The same lift kit sells to five different buyers. Your catalog is written for none of them.

A four-inch suspension lift for a 2021 F-250 gets bought by five people who share nothing except the part number.

The weekend trail rider wants to know if it clears 37s and whether the ride quality on pavement goes to hell. The commercial upfitter wants to know if it voids the chassis warranty, what it does to the payload rating, and whether the DOT-relevant headlight aim is still adjustable. The shop foreman quoting the job wants book time, whether the pitman arm comes out with a puller or a press, and what else he has to order today so the truck is not in the bay on Friday. The fleet manager wants total installed cost across nine trucks and whether it changes the PM interval. The restorer is not in this market at all, but he is in the next one over, and he needs to know if the finish is period-correct.

Now open the product page. It says the kit is engineered for superior off-road performance, includes premium components, and is designed to provide a comfortable ride on and off the road. It answered nobody. And if you pulled that description from a shared aftermarket data network, it answered nobody on four thousand other storefronts at the same time.

The average buyer does not exist

Shared product data has a structural constraint that is easy to miss because it sounds like a feature: one record, distributed everywhere. A brand writes the description once. A network like ASAP formats it, validates the fitment, and delivers it to every member dealer who requests that line. Nobody re-keys a spreadsheet. Everyone publishes on Tuesday.

The constraint is that a record written for everyone has to be written for the average, and the average aftermarket buyer is a statistical artifact. Real buyers arrive with a vehicle, a use case, a skill level, and a budget, and they are filtering hard on all four before they read a word of marketing copy.

This is more acute in the aftermarket than almost anywhere else, because fit is binary and suitability is not. Four brake pad sets fit the truck. Which one is correct depends entirely on whether it tows 9,000 pounds twice a month, sits in stop-and-go delivery routes, or gets driven to Cars and Coffee. The fitment table says all four are right. Only one of them is.

What each buyer is actually filtering on

The useful exercise is to stop thinking about audience segments and start listing the specific fields each buyer needs to see before they will commit. Same category, same catalog, wildly different fields:

Buyer	Deciding question	Attribute that answers it
DIY driver	Can I do this in my driveway?	Special tools required, press required, torque specs published
Pro installer	What's my book time and what else do I need?	Labor time, included hardware, required-but-not-included parts
Fleet / municipal	What does it cost me over 100k miles?	Service interval, warranty term, rebuild kit availability
Enthusiast	Will it clear my setup?	Lift requirement, max tire diameter, backspacing, wheel offset
Restorer	Is it correct for the year?	Finish, date-code correctness, OE casting number
Commercial upfitter	Does it survive compliance?	GVWR impact, emissions legality, CARB EO number

Look at the right column. Almost none of those live as structured fields in a typical distributed record. Most exist in an install PDF, a footnote, a forum thread, or nowhere at all. Which means that on the specific question that decides the sale, your catalog is silent — and so is every competitor publishing the same file, which is why the whole category converts worse than it should.

Persona work is really attribute discovery

"Write for your persona" is advice that usually produces nothing but adjectives. The version that works is mechanical: for each buyer, find the questions they ask that your schema has no field for, then create the field.

Those questions are already sitting in signals most catalogs never read:

Internal site search. Queries that return zero results, or return a page of forty items, are a list of attributes your facets do not have. fits leveled 2019 ram 1500, low dust pads for towing, no drill install. Each one names a field.

Return reasons and one-star reviews. The most honest attribute research available. "Doesn't include the ABS tone ring" appearing eleven times is not a customer service problem, it is a missing boolean.

Owners groups and platform forums. Aftermarket buyers do their compatibility research in public. A thread where forty people argue about whether a lift pump kit reuses the factory sending unit is a specification the brand never published and every buyer needs.

The counter phone. Your own people answer the same six questions all day. Every one of those is an attribute that should be on the page.

What a competitor added. When one listing outranks three identical ones, look at what is different. Frequently it is a single field — a CARB EO number, a 50-state legality flag — that turned a generic page into the answer to a specific query.

This is the same loop that decides whether an AI shopping engine cites you. Models expand a question like "will this kit fit my leveled Ram" into sub-questions about clearance, required hardware, and alignment, then compare sources fact by fact. A page that answers the sub-questions in labeled fields gets used. A page with a marketing paragraph does not, however good the part is.

One record, more depth — not five pages

The wrong implementation is a page per persona. That manufactures thin, near-duplicate content and makes the ranking problem worse.

The right implementation is one product record carrying enough structured attributes that every surface can select what it needs:

• The category page facets narrow on lift requirement and tire diameter for the enthusiast, and on labor time and included hardware for the shop.
• The product page leads with the fact that buyer's traffic source implies — the query that brought them in tells you which question is live.
• The feed exposes all of it, so marketplaces and answer engines can match on the specific attribute rather than on a title.
• The copy is written once but written to be decisive, naming the vehicle, the use case, and the tradeoff, instead of gesturing at superior performance.

That is a depth problem, not a duplication problem. And it is precisely the depth a shared network file cannot have, because a record built to be distributed identically to four thousand dealers cannot also be aimed at your buyer.

The practical order of operations

1. Take the shared baseline. Fitment, spec blocks, image links, install PDFs. Do not rebuild what a network already delivers correctly.
2. Pick your two real personas. Not five. Most dealers make their money on two, and the honest answer is usually visible in order data.
3. Mine their questions from site search, returns, reviews, forums, and the counter, and turn each recurring question into a proposed attribute.
4. Govern the values before you fill them, so lift requirement does not arrive as 2", 2 in, and two inch across three brands.
5. Extract from source documents — the install manual already contains the U-bolt length and the re-torque interval; it just is not in a field.
6. Rewrite the baseline copy for the buyer, and keep a citation on every value so a disputed spec is settleable instead of arguable.

Steps three through six are the part that does not scale by hand across 14,000 SKUs, which is the work Anglera automates: reading the signals, proposing the attributes your schema never had, normalizing the vocabulary, mining specs out of supplier PDFs with per-value provenance, and writing all of it back to your PIM.

The network hands you and four thousand competitors the same paragraph about superior off-road performance. The dealer who instead answers will it clear 37s on a leveled truck, and do I need longer U-bolts is the one the trail rider, the shop foreman, and the language model all end up choosing.