Describe what you want
Plain language in, parsed intent out: category, price ceiling, condition floor, brand constraints. The parse is shown, so you can correct it.
MarginMap normalizes messy listings into canonical products, computes landed cost, and scores each offer against completed-sale comps. Buyer mode optimizes the purchase. Reseller mode optimizes the exit.
item + shipping + tax, per offer
recency-weighted median, IQR trimmed
source, timestamp, match confidence
missing data is labelled missing
Plain language in, parsed intent out: category, price ceiling, condition floor, brand constraints. The parse is shown, so you can correct it.
Offers resolve to one canonical variant, total to item plus shipping plus tax, and rank against a recency-weighted median of completed sales.
Buyers get a value score with its factor breakdown. Resellers get net proceeds, ROI, breakeven and a seven-stage pipeline.
Each of these is a design decision with a price consequence. Read the reasoning before you trust the output.
How MarginMap collapses messy marketplace titles into one canonical variant so price comparisons describe the same physical thing.
Item price plus shipping plus tax, computed per offer. How MarginMap handles unknown tax, free-shipping thresholds and cross-border charges without guessing.
Fair market value from completed sales using a recency-weighted median with outlier filtering — never from active listings.
Every score, median and recommendation in MarginMap links to its source records, retrieval timestamps and match confidence.
Expected profit, ROI, breakeven and liquidity computed from landed cost, sold comps, marketplace fees and holding time.
Category commissions, payment processing, promoted listings and shipping subsidies modelled per marketplace and per category.
Mapping inconsistent seller condition language onto a single grade ladder so comps and offers describe comparable goods.
The seven-stage pipeline from watchlist to sold, with alerts, saved evaluations and post-sale accuracy feedback.
Open the workspace, search in plain language, and open the evidence drawer on the first number you doubt.