ItemFits Research · methodology
How ItemFits data is counted
The definitions, sources, windows and exclusions behind every number in ItemFits research. If you are citing a figure, this page tells you exactly what it counts. For how the fit checker itself decides whether something fits, see our fit methodology and the formulas.
Definitions last changed · Publisher: ItemFits LLC
Principles
Rules every figure follows
Every number carries its own window
Fields began being recorded at different times, so a single report mixes windows: fit decisions since March 2026, locations since July 2026, item names since August 2026. Each figure shows its own range. Figures with different windows are never placed side by side as if they were comparable.
Partial months are labelled
A window that ends before the month is over is marked “partial”, and partial months are drawn lighter in charts. A partial quarter is compared with a full one only when the report says so.
Internal accounts are excluded
The three accounts ItemFits uses for testing are removed from fit decisions, conversations, questions and new accounts. Page views and request locations are not filtered; internal traffic is a negligible share of both.
Anonymous and registered users
Most fit checks are run without signing in. Fit decisions, conversations and questions include anonymous use. New accounts count only registered sign-ups.
Aggregates only
We publish counts and shares, never an individual conversation, name, email or precise location. Rankings leave out any item or vehicle too rare to be shown without identifying someone.
ItemFits users, not everyone
These figures describe the questions people brought to ItemFits. People usually check when they already suspect a problem, so a won’t-fit share describes our users’ questions, not how often furniture fits in general.
Quarterly reports do not change
A quarterly report is a snapshot of the data on the day it was published. Later data does not update it. If we find an error, we correct it and record the date, what changed and why on the report itself. The Fit Check Dataset page is different: it is updated as new data arrives.
Definitions
What each term means
Each statistic on a research page links to its definition here.
- Fit decision
- One outcome computed by the ItemFits fit checker for one item against one space or route, saved to our fit-check results record. It includes every outcome the checker returns: fits, won’t fit, and requests for a missing measurement.
- The record began in March 2026. Fit checks run before then are not counted anywhere.
- The three internal test accounts are excluded. Anonymous (signed-out) checks are included.
- A person who re-runs a check after changing a measurement produces a new fit decision. We count decisions, not people.
- Fits, tight fit, very tight fit, likely won’t fit, won’t fit
- The verdict family a fit decision belongs to. “Fits” means the item clears every constraint with room to spare. “Tight fit” and “very tight fit” clear with a small clearance the checker flags. “Likely fits” and “likely won’t fit” are hedged answers given when a decisive size is a catalog standard or an estimate rather than a measurement. “Won’t fit” means no orientation clears.
- The fits family = fits, tight fit, very tight fit, likely fits, fits tilted, and fits but won’t close (a container lid).
- The won’t-fit family = won’t fit and likely won’t fit.
- A request for a missing measurement, an unknown outcome and a dimension lookup are not verdicts.
- Definite answer
- A fit decision whose verdict is in the fits family or the won’t-fit family. Requests for a missing measurement, unknown outcomes and dimension lookups are not answers and are left out of every “share of answers” figure.Source: ItemFits fit-check records
- Clearance
- The room to spare at the tightest point of a fit, in the orientation the checker chose. “Under 2 inches” counts fits whose clearance is below 50.8 mm.
- Only verdicts whose name begins with “fits” (fits, tight, very tight, tilted, won’t close) that carry a recorded, non-negative clearance are counted. Hedged “likely fits” answers are left out.
- Asked for a measurement
- A fit decision where the checker stopped to ask for a missing size instead of assuming one. It is counted as a fit decision but never as an answer.Source: ItemFits fit-check records
- Conversation
- A chat session with at least one question typed into it, across the ItemFits web app, the Chrome and Edge extension and the Shopify storefront assistant.
- Sessions belonging to the three internal test accounts are excluded.
- Conversation records begin in January 2026.
- Question
- One message a person typed into a conversation. Replies and system messages are not counted.
- Internal test accounts are excluded.
- New account
- A registered sign-up with an email or Google login. Anonymous browser sessions, which the app creates for everyone, are not accounts.
- The three internal test accounts are excluded.
- Page view
- One of our own tracked page-visit events on itemfits.com. Visitors who block scripts are not counted, so this undercounts real visits.
- Page views are not filtered for internal accounts; internal browsing is a negligible share of the total.
- A monthly average over a window that ends in a partial month includes that partial month, which pulls the average down.
- Google search impression
- A time an itemfits.com page appeared in Google search results, as reported by Google Search Console.
- Our Search Console mirror begins in September 2025. Google reports data with a lag of a few days, so the latest days of a window may be missing.
- Google click
- A click from a Google search result to an itemfits.com page, as reported by Google Search Console.Source: Google Search Console · Recorded since
- Search reach
- Distinct search queries, distinct pages and average position over the last 28 days of Search Console data available when the figure was taken. It is a rolling window, so it is pinned to the date it was read.Source: Google Search Console
- Item checked
- A distinct normalized item name attached to a fit decision, such as “sectional sofa” or “kayak”. Spelling variants that normalize to the same name count once.
- This field began being recorded in August 2026; earlier fit decisions carry no item name.
- Item family
- A group of item names that describe the same kind of object (sofas and sectionals, mattresses and beds, TVs). Names are grouped by keyword; generic names such as “item” or “box of stuff” that match no family are left out rather than guessed.Source: ItemFits fit-check records · Recorded since
- Space type
- The kind of place a fit decision was about: a vehicle, a doorway, a route or hallway, a room, a box or container, an elevator, or stairs. Each checker route maps to exactly one space type; anything that maps to none is “other”.
- Vehicles = single and multi-item vehicle checks. Doorways = openings and several items through one door. Routes and hallways = passages, turns, hallways and multi-step routes. Rooms = room and inventory fits. Boxes and containers = containers, slots and storage. Elevators = elevator checks. Stairs = stair checks.
- Vehicle pick
- A vehicle chosen from the vehicle grid during a fit check. Rankings of named models count only picks of a specific model year range; generic body-style picks (“hatchback”) count toward body styles only.Source: ItemFits vehicle-grid picks
- Country and US state
- The country (and, in the US, the state) a request to the fit checker came from, taken from the request’s network location. We store the country and region only, never a precise location.
- Recorded since July 2026. Shares are shares of requests to the fit checker, which include follow-up turns, so they are not a count of people.
- Internal test traffic is not removed from this count; it is a negligible share.
- Language
- The languages the ItemFits interface is published in. This counts the product’s languages, not the languages people typed in.Source: ItemFits product
- Shareable result page
- A fit answer saved with a permanent link that can be sent to someone else, with the calculation attached.Source: ItemFits shareable result pages · Recorded since
- Vehicle catalog
- The measured vehicle records the calculator solves against: models, generations (model-year ranges) and cargo configurations (a cargo area with a specific seat state).Source: ItemFits vehicle catalog
- Engineering and validation counts
- Counts read from our code repository and validation programs at publication time: hand-verified known-answer cases, automated tests, and checks that run on every change. They describe how the checker is tested, not how often it is right in the field.Source: ItemFits validation programs, ItemFits code repository
- Repeat usage
- People (a signed-in account, or else a browser session) with two or more fit decisions in the window.Source: ItemFits fit-check records
- Merchant fit check
- A fit check run by a shopper inside a retailer’s store through the ItemFits Shopify app. Not yet published: too few stores are live for store-level figures to mean anything.Source: ItemFits product
Data sources
Where the numbers come from
Search figures come from Google Search Console. Everything else comes from ItemFits’ own production records, read with a query that changes nothing.
| Source | What it provides |
|---|---|
| ItemFits fit-check records | Fit decisions, verdicts, clearances, space types, item names |
| ItemFits conversation records | Conversations and questions |
| ItemFits account records | Registered accounts |
| ItemFits page-visit events | Page views |
| ItemFits request locations | Country and US state of requests |
| ItemFits vehicle-grid picks | Vehicle models and body styles picked |
| ItemFits shareable result pages | Answers saved with a permanent link |
| Google Search Console | Search impressions, clicks, queries, pages and average position |
Reproducible vs. manual
Four kinds of statement
Not every statement in a report is a measurement. Each figure falls into one of four classes, and only the first can be regenerated automatically.
A · Reproducible from production
Counted by a committed, read-only query against ItemFits production records (fit decisions, conversations, questions, accounts, item families, space types, vehicle picks, locations, result pages). Anyone with access to the query and the records gets the same number for the same window. They are still reviewed by a person before publication.
B · Verified by hand
Counted once by a person at publication from a system with no standing query: the vehicle catalog, the validation programs and the code repository. They are stated as of the publication month.
C · External
Reported by a third party. Google search impressions, clicks and positions come from Google Search Console, which revises its own recent data; partner credits are checked on the partner’s own page.
D · Editorial
Statements about how the product works (the solvers are deterministic, no verdict is written by a person, the languages the interface is published in). They are facts about ItemFits, not measurements of its users, and are checked against the product before each edition.
Corrections
How we correct a published figure
A quarterly report is a record of the data on the day it was published. When newer data arrives, the report does not change; the Fit Check Dataset page carries the current figures instead.
If a published figure turns out to be wrong, we correct it on the report and add a visible note with the date, what changed and why. Inconsistencies we have found but not corrected are listed under “Data notes” on the report itself.
Found something that does not add up? Email support@itemfits.com.