Hardest Part

Accurately predicting the clearing price and margin potential of unstandardized wholesale lots based on messy listing data, as a single overconfident algorithm execution directly destroys the buyer's capital.

Min Viable Scope

Build a read-only pricing engine that outputs a maximum bid recommendation for a single asset class like off-lease fleet vehicles directly into a buyer's workflow. Deliberately leave out autonomous bid execution and multi-category support.

Cold Start Problem

Algorithms require historical win and loss data to calibrate pricing models, which you lack on day one. Break this by scraping historical auction results and shadowing a mid-sized buyer's manual bids for 30 days to train the initial heuristics.

Time To First Value

1-2 weeks of model calibration to ensure the agent bid recommendations align with target margins before activating real-time execution

Data Moat Available

true

Technical Difficulty

High

Wedge

Begin with independent used auto dealers bidding on structured digital wholesale platforms. This niche offers high daily transaction volume, standardized vehicle condition reports, and immediate pain around securing inventory against large franchised networks. Expand outward by adapting the evaluation engine to heavy equipment auctions, then unstructured retail liquidation pallets, and finally industrial salvage.

Timing

Multimodal models now reliably parse messy, non-standard lot descriptions and unstructured condition photos to predict defect rates. Concurrent expansion of API access across major digital auction platforms enables instant programmatic execution.

Why This ICP

Mid-market wholesale liquidators and dealers operate on high volume and thin margins where fractional improvements in acquisition costs directly drive net profit. They face immediate daily inventory constraints, making them early adopters for scalable procurement tools.

Size Of Prize

~45000 mid-sized wholesale dealers and liquidators in the US employ procurement teams, spending roughly ~$60000 annually per firm on dedicated bidding labor, yielding an addressable market of ~$2.7B.

Gap Narrative

Wholesale buyers manually evaluate and bid on hundreds of daily auction lots using static spreadsheets and human intuition. They miss profitable inventory due to limited processing bandwidth and overbid on dead stock due to inaccurate spot-pricing. They require autonomous agents that instantly analyze lot condition reports, calculate projected resale margins, and execute bids across multiple platforms simultaneously.

Defensibility

Defensibility compounds through proprietary pricing data and margin feedback loops. Every winning bid and subsequent retail sale logs the actual realized margin, continuously calibrating the bidding model for that specific dealer's market. A new entrant using off-the-shelf models lacks this historical transaction ground-truth and structurally overbids or misses margins.

Why This Thesis

Bidding is an execution-heavy, time-bound loop that requires continuous monitoring and instant decision-making. An Agent directly performs the evaluation and transaction steps without human intervention, structurally matching the required velocity of real-time auctions.

SAM

~$400M-600M across deregulated US and European power markets

SOM

~$15M-30M

TAM

~15k global wholesale energy trading entities × ~$100k/yr ≈ $1.5B

Growth Rate

~18-25%/yr, driven by renewable energy intermittency increasing short-term price volatility and requiring faster execution speeds

Paid Comparable Spend

~$150k-300k/yr per desk on quantitative analysts, 24/7 manual execution trader shifts, and legacy ETRM automation modules

Kill Thresholds

• Autonomous execution rate < 20% after 30 days of live deployment
• Bid latency > 200ms during market price spikes
• Zero conversions to $100k annual contracts after 3 successful pilots
• Custom ETRM integration costs > $25k per desk
• Compliance review cycle exceeds 90 days

Leading Metrics

• Time-to-first-automated-trade (hours)
• Autonomous execution rate (% of bids placed without human override)
• Bid-to-market latency during peak volatility (milliseconds)
• Manual kill switch activation rate per week
• ETRM API integration setup time (days)

What Proves Right

Trading desks deploy the agent on live deregulated power markets and allow it to execute real-time bids autonomously without human override. The platform captures spread during high-volatility events, effectively replacing overnight manual trader shifts. Customers transition from pilot agreements to $100k annual contracts within 90 days after verifying the agent's win rate outpaces manual execution.

What Proves Wrong

Risk and compliance officers block autonomous execution, restricting the product to a read-only dashboard that fails to displace human traders. The agent executes bids with excessive latency during volatility windows or breaches predefined risk limits, triggering manual kill switches. Integration with legacy ETRM systems demands custom engineering per client, destroying margins and stalling deployment.