Free Shopify store auditSpeed, SEO and conversion leaks — no cost, no obligation.
Claim it
Thriftizer Solutions LLPShopify Select Partner
Book a Growth Audit
Analytics Nov 21, 2026 8 min read

Sidekick Metafield Reports: Asking Shopify Which Fabric Actually Sold

Sidekick can now answer analytics questions from metafields and metaobjects. We tested it on a post-BFCM sell-through question and the first answer was wrong three ways. Here's the cleanup it needed.

Sidekick Metafield Reports: Asking Shopify Which Fabric Actually Sold

Yes, you can now ask Shopify's Sidekick which fabric sold best, and it will answer from your metafields rather than making you export a CSV. The capability landed on 17 November, which means this is the first BFCM dataset anyone can interrogate that way. The catch is that shopify sidekick metafields analytics is only as honest as your metafield definitions. If your fabric values are a free-text field where someone typed "Linen", "linen", "100% Linen" and "Lin" across four seasons, Sidekick will cheerfully report four fabrics and you will make a buying decision on fiction.

We ran the test on a Canadian apparel store the week after Cyber Monday. Here's what worked, what broke, and the cleanup that had to happen first.

What Sidekick will and won't answer in the admin

Sidekick sits in the Shopify admin and answers questions in plain language. For analytics, that now extends to product data held in metafields and metaobjects, so a question like "what was units sold by fabric last month" resolves against a structured attribute instead of only against product titles, vendors and tags.

What it does well: aggregating orders by an attribute, filtering by date range, comparing two periods, and giving you the answer in the admin in under a minute. What it does not do: judge whether the attribute is meaningful. It has no opinion on whether your 11 fabric values should be six. It will not tell you that 340 products have no fabric value at all and have been silently dropped from the denominator. That omission is where most wrong decisions start.

It also won't reconcile across apps. If your returns live in a third-party returns app and never write back to Shopify, Sidekick is reporting gross sales and calling it performance.

The question we put to it after BFCM

The real question a merchandiser has in early December is not "what sold". It's "what should I reorder for spring, and what do I mark down on Boxing Day". For an apparel catalogue, that usually resolves to fabric, fit and supplier.

So the question was: for the BFCM weekend through Cyber Monday, units sold and net revenue by fabric, with returns deducted.

Sidekick answered in about twenty seconds. The answer was wrong in three separate ways, and none of them were Sidekick's fault.

Why the first answer was wrong

First, the fabric metafield was single-line text, not a metaobject reference, so every typo since 2022 was its own category. Eleven values, six real fabrics.

Second, roughly a fifth of the catalogue was missing a value entirely. Products added by a part-time merchandiser during the autumn drop, where the field exists in the definition but was never filled. Those orders simply did not appear in the breakdown. Revenue that nobody argued about, because nobody saw it.

Third, the metafield lived at the product level, not the variant level. Fine for a store where fabric is a property of the garment. Not fine for the three styles sold in both cotton and a cotton-linen blend as separate variants. Those rolled up under whatever the product-level value said.

We get this wrong on the first pass more often than we'd like. The fix is boring and it takes a morning.

Metafield definitions that make the answers trustworthy

The best practice for metafield definitions in Shopify is not complicated, it's just rarely enforced:

  • Use a constrained type. A list of choices, or a metaobject reference, instead of single-line text. The moment a human can type the value, the data is dirty within a quarter.
  • Put the attribute at the level where it actually varies. Fabric is usually product-level. Size and colour are variant-level. Country of manufacture, which matters enormously if you ship into the US, is often variant-level because the same style gets made in two factories.
  • Namespace deliberately. custom.fabric is fine for one field. Once you have fourteen, you want product.fabric, sourcing.supplier, compliance.origin so a person reading the list six months from now knows which team owns which.
  • Decide what null means before you backfill. "Unknown" and "not applicable" are different answers and lumping them together ruins the denominator.
  • Mark the field as filterable where it has customer value. The same clean fabric taxonomy that gives Sidekick a straight answer is what powers a usable collection filter, which is the other reason to do this properly. We use FilterPro for that side of it on larger catalogues.

Backfilling is the part people avoid. For a 900-SKU catalogue, a bulk edit against the metafield column in the admin, or a Matrixify import, takes an afternoon. Budget for someone to actually look at the garments they're unsure about rather than guessing.

Metaobjects, and the supplier breakdown almost nobody has set up

Metafields hold a value. Metaobjects hold a record. The difference matters the moment you want a supplier-level sales breakdown, because a supplier is not a string, it's an entity with a lead time, a currency, a country of origin and a minimum order quantity.

Set up a supplier metaobject with those fields, point a product metafield at it, and you can ask for revenue by supplier and get a real answer. Then the follow-up questions become possible: which supplier's goods carry the highest return rate, which ones sit on more than 90 days of cover, which ones are made in a country that changes your duty position when you ship to a US customer.

That last one is not theoretical for Canadian merchants. Origin drives whether a cross-border order clears duty-free or not, and the de minimis rules have moved. If origin is a free-text note in a Google Sheet, nobody can answer the question at speed. If it's a field on a supplier metaobject, you can ask it in the admin.

Metaobject reporting has a limit worth knowing: Sidekick reads the reference, so the metaobject's own fields need to be populated and typed sensibly. An empty country_of_origin field on the metaobject is the same blind spot as an empty metafield on the product, one level deeper and harder to spot.

The arithmetic Sidekick won't do for you

Units sold by fabric is the wrong metric on its own. Sell-through against units received, net of returns, with margin applied, is the one that tells you what to buy. Here is the version we ran, with real numbers from the cleaned dataset, in CAD.

Linen: 980 units received for the season, 412 sold over the BFCM weekend at an average of $89 after discount. Returns ran at about 11%, so 45 units came back. Net units 367. Net revenue $32,663. Landed cost $28 a unit, so COGS on net units is $10,276. Gross margin $22,387. Divide by the 980 units committed: $22.84 of margin per unit received.

Merino: 640 units received, 408 sold at $129. Returns 4%, so 16 back, 392 net. Net revenue $50,568. Landed cost $51, COGS $19,992. Gross margin $30,576. Across 640 units received: $47.78 per unit received.

Linen sold more units. Merino made more than twice the margin on every unit of capital committed. If you had asked "what was my best seller" and stopped there, you would reorder linen.

One caveat we had to apply by hand. Canada Post disruption pushed delivery windows out during the weekend, and the returns and cancellations that followed were not evenly spread across fabrics, because the heavier merino pieces had already been routed to a courier fallback while lighter linen goods were still going via Canada Post. Some of that 11% is a carrier story, not a product story. Sidekick cannot know that. You can.

Bilingual values, and why Quebec makes this a compliance question

If you sell into Quebec, your product attributes are not purely an analytics concern. Bill 96 obligations mean French needs to be present on the storefront, and metafield values that render in the customer-facing part of a product page are part of that. "Brushed cotton twill" sitting in an English-only text field on a product template is a gap.

Metaobjects handle this better than raw metafields, because you can translate a metaobject entry once and have every product referencing it inherit the translation. Translate 240 free-text fabric strings instead and you'll be doing it again next season.

There's an analytics benefit too. One canonical record per fabric means the French and English storefronts report into the same bucket. Separate text fields per locale means you get "Linen" and "Lin" as two rows and have to add them up yourself. We've seen that split go unnoticed for a full year on a bilingual store.

Where ShopifyQL and custom reports still beat asking a question

Sidekick is fast for the question you think of at 9am on a Tuesday. It's not where you build the report you run every month.

If you want custom dimensions in Shopify analytics that survive, build the custom report, save it, and let the team run it. ShopifyQL in the analytics section handles grouping and filtering against your data with more precision than a conversational prompt, and the output is reproducible. Nobody has to re-prompt their way back to last month's definition of the question.

Our rule of thumb: explore with Sidekick, operationalise in a saved report. The conversational layer is best at discovery and at the questions too occasional to justify building anything.

Also worth saying plainly: if the attribute you care about drives real money, it belongs in your planning spreadsheet or your ERP anyway, pulled via API. The admin is for looking, not for forecasting.

Questions we get asked about this

Can I analyse sales by any metafield? Product and variant metafields with proper definitions, yes. Customer and order metafields are patchier. If a field was created by an app and has no definition attached, expect it to be invisible.

What does Sidekick do in the admin beyond analytics? Bulk edits, setting up discounts, drafting product copy, explaining settings. The analytics side is the newest and, for a merchandiser, the most useful.

How many values should an attribute have? Fewer than you think. If fabric has more than about a dozen values on a mid-size apparel catalogue, you have a taxonomy problem, not a rich dataset. Roll up the long tail into a parent value and keep the detail in a second field.

Do I need Shopify Plus for this? No. Metafield definitions and metaobjects are available on standard plans. Plus gets you more metafield capacity and B2B-specific fields, but for fabric-level reporting a Basic store is fine.

Before Boxing Day

You have a few weeks. Pick the single attribute that drives your reorders, give it a constrained definition, backfill the gaps, and then ask Sidekick the question. If the answer surprises you, check the null count before you believe it.

If the cleanup is bigger than an afternoon, or your product data is scattered across tags, a legacy app and three spreadsheets, that's the sort of thing our team in Canada-focused Shopify work handles as a fixed-scope project. A free audit will at least tell you how dirty the data is before you commit to the spring buy.

Next post

Ready to scale your D2C brand profitably?

Let's build a growth engine that drives more traffic, more conversions and more profit.

Book a Growth Audit
📅 Free Audit💬 WhatsApp