photosell

Identify cognitive barriers to household asset liquefaction.

• • Audited resale market entry friction
• • Observed household inventory patterns
• • Mapped valuation latency

• • Assumed the barrier was shipping, it was actually valuation effort

• • If valuation takes >30 seconds, the asset remains dormant

• • Pivoted to a Computer Vision-first identification model

Map CV pipeline to real-time market pricing indices.

• • Created image classification flows
• • Mapped pricing API integrations

• • Initial maps were too optimistic about object condition detection

• • Condition is subjective; system must allow for tiered valuation

• • Added multi-state condition assessment to the CV loop

Test identification precision and conversion speed.

• • Ran tests with 500 household items
• • Measured "snap-to-price" velocity

• • Users found the categorization too generic

• • Users value market-specific metadata over generic labels

• • Tuned AI to pull channel-specific attributes (e.g. eBay/Reverb)

Build the CV pipeline and pricing indexer.

• • CV classification engine
• • Pricing aggregation layer
• • Snap-to-sell UI

• • Over-engineered the listing automation features early on

• • Accuracy in valuation is the hook; ease of listing is the closer

• • Focused on sub-second classification and price confidence

Calculate liquidation velocity and reclaimed equity.

• • Identification precision
• • Valuation accuracy
• • Time-to-list

• • Metrics didn't account for items that were identified but not sold

• • Confidence in price is the primary driver of sell-intent

• • Introduced a "price confidence score" for every asset