FAQ

Will this change anything in my store?

Nothing automatically. The service is read only and writes nothing to your store. The file you get back carries only the measured columns, and what you import, and when, is entirely your decision.

Do you need access to my store?

No. Public data only. No login, no admin access, no customer data. A run reads the public pages of your store, the same ones any shopper's browser fetches. There is nothing to install.

Where do the values come from?

Your own copy, or a named authority, with a citation either way. A value that cannot be proved from one of those is refused rather than shipped.

What does refused mean?

A deliberate withholding. Where nothing supports a value, the gap ships as a gap, because an absence is safer than a guess. Every refusal appears in your report with its reason.

Can't I just use AI to do this?

You could, and inparticular is built on the same kind of AI underneath. The difference is what we put around it. Ask a general model to fill your catalogue and it hallucinates, confidently inventing a material or a size your copy never states, and you cannot tell its inventions from its facts. Run it across thousands of values and it turns inconsistent, answering the same question two ways and breaking the Shopify import as it goes. Doing this properly means catching every hallucination, holding one standard across the whole catalogue, citing every value to its source, and handing back a file that imports clean. That is the job. inparticular is the machine that does it. The AI reasons. inparticular proves every value, refuses what it cannot prove, and never lets a hallucination or an inconsistency reach your file.

Doesn't Shopify already infer attributes for free?

It does, and it is useful. The word to hold onto is infer. Shopify's own documentation notes that inferred values vary in accuracy, are not editable by you, and reach the AI surfaces as signals rather than facts you authored. Inference can only work from what is already there. Where a fact sits in your prose but not in the fields the machines read, inference has to reach for it. Where the fact was never recorded at all, there is nothing to infer from, and it cannot be known. inparticular fixes both. We recover the fact from your own copy or a named authority and put it in the machine-readable field where it belongs, proven and cited. The more of your data is real and structured, the better every surface reads it, inference included. We do not compete with the inference. We give it the truth to work from.

How do AI surfaces read these fields?

As structured source data. A shopping assistant works from the machine-readable fields in your feed, not from your prose, and the file lands every value in exactly those fields. The import guide that comes with your report walks the feed mapping step by step, naming the field each column fills.

Why per product and not per SKU?

Every variant is included. A product with forty variants is one product on the rate card, so the price is per product, never per SKU and never per variant.

What if you find conflicts?

When two places in your own data disagree about the same attribute, we show you both and choose neither. Your variant might read Mid Blue where your own tag reads blue. We surface the pair, cite each to where it came from, and leave the decision to you, because guessing which one is right is the one thing we refuse to do.

What happens after I pay?

The file goes to the email you gave us, usually within a couple of days. It arrives the same shape as your free sample, every value cited, ready for the Shopify import.

Anything else goes to hello@inparticular.ai. Back to inparticular