The agent never even considered Temu.
In four journeys Temu was not among the options the agent weighed and its site was never requested; no retailer was recommended.
Open-market search Four runs · open-market search · one research-agent configuration · no purchase made · 2026-09-20 · all 4 journeys of this decision
In four recorded research runs for a set of kitchen storage containers, the agent searched Amazon, Walmart and Target, and in one run Wayfair, Crate & Barrel, The Container Store and Lowe’s. Temu did not appear among the options in any run and temu.com received no request. Two runs ended with no retailer recommended; two ran out of page reads.
- The customer’s requirements
- Five matching containers with lids, glass or BPA-free plastic, 1 to 2 litres each, delivered to Texas before a dinner in ten days; from each retailer’s own pages, the delivered total, the delivery promise, the return window and how quickly a refund is paid.
- The information used
- Target: a 90-day return window read on target.com. Walmart: answered three of four visits with a bot check. Amazon: listings from third-party sellers. Nothing about Temu, because nothing about Temu was read.
- The resulting recommendation
- None. Two journeys ended with Target, Walmart and Amazon unresolved on delivery or refund speed and OXO, Rubbermaid and Pyrex recorded as rejected; two journeys exhausted their page reads first.
The full evidence, every option weighed and every source, follows below and on each journey page.
The customer
A customer in Texas who needs five matching containers before a dinner in ten days, with typed conditions on the delivered total, the delivery promise, the return window and the speed of a refund, all to be read on the retailer’s own pages. Two of the four customers remembered a discount-marketplace order that arrived after 26 days and a refund that took six weeks and three chats; the other two carried nothing.
What the agent researched
Between 6 and 13 page reads per journey. amazon.com, walmart.com and target.com in every journey; wayfair.com, crateandbarrel.com, containerstore.com and lowes.com in one. Walmart answered three of the four visits with a bot check; Crate & Barrel and Lowe’s refused the one request each received. No journey searched for Temu, named Temu, or requested temu.com.
What it believed
That Target offers a 90-day return window, read on target.com. That the Amazon listings it found were sold by third-party sellers, which left the return and refund conditions unconfirmed. That the branded options it priced, OXO, Rubbermaid and Pyrex, did not fit; the record marks them rejected without a readable reason we can quote.
The decision
No retailer, in the two journeys that reached a decision. Target, Walmart and Amazon stayed unresolved on delivery time or refund speed. The other two journeys ran out of page reads before any option was resolved.
What was never established
Whether Temu could have met the four conditions. The agent did not look. This is the same shape as the Amazon finding in the vacuum study: an option that never enters the comparison cannot lose it or win it.
Fresh versus persistent memory
The two customers who remembered a slow discount-marketplace order did not consider Temu; neither did the two who remembered nothing. The record therefore does not show that the memory steered the agent away from Temu; it shows that in this search Temu was not where the agent went.
What this means for the brand
For a retailer whose case rests on price, the first question is not whether the price wins but whether the agent ever sees it. In these four journeys, it did not. Whether that holds for other customers, other requirements or other agents is not established here.
What could be tested
The same customer and requirements with Temu named as an option, to see whether the delivered total, delivery promise and refund timing can be established from Temu’s own pages and how the recommendation changes. The baseline is these four journeys.
The journeys
- 1Customer missionTemu decision — same customer with relevant memory
- set of five matching containers with lids
- each container 1-2 litres (glass or BPA-free plastic)
- delivered total at most the customer's budget ceiling
- delivery within 8 days, as stated on retailer's own pages
- 2Evidence16 searches, 8 pages: amazon.com, walmart.com, target.com
- 3BeliefNothing established about the brand
- 4DecisionNever entered the comparisonNONEAfter 24 page/search actions, no offer could be confirmed against all required facts using the retailer's own pages. Walmart.com and Amazon.com blocked every direct page …
- 1Customer missionTemu decision — same customer with relevant memory
- delivered total for the set of five ≤ $45 USD, as stated by the retailer
- delivery within 8 days of ordering, as promised on the retailer's own pages
- return window of at least 30 days, as stated by the retailer
- refund paid within 10 days of return receipt, as stated by the retailer (preference)
- 2Evidence13 searches, 11 pages: walmart.com, target.com, help.target.com, amazon.com
- 3BeliefNothing established about the brand
- 4DecisionNever entered the comparisonNONEThe best physical match found was the NutriChef 5-Piece Superior Glass Food Storage Containers Set sold via Target.com's Target Plus marketplace: all five containers are …
- 1Customer missionTemu decision — customer with no relevant memory
- 2Evidence11 searches, 13 pages: amazon.com, wayfair.com, walmart.com, target.com +4
- 3BeliefNothing established about the brand
- 4DecisionRan out of page reads before decidingNo decision: the action budget ran out after 13 page reads, before a readable Temu page.
- 1Customer missionTemu decision — customer with no relevant memory
- 2Evidence17 searches, 6 pages: walmart.com, target.com, amazon.com
- 3BeliefNothing established about the brand
- 4DecisionRan out of page reads before decidingNo decision: the action budget ran out after 6 page reads, before a readable Temu page.
Limits
- The mission required delivery to Texas within ten days, and two of the four customers remembered a slow discount-marketplace order. Both favour faster retailers. The next release runs the same task with no deadline and no memory, to see whether Temu enters the comparison then.
- Four journeys on one day, two per memory condition, one agent configuration.
- Open-market search: the agent chose which retailers to research; nothing in the brief named or excluded Temu.
- Bot checks are what the harness saw on 20 September 2026; a different client may be treated differently.
- Every journey run with Claude on the live web. Every source, belief and decision published.
Reproduce this experiment → the frozen customer mission, ready to run in your own agent.
Investigated with Accord. Accord is the decision-intelligence system behind these experiments. It reconstructs how customer agents evaluate services, identifies what remains unresolved, and tests what changes when the enterprise provides better information or a more authoritative answer. For brands and CX teams →