Shopify's built-in collection page filters come from Search & Discovery, the free first-party app, and they can be built from product type, vendor, price, availability, variant options, tags and product metafields. That last source is the whole game for a handloom catalogue. Colour and size are the defaults because the tooling was designed around apparel brands that sell the same shirt in five colours and four sizes. A kasavu store doesn't work that way. Your buyer wants to narrow by weave cluster, border width, zari type, drape length and occasion, and none of those exist until you create them as metafields first.
How to add product filters to a Shopify collection page
The mechanics take about ten minutes, which is why so many stores stop at colour and size and call it done.
- Install Search & Discovery from the Shopify App Store. It's free and first-party.
- Open the Filters tab, add a filter, and pick its source: a variant option, a tag, price, availability, or a metafield definition.
- Rename the label so it reads the way customers speak. "Border" beats "custom.border_width".
- Drag the filters into priority order. The first two get used; the sixth rarely does.
- In the theme editor, confirm the collection template has the filtering block enabled. Dawn and most Online Store 2.0 themes ship with it; older themes often don't support it at all.
Shopify hides filter values that return nothing inside the collection you're viewing, so a "Chendamangalam" option won't show up on a churidar material collection. That behaviour is useful and it also hides your data-entry gaps, which is the trap in the next section.
Why Shopify's default filters miss your products
Three reasons, in the order we usually find them.
The product has no value for the facet. Filters are an exact match against stored data. A saree with the weave metafield left blank disappears the moment anyone clicks a weave filter. Not demoted. Gone. On a catalogue built over two Onam seasons by three different people, the blank rate on any given attribute is usually somewhere between 10% and 40%.
The values are inconsistent. "4 inch", "4in", "4 inches", "4\"" and "four inch" are five facet values describing one border. The filter drawer becomes a list of near-duplicates and each one returns a handful of products.
The attribute lives in the title or description. Plenty of handloom stores put everything in the product title: "Kuthampully Pure Cotton Kasavu Saree 6.3m with 4 inch Zari Border". Great for search snippets, useless for filtering. Liquid can't facet on a substring.
Fix the data before you fix the filters. We run a spreadsheet export, count blanks per attribute, and refuse to configure the facet until the blank rate is under 5%. It isn't glamorous work, but a filter that silently drops a third of your catalogue is worse than no filter.
The facets a kasavu buyer actually uses
From how people browse and what they ask on WhatsApp before ordering:
- Weave or cluster β Balaramapuram, Chendamangalam, Kuthampully, plus generic handloom and powerloom. Buyers from the Gulf often search by cluster name because that's what the family at home asked for.
- Zari type β real/pure zari, tested or half-fine, copper, artificial. This drives a price gap of several thousand rupees on otherwise identical-looking sarees, and it's the single most common pre-purchase question.
- Border width β in inches, banded: under 2", 2β4", 4β6", above 6". Don't store 37 distinct widths.
- Length and format β 5.5m saree, 6.3m with blouse piece, set mundu (two-piece), single mundu, dhoti, kasavu material by the metre. Format matters more than length, and they're frequently conflated into one messy field.
- Cotton count β 80s, 100s, 120s. A knowledgeable buyer filters on this. A gifting buyer never touches it, so rank it low.
- Occasion β Onam, Vishu, wedding, daily wear, gifting.
Note what's missing: colour. For most kasavu catalogues the colour facet puts 70% of products in one off-white bucket, and a filter that only removes 30% of the results isn't doing anything. Keep it if you also sell coloured handloom cottons. Drop it if you don't.
Four facets, not eight: the arithmetic
Say you list 900 products. Six weaves Γ five border bands Γ four formats = 120 combinations, or an average of 7.5 products per fully-filtered view. Add occasion and cotton count and you're at 1,440 combinations, under one product each. The customer clicks three filters, sees an empty grid, and leaves.
So: show four or five facets, allow multi-select inside each one (Shopify's filters OR within a facet and AND across facets), and order them by how often they're used rather than how proud you are of the data. On mobile β and in Kerala that's the overwhelming majority of your traffic β the drawer should show result counts next to each value so nobody clicks into a dead end.
Shopify metafields as filter options: setting them up properly
Go to Settings β Custom data β Products β Add definition. For weave, use namespace custom, key weave, type List of single-line text, and under validations set limit to specific choices with your cluster names typed once. That validation is the piece people skip, and it's the difference between a clean facet and the five-ways-to-write-four-inches problem. Do the same for zari type, border band and format. Make sure the definition is exposed to the storefront, otherwise Search & Discovery won't offer it as a filter source.
Supported types for filtering are limited β single-line text, lists of single-line text, boolean and number are the ones we rely on. Rich text and file references won't appear as filter options, so don't model anything you plan to filter on as a rich text field.
Once live, the URLs look like /collections/kasavu-sarees?filter.p.m.custom.weave=Kuthampully&filter.p.m.custom.border_band=4-6-inch. Worth knowing, because you'll want to link to specific combinations from email and Meta ads during the Onam run-up.
On tags. They work, they're faster to set up, and they rot. You're capped at 250 tags per product, tags carry no validation, and within a year you have weave-kuthampully, Kuthampully and kuthampully_weave all in the same store. Use tags for temporary merchandising (onam-2025, clearance) and metafields for anything structural. If you're migrating an existing tag mess, Shopify's bulk editor handles a few hundred rows; past that, a Matrixify import or a quick script is less painful.
Improving on-site search for a Malayalam and Manglish catalogue
Filters handle browsers. Search handles people who already know what they want, and they type it in a transliterated mix that no exact-match engine handles well. "Settu mundu", "set mundu" and "mundum neriyathum" are the same product. "Kasav", "kasavu" and "kasava" are the same word. Search & Discovery lets you define synonym groups, and twenty minutes spent writing twenty of them recovers more revenue than most theme tweaks. Pull the actual query list from Shopify's search analytics, sort by searches with zero results, and start there.
Where native search runs out is descriptive queries: "cream saree thick gold border for Onam", "light cotton set mundu under 3000". Keyword matching returns nothing useful because no single token matches. Semantic or AI search embeds the query and the product together and returns the right fifteen products even with no exact word overlap. On a catalogue above roughly 500 SKUs with heavy long-tail search, that's where the lift is. Our own app, FilterPro, adds metafield-driven filters and AI search together, which is the combination most handloom catalogues need; there are other good options and the point stands regardless of which you pick.
One blunt opinion: don't make payment or shipping a facet. "COD available" and "ships to UAE" belong as badges on the card and rules at checkout, not as filters. Every time we've seen them as facets, the data went stale within a month.
Faceted navigation SEO risks worth taking seriously
Shopify's canonical tag on a filtered collection URL points back at the unfiltered collection, so you're unlikely to create duplicate-content penalties. The real cost is crawl waste. Four facets with multi-select produce thousands of valid URL combinations, and Googlebot will happily spend its budget on ?filter.v.price.gte=1200&sort_by=created-descending instead of your new Onam collection.
What we do:
- Edit robots.txt.liquid (editable on all plans) to disallow crawling of sort and price parameters, keeping the attribute filters that could matter.
- Never link filter combinations from the footer or mega-menu. If it isn't linked, it's mostly not crawled.
- For the three or four combinations with genuine search demand β "Kuthampully kasavu saree", "set mundu online" β build a real collection with its own handle, a 150-word intro and an FAQ. A filter URL will not rank for a commercial query. A merchandised collection page will.
Resist building fifty of those. Each one needs unique copy and ongoing stock, and thin collection pages are a bigger liability than missing ones.
Merchandising the collection, not just filtering it
Filters narrow. Merchandising decides what people see first, and on a handloom store the default "best selling" sort will show you last Onam's winners for the whole of August. Search & Discovery's product boosts let you pin specific products to the top of a collection or a search term. Set in-stock-first as the base sort, boost the current season's range for six weeks, then pull it back.
Two Kerala-specific adjustments. First, the price facet renders in the customer's presentment currency, so a Gulf buyer sees AED bands β check that your bands make sense after conversion rather than reading as βΉ2,000-equivalent oddities. Second, Onam traffic arrives in a compressed two-week window, and filter interactions hammer the collection template. Dawn-based themes re-render the product grid through the Section Rendering API, which is fast; many third-party filter apps re-render client-side and add 300β600ms per click. Test it on a 4G connection in August, not on office wifi in June. If the collection page is already slow before filters, fix that first β our notes on Shopify speed optimisation cover where the time usually goes.
Quick answers
Do I need an app to filter on metafields? No. Search & Discovery is free and handles metafield filters natively. You'd add a paid app for AI search, filter analytics, swatch rendering or filtering on inventory across multiple locations.
Do I need Shopify Plus for this? No. Everything above works on Basic. If someone is quoting you an upgrade to get better filters, that's not where the Plus value sits.
How long does the data cleanup take? For 800β1,000 products with four new attributes, budget two to three weeks of part-time work by someone who knows the stock. We get the first pass at border bands wrong fairly often, because what the weaver calls a 4-inch border and what measures 4 inches aren't always the same thing.
Pick one collection β your highest-traffic saree collection β export it to CSV, and count how many rows are blank in the four columns you'd want to filter on. That number tells you whether this is a configuration job or a cataloguing job. If you'd rather have someone else do the counting, our team in Bengaluru works with handloom and textile merchants across Kerala, and a free store audit will flag the facet gaps along with everything else.