The challenge
Homebox had been running TikTok Sales campaigns in Saudi Arabia on TikTok’s native "All Products" catalog feed — a setup that was generating results, but leaving one critical question unanswered: could a curated, performance-led product set do more with the same budget?
Feeding TikTok’s algorithm the entire catalog indiscriminately meant every SKU, regardless of how well it actually converted, competed for the same delivery. Homebox needed rigorous proof rather than assumptions — specifically, whether focusing delivery on a smarter subset of the catalog could move the needle on ROAS, CPA, and conversion rate without sacrificing the purchase volume the "All Products" campaign had already established. The challenge was designing a test tight enough to answer that question cleanly, isolating product selection as the one true variable.
THE SOLUTION
To settle the question, ROI Hunter designed a controlled two-way split test that isolated the product set as the single variable, holding targeting and bidding parameters equal across both campaigns. The test ran over four weeks, with matched daily budgets across both arms.
The structure was kept as clean as possible. A business-as-usual (BAU) campaign continued running TikTok Sales with the native All Products feed as the baseline. In parallel, a Test campaign ran on ROI Hunter’s curated Bestsellers product set — a segment which classifies products into meaningful, performance-based clusters using live signals from Homebox’s own tracking rather than static, manually maintained rules. The Bestsellers segment surfaced only the most engaging and converting products from the Homebox catalog, continuously updated as performance data changed.
The underlying hypothesis was straightforward: presenting shoppers with a curated selection of best-selling, highly relevant products would sharpen audience engagement, lower media costs, and improve purchase efficiency — without restricting scale. Performance was tracked directly through the TikTok dashboard throughout the test window, giving the team a clear, apples-to-apples read on how each arm performed across the funnel, from impressions and clicks down to purchases and revenue.
THE RESULTS
Four weeks in, the Test arm delivered a clear verdict. ROAS rose 64% compared to the Control arm, confirming that the curated Bestsellers selection drove meaningfully stronger revenue efficiency for the same spend.
The gains continued down the funnel. CPA fell 19%, and the click-to-conversion rate improved 37%, a clear signal that the traffic the Bestsellers feed converted better than the "All Products" catalog.
The Bestsellers campaign achieved this while running on comparable spend and with fewer impressions and clicks than BAU — proof that the efficiency gains came from smarter product selection, not simply more exposure.
These efficiency improvements did not come at the expense of volume: the Bestsellers campaign also drove 24% more purchases than BAU, putting to rest any concern that narrowing the product set would limit scale.
The test gave Homebox a data-backed foundation to restructure its TikTok Sales strategy around a curated, performance-led product set — and a repeatable testing framework for validating catalog segmentation decisions in future campaigns across its other GCC markets.
