Customer distillation · adaptive retrieval

Customer signals,
distilled into search that adapts.

Seekly converts anonymous shopping history, rating behaviour, live preferences, and intent changes into safe retrieval signals—then ranks and explains every result in real time.

Technical score Run evaluation to calculate
Hit Rate @ 10 Baseline —
Mean reciprocal rank Baseline —
Turns to convert Baseline —

Live customer journey

Shopping conversation

Vercel demo notice: starting a conversation may take 20–30 seconds while the 50,000-product catalog initializes. The locally hosted dashboard does not have this serverless delay.

Seekly shopping agent Online · responds with ranked products
Select a case

Ready when you are.

Choose a public test case to watch Seekly clarify needs, rank products, and explain each result.

Evaluator controls the customer's next reply

Technical evidence

Performance by scenario

No evaluation running
Scenario Cases Hit @ 10 MRR MTTC Sector score
Run an evaluation to see scenario metrics.

Technical score = 50% Hit Rate@10 + 30% MRR + 20% conversational efficiency.