An ecommerce SEO agency decides which parts of a catalogue are worth ranking, then makes a large site capable of ranking them without somebody hand-editing every page. Both halves are load-bearing. Choosing badly wastes the work, and choosing well on a site that cannot be crawled wastes the choice.

Four jobs sit under the label.

1. Deciding where the money actually is. Which categories, which products and which buying questions deserve investment, weighed against margin, stock depth and who currently holds the positions. That is a commercial decision before it is a search one.
2. Making a large site legible. A store manufactures addresses on its own through filters, sorting, variants and pagination. Somebody has to decide which of them a search engine should keep, because left undirected a two-thousand-product catalogue can present several hundred thousand pages and get its important ones visited once a month.
3. Making the pages worth choosing. Collections that answer the question a shopper arrived with, product pages carrying something the supplier text does not, and machine-readable facts so price, availability and ratings show up in the listing.
4. Protecting what has already been earned. Products end, seasons rotate, platforms get replaced. Each of those events can quietly delete years of accumulated equity, and having a policy for them is the least glamorous part of this discipline and one of the highest returning.

Brand Vision has been running since 2018, with more than 500 projects delivered for roughly 150 brands and 250 or more verified five-star reviews behind that record. The independently collected figure is a Clutch score of 5.0, built on interviews with 64 or more clients. The average increase in leads recorded across past engagements is 250%, and that figure needs its limits attached. It pools accounts that differ enormously in size and in the condition they arrived in, and it reports finished work instead of forecasting your store.

Two things worth checking before you commit a budget. Look at stores we have worked on and at the senior team on the account, by name. Reading a crawl log and writing a buying guide are unrelated skills, and the roles we recruit for are split on purpose so one person is never asked to do both badly.

Nine, and no store needs all nine at once. Which ones apply comes out of the catalogue, the platform, and how much of the site is currently being indexed against how much should be.

Ecommerce technical SEO. Crawling, index coverage, canonicals across variants, pagination, speed on the product and collection templates, and rendering. It shares its method with technical SEO everywhere else and differs in one respect, which is that a single template fault repeats itself across ten thousand addresses.

Category and collection page strategy. Which collections should exist, what each one owns, what appears on the page, and how they nest inside each other. Built from how people shop instead of from how the warehouse is organized.

Product page optimization at scale. Titles, descriptions, specifications, imagery, reviews and internal links, applied through templates and prioritized so effort lands on the products capable of paying for it.

Faceted navigation and crawl control. Deciding which filter values earn a place in the index, which stay reachable and unindexed, and which get shut off entirely.

Product feed and structured data. Product, offer and review markup that agrees with the visible page and with your merchant feed, so price and availability are stated once and correctly in every place a machine reads them.

Site search and internal linking as merchandising. Your own search logs read as demand evidence, and links placed to move rankings and revenue at the same time.

Migration and replatforming protection. Address inventory, collection mapping, redirect maps that account for discontinued items, feed resubmission, and verification after cutover instead of only before it.

Seasonal and out-of-stock handling. A written policy for what happens to a page when the product goes away, applied by whoever manages the catalogue every day.

Reporting tied to revenue. Organic revenue and orders by collection, with margin where you are willing to share it, so a specific investment can be judged instead of a blended total.

None of the nine is sold as a fixed package. Where the whole picture needs establishing first, the broader audit covers content and authority alongside the technical condition. Most of what matters here ships through templates, so whoever owns the theme code joins the work early instead of receiving a list of requests in month three.

The largest opportunity in a catalogue is almost never the term with the highest search volume attached to it, and a keyword list ordered by volume will bury the one that actually pays. Volume proves a market exists. It says nothing about whether you can serve that market profitably.

Five things get weighed before anything is committed.

• Demand and how contested it is. How many people search, in what words, and who holds the positions today. A term owned by three marketplaces and a manufacturer is a different proposition from one owned by two small retailers.
• Gross margin on what the term sells. A category you sell at nine points has to move enormous volume to fund the work. One at fifty-five points funds itself inside a quarter.
• Stock depth and reliability. Ranking a collection where you carry four items against competitors carrying ninety produces a page that disappoints everyone who lands on it. Ranking a category you run out of every six weeks produces the same disappointment intermittently.
• Strategic importance of the category. Some categories exist to win a customer who then buys the profitable thing. Some are the profitable thing. Some are being quietly wound down and nobody told marketing.
• Intent inside the phrasing. Somebody typing a category with a specification attached is much closer to paying than somebody typing the bare category, and the bare category usually carries ten times the volume.

A worked version. Twenty-two thousand monthly searches sit on a category where you stock two items, hold single-digit margin, and currently rank eleventh behind two marketplaces. Thirteen hundred searches sit on a subcategory where you carry sixty items, hold healthy margin, already rank sixth, and get asked about it by customers every week. The second one is worth building first, and it looks like the wrong answer on any spreadsheet ordered by volume.

One demand source almost nobody uses. Your internal site search log records what visitors asked for in their own words, including every query that returned nothing. It is free, it describes your customers instead of a national average, and it reliably names three collections you should build and two you should retire. We read it in week one.

Where the profitable terms sit out of reach for now, paid shopping buys the position while the slower work compounds. Margin structure decides more of this than most operators expect, and for food and drink brands the arithmetic is tighter than in almost any other category we work in.

Collections carry most of the commercial rankings in a catalogue, and most stores put most of their effort into product pages. Both have to work. The order of investment is the thing that gets decided wrongly.

What collection pages win. The terms with real money behind them, because those describe a set instead of a single item.

• Category and subcategory language, which is how somebody shops before they have chosen anything
• Category with an attribute attached, meaning a material, a finish, a size range, a compatibility or a price band
• Browsing and comparison intent, where a shopper wants options laid out instead of one product argued for
• Links from other sites, since a guide or a roundup points at a category far more often than at one product

What product pages win. Narrower traffic that converts hard when it lands.

• Model numbers, part numbers and exact product names
• Queries where the shopper already knows precisely what they want
• Long-tail specifics on items exclusive to you
• The transaction itself, which is why the page earns attention even when it will never rank

The common failure is a collection page that is a product grid and nothing else. No opening that answers the question somebody arrived with, no guidance on choosing between the options, no links down to the narrower collections beneath it, and a title assembled automatically. That page competes for a valuable term with nothing to say, and the retailer who wrote three useful paragraphs takes it. Five hundred words dumped under the grid fools nobody, and the opening has to be short enough that it does not push the products off the screen.

Then internal linking, which is the lever almost nobody treats as one. What you link from the homepage, from the navigation and from inside collection copy decides which pages accumulate authority. It also decides what sells. Those are the same decisions, which means merchandising and search are choosing from one list, and the store that runs them as one function gets paid twice for each choice. A deep collection nothing links to sits five clicks from the homepage and ranks exactly as the structure says it should.

Whether any of it gets crawled and kept is indexation work. What actually goes on the pages belongs to content strategy. And the fact that collections attract outside coverage more readily than product pages is worth building into which pages you promote.

Catalogue work is priced from template count and the state of the index, never from product count. A forty-thousand-item store running on six templates is less work than a four-hundred-product store that five agencies have modified over eight years.

How the phases usually run.

1. Crawl, catalogue and baseline. Two to three weeks. A full crawl, index coverage, analytics and search console access, the merchant feed, the site search log, and revenue by collection recorded before anything gets touched. Margin data too, where you are willing to share it, because the priorities change completely without it.
2. Opportunity model and architecture. Two to three weeks. Demand read against margin, stock and category importance, then the collection structure and the facet rules decided. This document is what the rest of the engagement executes against.
3. Crawl control and technical work. Weeks four to twelve. Your release cycle sets this pace and ours does not. Configuration changes ship in days, template changes wait for an engineering slot, and rebuilding a filter system is a project with its own timeline.
4. Collections and product content. Month two onward. New collections built, priority product pages rewritten, templates improved, and linking changes going out with each release.
5. Measurement and reallocation. Continuous. Re-crawl, re-check coverage, read revenue by collection, and move effort toward whatever moved.

What moves the number.

• Variant complexity, since a thousand products with twelve options each behaves like twelve thousand
• The platform, and how much of the template layer can genuinely be changed
• How many collections have to be created against how many merely need fixing
• Product content production, meaning photography and original copy, which is usually the largest line and the one most often missing from a budget
• Markets and currencies, and whether cross-border work from Toronto into the United States needs original content per market instead of a copy
• How much your team implements against how much we do

On timing, meaningful movement lands between four months and a year. Coverage and crawl fixes can show inside weeks, because removing a blockage is faster than earning a position. A promise of page-one placement within a month means the terms in question have no buyers behind them. A six-month minimum applies to ongoing programs, and the accounts, the data and the files stay in your name from the first week. Very large estates get scoped as enterprise SEO, and a fixed-scope review is a much smaller purchase that will tell you whether a program is warranted here at all.

On a broad head term where a marketplace, a big-box retailer and the manufacturer all sit above you, usually not, and the useful move is to stop spending there. Pretending otherwise is how a year of budget disappears into position nine.

Give the other side its due first. Marketplaces hold decades of accumulated authority, review counts in the thousands per item, a delivery promise almost nobody can match, and a name people search directly and then search inside. A system weighing that evidence is behaving sensibly when it puts them first. Any agency claiming a content plan overturns it is selling optimism.

Three places the advantage flips.

1. Depth inside a narrow range. A marketplace carries everything and understands nothing. A retailer holding ninety variants of one product type, with genuine specification detail and staff who know the difference between them, wins the specific queries a generalist cannot serve. The narrower the query, the weaker the marketplace becomes.
2. The questions that come before the purchase. Compatibility, sizing, materials, care, installation, and which of two models suits which situation. Marketplace pages answer these badly because they exist to transact. Answering them properly earns the visit before the comparison even starts, and most stores skip it because it does not look like commerce.
3. Demand for your own name. That kind of direct demand is the cheapest traffic in commerce, and it exists only if a shopper already knows who you are. That is a positioning problem more than a search one, which is why brand strategy sits upstream of this work instead of beside it.

There is a decision most stores avoid having. If you also sell through the marketplace, your listing there competes with your own site, and the marketplace usually wins that fight. Whether that is acceptable is a question about margin and control, and it belongs in the consumer marketing work before anybody optimizes a page.

The honest ceiling. Some categories are effectively closed. A commodity product with no exclusivity, no range advantage and no brand demand can be technically flawless and still lose every commercial term to a marketplace holding a thousand reviews. When a category reads that way to us, we say so in the first conversation and the recommendation becomes paid, retention and brand work instead of a search retainer.

There are four cases where we recommend against buying this, and all four get said out loud before anybody signs anything. Search is the slowest lever a store has, so pulling it while something further down the funnel is broken wastes the money twice.

1. Traffic is already adequate and the store loses people. Sending more visitors into a checkout that leaks multiplies the loss instead of fixing it. A UX audit of the path from collection to confirmation returns faster than any ranking gain, and the two get confused constantly because both show up as a conversion rate.
2. Nobody searches your category by name. New product types, invented categories and impulse products get discovered in feeds and from creators instead of in a search box. Demand of that kind has to be manufactured before anything can capture it, which makes paid media and creative the honest first spend and search a slow second channel.
3. A replatform is already scheduled inside two quarters. Auditing an architecture already queued for deletion bills real money against an outcome that cannot arrive. Review the incoming build while somebody can still change the address structure and the filter system.
4. The arithmetic does not close. Take realistic incremental sessions, your conversion rate, your average order value and your gross margin, and work out what the program has to produce to pay for itself. On a store with a small basket and thin margin that number is sometimes unreachable at any position, and we show you the calculation instead of talking around it.

A fifth case deserves naming, which is urgency. If the business needs revenue this quarter, none of this delivers it, and agencies that pretend otherwise lose the account in month five anyway.

The smaller version is worth spelling out, because outright nothing is rarely the right answer either. A fixed-scope piece covering crawl control and collection architecture, delivered in a few weeks with no retainer attached, leaves you with the two decisions that matter most and your own team executing them. Plenty of stores buy exactly that and come back a year later, which is a reasonable way to run a business. Where the constraint turns out to sit somewhere else entirely, the rest of the search practice or another discipline is the right conversation and we would rather have it early.

Almost none of them, and getting this call right is worth more than every other technical decision in a catalogue put together. Faceted navigation means the filters a shopper uses to narrow a listing, and on most builds every selection produces a new address.

The arithmetic is what makes it dangerous. Five filters with six values each, combinable in any order, generate tens of thousands of addresses out of one collection. Add sorting and pagination and a store with two thousand products can hand a search engine several hundred thousand near-identical pages. Crawl effort then goes on colour-and-price combinations while the collections that earn revenue get requested every few weeks.

Three tiers, decided once and written into the templates.

Indexed. A single filter value with genuine search demand behind it and enough products to justify a page. Category with a brand, a colour, a size or a material, where people actually type that phrase. These stop being filtered views and become real collections with their own title, their own opening and their own internal links. The test is whether you would build the page even if the filter did not exist.

Reachable and not indexed. Combinations a shopper legitimately uses and nobody searches for. The links keep working, products stay discoverable through them, and the pages themselves are kept out of the index.

Shut off from crawling. Sorting, pagination parameters on filtered views, session and tracking strings, and any combination of three or more facets. None of it serves a searcher and every request spends budget you needed elsewhere.

Two things worth knowing before anybody acts. Preventing a crawler from requesting an address does not remove an address already in the index, because the instruction to drop it has to be read on the page, so the order of operations decides whether this works. And the platform sets the difficulty. On Shopify filtering runs through tags and metafields with parameters attached, which limits how far you can restructure and makes this largely a configuration exercise. On WooCommerce a filter plugin will produce parameter combinations without any ceiling, and the theme is usually where the problem was introduced.

At sites at that scale this is a rules problem instead of a cleanup task, because anything cleared by hand regenerates within a month.

You do not fix four thousand product pages by hand. You fix the two hundred that can earn something, then fix the template that generates the rest. The starting problem is nearly always the same, which is that the descriptions came from the manufacturer.

Every retailer selling that item publishes the same supplier paragraph. From a search engine's position there are forty pages saying identical things about one product, and the site with the most authority wins by default. Your page gives no reason to be chosen, and it gives a shopper nothing to decide with either.

How the work gets sequenced.

1. Segment the catalogue by what each product could realistically earn. Demand on the model name, margin, stock reliability, and whether the item is exclusive to you. Most catalogues split into a small tier worth individual attention and a long tail that only has to be correct.
2. Rewrite the top tier properly. Specifications written as text instead of trapped inside an image, what the product suits and what it does not, sizing and fit, a comparison against the obvious alternative, and the questions your support team answers about it every week.
3. Make the template earn its keep across the tail. A template that assembles a genuine description from structured attributes beats both the supplier paragraph and a thin sentence written under protest.
4. Treat reviews and customer questions as content production. Customer language on the page is original by definition, it covers phrasings no copywriter would predict, and it accumulates without a brief. It is the only content in a catalogue that scales for nothing.
5. Get the structured data and the feed agreeing. Product and offer markup carrying price, currency, availability and condition, review markup only where the reviews are real and visible on the page, and the same facts in your feed. A feed claiming one price while the page shows another gets listings disapproved and gets trust withdrawn quietly.

The feed deserves its own attention. It is usually the most complete machine-readable version of your catalogue, and on most accounts nobody has audited it since launch, so a wrong attribute travels into surfaces you never look at, including the AI search optimization answers now recommending products to buyers. In regulated categories the copy carries risk alongside opportunity, so for health and wellness brands claim language gets flagged for your reviewer before it ships. Schema and template changes need a deploy, which is why the development side is in the room for this.

An address that spent three years earning positions and links should not vanish because a shelf is empty. Stores delete product pages as a housekeeping habit and lose visibility nobody ever connects back to the deletion.

The policy has to be written once and handed to whoever manages the catalogue, because these decisions get made daily by people who are not thinking about search.

Temporarily unavailable. Keep the address live and indexed. State plainly that it is out of stock, give a return date where you have one, offer a notification, and put the closest alternatives on the same page. Availability in the structured data has to match what the page says, since a mismatch gets the listing pulled.

Seasonal. Keep the page and its collection live all year. A page deleted in February and rebuilt in October starts from zero every cycle, while the same address held for four seasons compounds. Writing something genuinely useful about the season is what makes the collection rank before the season opens instead of during it, and brands with seasonal ranges live or die on that timing.

Discontinued with a direct replacement. A permanent redirect to the successor, which carries the accumulated signals and lands the shopper somewhere they can still buy.

Discontinued with nothing comparable. Redirect to the parent collection, where the underlying intent is still served. Where the page holds real external links and still receives traffic, keeping it live with alternatives on it is frequently worth more than any redirect.

A variant gone while the product continues. Nothing happens to the product page. Removing a size should never remove an address.

The equity point is the one that gets missed. Product URLs collect references from reviews, roundups, forums and social posts over years, and that kind of earned link is the hardest thing in search to replace. Deleting the page throws them away silently. A thousand dead product addresses returning soft errors also teaches a crawler that your catalogue is unreliable, which costs you attention on the pages that are still selling.

Worth adding that the customer decides whether any of this pays off. UX research on what somebody does when they land on an unavailable item usually shows that a clear alternative recovers the sale and a dead end loses it.

Almost never, because the constraint people blame on the platform is usually a theme or configuration decision, and a migration risks far more than the constraint was costing. Platforms do shape what is possible though, and understanding those limits before anybody proposes a fix saves a quarter of wasted argument.

On Shopify. The address structure is fixed. Products sit under /products/ and collections under /collections/, and you cannot nest a collection inside a folder path or place a product inside its category. The same product is also reachable through a collection path, so canonical handling matters more here than elsewhere. Filtering runs on tags and metafields, and raw server logs are not available to you. What most people overestimate is the cost of those fixed paths. A flat structure with deliberate internal linking competes perfectly well, and the hierarchy people want in the address bar can be expressed through breadcrumbs, navigation and collection nesting instead, which is how we structure a Shopify build from the first sprint.

On WooCommerce. You control the permalink structure completely, so categories can nest, the product base can be removed, and the addresses can be designed to match how people shop. You also inherit raw logs, full template access, and every performance and security obligation that comes with them. That freedom is where the damage usually originates, since a filtering plugin will generate parameter combinations without any ceiling and nobody notices until a crawl returns four hundred thousand addresses. So a WooCommerce build gets its filter rules written before launch and not after.

When replatforming genuinely is the answer, the reasons are commercial. The checkout cannot do what the business needs, integrations keep failing, nobody on your side can maintain the stack, or the total cost has stopped making sense. Search is rarely the deciding factor and it is always a constraint on how the move gets executed.

What protects a move once it is decided.

• A benchmark taken before anything moves, covering index coverage, positions, revenue and how the site is currently crawled
• Every collection and product address mapped one to one, including discontinued items still receiving traffic
• The facet architecture decided before launch, since inheriting the old parameter mess into a new platform is the most common own goal in commerce
• Feeds resubmitted and re-approved, because a mismatch after cutover suspends listings in the exact week you cannot afford it
• Verification on the live domain within the hour, then coverage watched against the benchmark for a full quarter

Where the move is the actual project, this work sits inside the website redesign engagement instead of alongside it.

Reporting runs against organic revenue, orders and margin by collection, and never against impressions or sessions. A search report leading with traffic, in a business that takes payments all day, is avoiding the only question anybody should be asking.

What arrives every month.

• Organic revenue and orders, held apart from branded demand. Growth on your own name tracks everything else the business does. Growth on category and product language tracks this work specifically.
• Revenue by collection and by product set, so a particular investment can be judged instead of absorbed into a total. This is the figure that says whether building six collections was worth the money.
• Contribution after margin wherever we can see it, since a collection producing volume at nine points is a very different result from one producing half as much at fifty.
• Positions across the tracked set, including the terms we deliberately left alone and the reason for leaving them.
• Index coverage, meaning how many collection and product addresses are indexed against how many should be, which on a catalogue is the earliest warning signal available.
• Crawl distribution, where logs are obtainable, so you can see whether requests are landing on revenue pages or on filter combinations.
• Appearance in generated answers for the product questions your buyers ask, reported as share across a fixed set because there is no position to report.
• Policy exceptions, meaning unavailable, discontinued and orphaned pages that drifted outside the rules agreed at the start.

Two limits get stated plainly instead of smoothed over. Your platform's revenue figure and your analytics revenue figure will never agree to the dollar, so one gets nominated as the source of truth at the start and the gap gets explained once instead of relitigated monthly. And last-click reporting undervalues organic in any store where paid closes the sale, which is why first-touch and assisted views sit beside it from the beginning.

The commitments behind all of it are the same ones that hold everywhere in the monthly search program. Conversion tracking gets verified before any number is judged. A disappointing quarter reaches you from us with the diagnosis and the change of plan attached, before you find it in a dashboard. And where organic runs alongside email, paid and retention, those numbers get read together, because the rest of a consumer program decides how much of a first order is even attributable to one channel. Send us your numbers and we will tell you plainly which parts of the current reporting are flattering you.