Pearfit Book Brand Pilot
Size intelligence for apparel brands · Closet-matched fit for shoppers

Lower returns for brands.
Better-fitting clothes for every shopper.

Size is one of the main reasons online apparel comes back. Pearfit turns customer reviews into numerical fit values for each size, and a separate comparison engine measures a garment against clothes the shopper already owns.

Launch 6-Week Pilot Investor inquiry
For brands: fit numbers for every size For shoppers: compared with a garment they own Review text stays on the machine that scraped it
studiomercer.com/products/washed-oxford Sample silhouette
Target Size M (58 cm chest)
Your Uniqlo (M · 56.5 cm)

Woven Shirts · 195 gsm organic cotton

Washed Everyday Oxford Shirt

$88

Fit values Per size, from reviews Runs small, true to size, and runs large
One row Per size of a SKU Review count, skew, verdict, shrinkage
Blank When a measurement is missing Nothing is filled in to make a size look closer
Shared Fit table for every machine Review text stays in the local scrape file
01. Two-Sided Value Engine

Built to protect brand margins and
eliminate shopper sizing guesswork.

01 · For Apparel Businesses · Margin Recovery

Read how each size actually fits, before you guess from a handful of reviews.

Shoppers who are unsure often buy two sizes and send one back. Pearfit counts what reviewers say about each size: runs small, true to size, runs large, shrinkage, and sentiment.

  • One verdict per size: Review count, fit skew, and the share of reviews in each direction.
  • Shrinkage and opacity, counted: Mentions are tallied. The wording of the reviews is not copied into the shared table.
  • The same table everywhere: A scrape on any machine writes that SKU’s fit numbers to the shared database.
Fit numbers, per size The ROI calculator below uses your own assumptions
02 · For Consumers · Closet-Anchored Confidence

Compare a garment with one the shopper already wears.

The comparison engine measures a candidate size against an owned garment. The preview on this page is a sample of that idea, using made-up shirts and a score computed only for the illustration.

  • Sample overlay: Chest, shoulder, and length for one oxford and four reference shirts.
  • Sample review split: The size matrix below uses a fixed set of review percentages.
  • No body scan: The sample asks the shopper to pick a reference garment.
Sample preview The match score on this page is not the comparison engine
03 · Network Effect · Compounding Fit Graph

Each scrape adds that SKU’s fit numbers to the shared table.

A second scrape of the same SKU replaces its size rows. Full review text stays in the local file on the machine that scraped it.

Shared fit rows commercial_skus and sku_size_metrics
02. Sample fit preview

A sample of a size preview, using made-up review splits.

The percentages below are sample data for one heavyweight tee. Pick a usual size to see how this page turns that sample split into a size note. The score is computed here for the illustration.

Per-Size Review Distribution Runs small True to size Runs large

Sample note for L–XL: shoulders wide (38% of mentions) · body length long (24%)

03. Proprietary Fit Architecture

Proprietary ingestion. Deterministic geometry.

The scraper reads a product page and classifies reviews with deterministic rules. Fit numbers are stored per size. A separate comparison engine measures a candidate against a garment the shopper already owns. Missing measurements stay blank.

  1. 01. Ingest

    Read the reviews on a product page

    The scraper loads the page and classifies each review by size, fit direction, shrinkage, and sentiment.

  2. 02. Store

    Keep the numbers

    Each size gets a row: review count, fit skew, verdict, and shrinkage mentions. Full review text stays on the machine that scraped it.

  3. 03. Compare

    Measure against an owned garment

    The comparison engine scores the sizes it has measurements for. A missing centimeter stays empty.

  4. 04. Replace

    A later scrape updates that SKU

    Scraping the same SKU again replaces its size rows in the shared table. It does not add a second copy.

Missing measurements stay blank. The comparison engine leaves an unknown centimeter empty. The match score in the sample above is a separate illustration on this page.
04. Business Case & Unit Economics

Model your annual return savings with Pearfit.

Enter your own monthly orders and return rates. The reduction slider is your assumption. This calculator does not report a measured Pearfit result.

Projected Annual P&L Impact
Fit returns avoided / yr —
Processing fees saved / yr —
The “Bracketing Tax” Breakdown (38% multi-size carts) —
— —
Without Pearfit (Multi-Size Cart) Shopper buys M + L to test fit at home → the extra size is returned, and that unit is out of stock while it travels back.
With Pearfit (Closet-Matched) Shopper sees exact cm overlay → orders 1 size with conviction, preserving outbound & return margin.
Total Annual Margin & Retained GMV Impact (incl. unrefunded orders) —

05. Get Started with Pearfit

Start a 6-week pilot on one category

  1. Scrape one category: The scraper reads reviews and stores numerical fit values for each size.
  2. Read the shared table: Those fit rows are in the same database from any machine that runs the scraper.
  3. Decide what to look at next: A storefront size advisor is a later step. This starts with the fit table.

Sends this inquiry to customerservice@pearfit.app.

06. Frequently Asked Questions

Answers for brands and partners

What data does Pearfit need to launch a pilot?

Product page URLs. The scraper reads the reviews on those pages and stores numerical fit values per size. Full review text stays on the machine that scraped it. Size charts are not part of the shared fit table.

How do shoppers tell Pearfit what clothes they already own?

The sample on this page lets you pick one of four reference shirts. Comparing a real catalog with clothes a shopper owns is the separate comparison engine. This page does not store that choice.

How does Pearfit’s proprietary engine work without LLMs?

The scraper classifies review text with deterministic rules and stores the resulting numbers. When a measurement is missing, it stays blank. The sample score on this page is an illustration, separate from the comparison engine.

How do you measure ROI during the 6-week pilot?

The calculator on this page is a model you fill in. A pilot would compare your own fit-return rate before and after, using your orders.

How is shopper privacy handled?

The sample on this page stores nothing. The scraper keeps review text on the machine that scraped it. The shared database receives the fit numbers.