Albatross announced on 23 September that OfferUp had agreed to a long-term partnership to deploy Albatross’s adaptive discovery technology across its marketplace after a pilot lifted homepage feed engagement by 84%, according to its company announcement.
Key points
- Albatross says an A/B pilot increased homepage feed engagement by 84% and platform-wide listing views by 19%.
- The company says participating buyers discovered 3 million additional unique listings during the three-week test.
- OfferUp plans to replace recommendations based on popularity or recency with a feed that adapts to buyer behaviour, according to Albatross.
OfferUp’s pilot measures views and buyer contacts
Albatross said the initial A/B test increased listing views across OfferUp’s entire platform by 19%. It reported a 15% rise in total purchase intentions, which it defined as qualified buyers contacting sellers. The announcement also gave a 61% increase for homepage feed engagements, separately from its 84% figure for homepage feed engagement.
Each of those measures improved week over week during the test, according to Albatross. The company attributed the progression to its models learning from new interactions. Its account of the pilot also said engagements among buyers using its technology increased by 15%.
The buyers whose feeds ran on Albatross’s system discovered 3 million additional unique listings over three weeks, the company said. Kevin Kahn, its cofounder and chief executive, said those buyers also connected with 131,000 additional sellers. Those figures describe the group using the technology during the pilot rather than the whole OfferUp marketplace.
Nathan Garnett, OfferUp’s chief business officer, called the pilot results “exceptional” in the announcement and said the company was expanding the partnership. Scott Greenberg, OfferUp’s vice-president of advertising and partnerships, said the companies moved from the start of the project to a live experiment within weeks. He pointed to more discovered listings, seller contacts and conversations as the outcomes OfferUp wanted.
Albatross replaces OfferUp’s existing recommendation approach
Under the partnership, Albatross says OfferUp will replace recommendations driven by popular or newly uploaded listings. Its models instead update a buyer’s homepage feed during a session, using the buyer’s actions to select items. The planned deployment takes that approach across the marketplace following the initial test.
Albatross says the change can bring listings from less-established sellers to buyers whose activity indicates an interest in those items. Kahn described the division between the technologies this way: “LLMs understand words; perception models understand behavior.” The company says its system responds to changes in buyer behaviour as they occur.
OfferUp is a mobile marketplace for local buying and selling in the United States. Albatross describes marketplaces where listings are one-of-a-kind and thousands of new listings turn over each hour. Albatross said recommendation systems based on popularity or recency can favour sellers who already have visibility on such a marketplace.
Wallapop rollout precedes Albatross’s OfferUp deal
Albatross said it had previously rolled out its discovery technology with Wallapop in Spain, Italy and Portugal. The OfferUp agreement extends its marketplace work to a US company based in Bellevue, Washington. OfferUp was founded in 2011, according to the company announcement.
The Swiss company says its technology serves more than 50 million shoppers each month and handles more than 3 billion user interactions a month across more than 150 million products. Albatross says it is backed by Redalpine, MMC Ventures and Daphni.
