New21 benchmarks for auto-parts stores, with examples

Sells only what fits.Finds the sales you never saw.

For auto-parts stores. Selva checks every fit against the maker’s own document, asks the one question that settles the rest, and reports back in kept dollars.

Get the free check See how it works

01Sells only what fits

Shoppers now type the whole job, not a keyword. Google says the average search in its AI Mode is triple the length of a traditional query. Keyword search keeps the part and drops the rest: the underlined words below, which are usually what decides the sale.

Google, May 19, 2026 (source). The requests below are examples.

  • Fit“Front pads for a 2019 Mustang GT, the Performance Pack car”
  • Unsure“Rotors for my 2018 F-150. Not sure which brakes it has.”
  • Build“A 2.5-inch lift for my JL Rubicon, and shocks that work with it”
  • Legal“Cold air intake for my 2018 F-150 5.0. I live in California.”
  • True cost“A front bumper for my 2022 Bronco, shipped to 19087, all in”
  • Electronics“HDMI cable for 4K at 120 Hz, through my receiver”
  • Whole job“Everything I need to do the front brakes on my Gladiator”
  • Delivery“Wheels that arrive before Saturday’s trail ride”
  • Returns“Can I return a tuner once it’s been flashed?”
  • Symptom“Squeal from the front when I brake, 2021 Gladiator”
  • Agents“An AI assistant, asking for pads that fit a 2019 GT”
  • Electronics“A charger that fast-charges my 16-inch MacBook Pro”

One understanding

Every surface gives the same answer. Selva reads a shopper once, answers the same way everywhere, then tells you what happened.

Search

Reads the whole ask. Every need in the sentence, not just the keywords.

Brake pads for a 2019 Mustang GT with the six-piston Brembos, under $200
Vehicle2019 Mustang GTBrakessix-piston BremboPriceunder $200Positionfront
  • Ceramic street pads$89Fits your brakes
  • Low-dust performance pads$134Fits your brakes
  • Ceramic street pads$72Held back: made for the base brakes

Chat

Asks once. Only what your data can’t already tell it.

Will this cover fit my 2022 Ram 1500?

Which bed, the 5′7″ or the 6′4″?

5′7″

It fits. I left out the 6′4″ cover and liner, because bed length has to match.

Product page

Says only what your data supports.

Folding bed coverShort bed

Fits your 2022 Ram 1500, 5′7″ bedWhy this?

Sales desk

Your team sees the same answer, and why.

Answer
Fits the 5′7″ bed
Source
Fitment table, row R-1182
Rule
Bed length must match, v3

The same answer the shopper saw, word for word.

Assistant feeds

The facts assistants read, kept in step.

product
Folding bed cover
fits
2022 Ram 1500, 5′7″ bed
not_for
6′4″ bed
source
fitment R-1182
rule
bed length, v3

The proof

Four ways to say “fits”. One shopper model, a floor and a ceiling, and Selva loses the least at both.

Fit losses per 1,000 sessionsFloor Ceiling
97%vs 89%of shoppers reach a correct cart with Selva, against the year, make and model badge
1.4vs 4.0wrong parts per 100 orders. Each one is a return cycle that averages 9 to 10 days from start to refund.
15%less timeper correct cart once wrong carts count, and 72% less at the ceiling. The question adds seconds; a wrong part costs days.
Why “slower” was the wrong word

The clock only ran for winners. The model’s time to a confident cart averages successful sessions only. The badge store’s fast time leaves out the 11% of shoppers it fails at the floor, and 39% at the ceiling. Selva’s time includes the hard cases it gets right.

Fast because it guesses. At a badge store, a shopper who doesn’t know the deciding fact guesses, and the badge says “fits” either way. Browsing with no vehicle set reaches a cart in 9.5 seconds, a cart that is often wrong.

It times the fit task, not the visit. The seconds are an expert’s pace through the fit task: taps, reading, and checking the vehicle. They leave out the rest of a real visit, like browsing and weighing options, which is why no real visit looks like 46 seconds.

FloorCeiling
Badge65 s219 s
Selva55 s61 s

Time per correct cart: every session’s time, with each wrong or failed cart counted at the 300-second cap, divided by the correct carts. The cap is the shortest a run is allowed, the most generous setting for the badge. Successful sessions enter at their modeled geometric-mean time. The real number comes from timed runs on real sites, not from the model.

How we print numbers
A floor and a ceilingNever one number. The ceiling is unanchored, and it is never the selling number.
Costs beside winsThe deciding question adds seconds, and 1.5% of the shoppers it asks leave. Both are counted above.
Modeled until measuredFigures on this page are modeled. Your own figures come from your pages and your 90 days of orders.
No store namedYour numbers come to you in private, with reproduction steps and a window to dispute any finding.

Benchmark B21 · Agent shoppability

When an agent shops for your customer, it’s only as right as the answer your store gives it.

In Adobe’s July 2026 retail data, visits sent by AI assistants were up 62% in a year and converted 60% better than other traffic. More and more, the assistant does the shopping too. It reads your pages faster than any shopper, and believes them more. If a page says “fits” at year, make and model, the agent buys on it.

Adobe Analytics, July 2026 (source).

The same request, sent by an agent The gears question from the top of this page

“4.10 gears for a 2019 F-150 5.0, going to 35s”

Your store today

  1. Reads your pagesBoth gear sets say “Fits 2019 F-150”. Each is true for one axle, not both.
  2. Picks the cheaper one8.8-in. set · $349

Wrong axle shipped. A return that takes 9 to 10 days, and the install kit never sold.

With Selva

  1. Asks your storeIt depends on the rear axle. Count the cover bolts: 12 means the 9.75.
  2. Asks the shopper“12.”
  3. Gets the answer, with its proofFits: the 9.75-in. set, $389, from the maker’s document. The job also needs the install kit, $279.

Right gears, and the kit. $668, kept.

Example. Same agent, same request. The only difference is your store’s answer.
3 of 32 of 3A question passes only when three separate runs all get it right. A single lucky answer never prints alone.
  • Run 1Right part, right fit claim
  • Run 2Right part, right fit claimAdds “arrives Thursday”. Your store never said so.
  • Run 3Right part, right fit claim

Passes. All three runs are right.Fails. A price, stock or date your store doesn’t state sinks the whole question.

Fails a run
  • A wrong part, or a wrong fit claim
  • A price, stock or date your store doesn’t state
  • Ignoring a word like “SuperCrew” that decides the fit
Still passes
  • Sending the shopper to your page
  • Asking the one question that decides the fit

438 cards46% hard6 platforms

How Selva answers an agent

  1. A model reads.It turns the request into facts: the truck, the axle, the job. Models read and choose between options code gives them. They never decide what fits.
  2. Code decides.Every fit is checked against the maker’s document, by code. The same verdict feeds your product page, your cart and every agent.
  3. Code asks, once.Only when one fact would change the answer, and with how to tell. The questions are written and checked by technicians, never improvised.

Nothing your store didn’t say: no invented price, stock or delivery date. Sending the shopper to your page is a fine answer. A wrong fit claim never is.

How it reaches your money

As exposure, not yet counted: your AI-referred sessions, times the share of questions agents get wrong, times your conversion and average order. Your AI-referred share comes from your own analytics. Selva scores this benchmark through its own agent tools, and never asks a public assistant about your store.

This page is written for agents too: runselva.com/llms.txt

02Finds the sales you never saw

See what you never knew to look for.

Selling is half the job. The other half is knowing what is happening in your store. Your reports count what comes back: returns, tickets, searches with no results. Some of the costliest fit losses send nothing back at all.

The sale nobody sees

Fit losses per 1,000 sessions at a store that says “fits” at year, make and model.

Not in any report
$142

What your reports see · $50Fit returns. Each one comes back, with a label and a reason code.Returns, traced · $50Each one tied to exactly what the shopper was told before buying.

What they never see · $92Hidden fits: the maker lists the part, your store says no, and the shopper leaves. No return, no ticket, no review.Hidden fits, counted · $92Every time your store hides a part the maker lists for that vehicle, counted per 1,000 sessions and priced in your margin.

Counting it is the report’s job. Winning the sale back is the answer’s job, measured in the proof.

How this is counted

Modeled across 300,000 shopper sessions at the floor, where fit returns are anchored at 4% of orders; there, hidden fits are the largest single loss. At the unanchored ceiling, wrong-part returns are larger. Hidden fits rest on one estimate, 8% of maker-listed fits hidden, which the free check measures on your own catalog from your public pages.

Twenty-one questions about your store. How many could you answer today?

Swipe right if you could answer it for your store this afternoon, left if you couldn’t. Scroll sideways with two fingers to browse without answering. Turn a card over to see how Selva checks it.

Hard on purpose. On average, half of each benchmark’s cards sit in the two hardest of four difficulty bands.

Built to be hardShare of each benchmark’s cards in the two hardest of four difficulty bands
3 in 10cards are generated fresh at scoring time, so nobody building Selva has seen them.
1 in 7cards are shared anchors that every store on a platform is scored on, so results compare across stores and versions.
0verdicts graded by a language model. Answer rules run as code, against the maker’s document, a regulator’s record or your own page.
Beforea card runs, the run is written down and timestamped. The cards it draws are fixed by that record, never picked afterward.

Your report

The sales you never saw, counted. From what shoppers actually did, not what a dashboard guessed.

Kept dollars, net of returnsILLUSTRATIVE
With SelvaShoppers who never saw Selva
  • 61 front-pad returns last quarter were four-piston pads sold to six-piston cars.Changes the question Selva asks
  • Your wheel filter hid 6 black 17-inch wheels their maker lists for the JL.Changes what your selector shows
  • Your intake page says “50-state legal”. Its California approval covers the 3.5 EcoBoost, not the 5.0.Changes what you can ship to California

Examples of what your report finds, not results.

Once Selva serves your shoppers, 20 of the 21 go live.Same definitions, counted from what shoppers actually did. Accessibility stays a release check.
Demand in its own wordsNo-result searches split four ways: words you don’t map, parts you don’t carry, stock you’re out of, and fit nobody could confirm.
The shoppers behind every numberOpen any figure and read the searches behind it, in the shopper’s words, with the sample size beside it.
A new question, asked of the pastEvery search and answer is kept, so a question you think of today runs over everything since Selva started.

The record

Answers you can see through. Each one carries the claim, its source, the rule that applied and what was held back.

Selva means forest. The glasswing butterfly lives in the forests between Mexico and Panama, and its wings reflect about two percent of the light that reaches them, so you see straight through. We wanted answers like that.

Siddique, Gomard and Hölscher, Nature Communications, April 2015 (source)

{  "answer": "fits",  "item": "Folding bed cover",  "for": "2022 Ram 1500, 5′7″ bed",  "source": { "table": "fitment", "row": "R-1182" },  "rule": {    "name": "bed_length_must_match",    "version": 3,    "approved_by": "fitment lead"  },  "asked": { "bed_length": "5′7″", "times": 1 },  "held_back": [    { "item": "6′4″ cover", "rule": "bed_length" },    { "item": "6′4″ liner", "rule": "bed_length" }  ],  "saved_with": "order",  "store": "example"}
Why this?Example
Answer
Folding bed cover fits your 2022 Ram 1500, 5′7″ bed
Source
Your fitment table, row R-1182
Rule
Bed length must match, version 3, approved by your fitment lead
Asked
Bed length, once. The answer: 5′7″
Held back
The 6′4″ cover and liner, by the same rule
Saved
With the order, so any return traces back to what the shopper was told

Every answer leaves one of these. Selva says only what your data supports.

Free check

Start with the free check. It measures your hidden fits and false fits from your public pages, with intervals, before you share any data.

Write to Ken See the benchmarks
It reads, and touches nothingYour public pages only, at a human pace. The check and the replay change nothing on your site.
Your rates, measuredYour hidden-fit rate, which decides most of the money, and your false-fit rate, from public pages. Your fit-return share comes later, from your own returns.
Off with one switchEvery step can be switched off, and your pages are exactly as before.
On the site you run todayYour systems stay the system of record. Switch Selva off and your pages are exactly as before.
If anything failsShoppers see your normal store, at the same speed.
Yours to leaveExport everything in open formats, any day. Selva is never priced per question.

Write to Ken

Tell me your platforms and the part page that worries you most. The check comes back as a list you keep.

ken@runselva.com
Ken NavarroFounder, Selva, Wayne, Pennsylvania
267-265-9284