Your catalog has been almost ready for nine weeks.
Not blocked. Not cancelled. Just moving through a queue — a supplier file, a copywriter, a translation, a rejected upload, a person who is on holiday. Nine weeks in which the product exists, is in stock, and cannot be bought.
The number worth knowing is not how long onboarding takes. It is how much of that time is actually work. For most catalogs the answer is about one day in eight.
Working days for a 25,000-SKU catalog.
Same catalog, same channels, end to end.
Payloads accepted without a rejection cycle.
Waiting, not working. This is the real target.
Both bars below are to scale.
That is the entire argument. Set your catalog volume and watch what happens to the second bar — it does not shrink politely, it disappears.
Second bar is 0.12% the width of the first · held at a 3px minimum so it remains visible
At a blended $4.50 per enriched SKU.
At $0.35 per enriched SKU.
Before counting anything the delay itself costs.
Five stages. One of them is real work.
Timings are for a 25,000-SKU catalog going live across three channels. The line under each stage is the part nobody puts in the project plan.
Ingestion & mapping
Supplier feeds arrive as whatever the supplier felt like sending: CSV, Excel, XML, EDI, a folder of PDFs, a ZIP of images named by SKU if you are lucky. Aonex parses all of it in place. Nobody reformats anything.
Waiting on: the supplier to resend a file that opens correctly.
Extraction & normalisation
Dimensions, materials, voltages, certifications and compliance marks are pulled out of spec sheets and manuals, then converted to GS1 unit codes. This is the stage where a human being opens four hundred PDFs.
Waiting on: one person who understands the product taxonomy.
Copy & localisation
Titles, bullets and descriptions generated against your tone guidelines and channel character limits, in forty-plus locales. Not translated afterwards — written per locale, with the attribute set intact.
Waiting on: the copywriter's queue, and then a translation agency.
Channel payload mapping
Every payload is validated against live channel rules — Amazon SP-API, eBay Sell Inventory, Etsy Open API, Shopify GraphQL, WooCommerce REST — before submission. Errors surface here, in seconds, rather than three days later in a rejection email.
Waiting on: the marketplace to tell you which field was wrong.
Activation & sync
Listings go live across every regional channel simultaneously, with stock and price sanity checks running before anything publishes. No per-channel publishing rota, no staggered go-live.
Waiting on: someone to work through the channels one at a time.
Actual work: reading specs, writing copy, building the mapping sheet, fixing rejections.
Not work. Waiting for a file, a person, a translation, an approval, or a marketplace to answer.
This is why the speed-up is so large, and it is not a claim about how clever the models are. Automation does not make the seven days of work vanish — it makes the fifty-two days of waiting vanish, because software does not have a queue, a holiday, or an inbox.
A product that is in stock, priced, and not yet listed earns exactly the same as a product that does not exist.
The invoice you never receive.
Labour cost is the easy number and the small one. The expensive number is shelf time: every day a SKU sits in the pipeline is a day it cannot be sold, and you never get those days back.
SKU-days of sellable inventory that were never on sale. Not a projection — just volume multiplied by delay.
We deliberately stop short of putting a revenue figure on that, because we do not know your margins and any number we invented would be marketing. You do know them. Multiply.
What we can say is that the arithmetic is unusually unforgiving in three situations, and most catalogs hit at least one of them every year.
A nine-week delay on a twelve-week season does not cost you nine weeks of margin. It costs you the season.
First listing on a new product category captures review volume and ranking that later entrants spend a year buying back.
Vendors route their best allocation to the partners who can list fast. Slow onboarding quietly costs you the next range.
The questions you’re about to ask.
Our supplier data is genuinely terrible. Doesn't that break the whole premise?
It is the premise. Clean feeds were never the bottleneck — nobody needed a platform for those. Aonex is built for the column named DESC2, the dimension recorded in three different units across four suppliers, and the specification that only exists inside a scanned PDF. Where a value genuinely cannot be sourced from anything you have, it is flagged as missing rather than invented, and you get a list of exactly what to request from whom.
What formats can you actually ingest?
CSV, Excel, XML, JSON, EDI, PDF manuals and spec sheets, CAD drawings, and ZIP archives of images. Vision and document models read the ones that are not structured data at all. There is no supplier onboarding project, no template for them to fill in, and no requirement that they change anything about how they send you files.
How do you avoid the rejection loop on marketplaces?
By moving validation before submission instead of after it. Every payload is checked against live channel schemas — required attributes, value formats, character limits, category-specific rules — and non-compliant values are corrected or flagged before anything is sent. First-pass approval runs at 99.4%. The days you currently lose are not spent fixing errors; they are spent waiting to be told an error exists.
Ninety-two percent faster sounds like a number invented for a slide.
It is arithmetic on the stage table above, and you can check it. Fifty-nine working days against one hour and forty-two minutes for a 25,000-SKU catalog. The reason the gap is that wide is not that the software is miraculous — it is that roughly fifty-two of those fifty-nine days were never work in the first place. They were queue. Removing queue is a much easier problem than removing work.
What does the first launch actually look like?
Send one supplier feed. We ingest, enrich and validate it, then hand you the payloads and the rejection report before anything publishes. You approve, or you do not. Most teams run the first two or three feeds with full human sign-off, watch the confidence scores, and then move low-risk attribute classes to autonomous once they stop disagreeing with the output.
Start the clock on a real feed.
Pick the supplier feed that has been sitting in someone’s inbox the longest. We will run it end to end — ingestion, enrichment, validation, channel payloads — and hand you the output plus the rejection report before a single listing publishes.
One feed · Nothing publishes without approval · Output in 48 hours