, preserved
Our users asked for better shopping advice. Sponsored answers were the first ad format that actually helped — because we labeled them clearly, kept them out of the organic recommendations, and matched them to real purchase intent. Trust went up, not down.
Hana Okafor, Head of Product, Lumen AI

Key takeaways

  • Shopping AI can hit 5-10x the cross-vertical RPU mean if ad supply matches query-level commercial intent.
  • Clear visual separation of organic and sponsored content is the trust-preservation moat.
  • Surfacedd stacks with affiliate revenue rather than cannibalizing it.

Frequently asked questions

Why did Lumen’s RPU run so far above the network average?

Two compounding factors. Shopping queries carry the highest commercial intent on the Surfacedd network, so eligible CPMs are near the high-scenario range. And the clear separation of organic and sponsored blocks produced a CTR roughly 5x the AI chat baseline.

Does Lumen still run affiliate links?

Yes. Organic AI recommendations carry standard affiliate links; sponsored blocks are Surfacedd-native. The two streams compose and both produce meaningful revenue.

How much did the LlamaIndex integration cost in engineering time?

The wrapper integration was about 1.5 engineering days. Most of the time was spent designing the visual separation between organic and sponsored blocks — not the data layer.

Want numbers like Lumen AI’s?

Estimate your app’s potential in the revenue calculator, then start the integration.

Try the calculator →See the SDK