Best Ecommerce SEO: How to Judge the Work You Are Getting

Quick answer: The best ecommerce seo work is identifiable from its outputs rather than its promises. Look for an index plan, revenue reported by landing page, category pages that changed structure rather than word count, a frozen tracked keyword set, and a written argument about priorities you can disagree with.

Most articles on this subject are about choosing a provider. This one assumes you already have one, or an in-house team, or a freelancer, and that the real question in front of you is whether the work arriving each month is any good. That is a harder question than it sounds, because store SEO is largely invisible from the outside. Nothing looks broken. The reports go up and to the right. The invoices are paid. And it can take three quarters to notice that the visible activity and the commercial outcome have quietly detached from each other. What follows is a way to audit the work itself: the artefacts it should produce, the tests you can run on your own store without tools, and the timeline of evidence you are entitled to expect. It is written to be used on any provider, including the way we scope our own e-commerce SEO engagements.

Quality Is Legible in the Output, Not the Relationship

The single most reliable predictor of a good store engagement is not responsiveness, not reporting frequency, and not how pleasant the calls are. It is whether the work produces artefacts that would still make sense to a stranger.

Good catalogue work leaves documents behind. An index plan, a demand map, a create-merge-filter decision list, a template change log. These are working documents rather than client-facing deliverables, which is why many buyers have never seen one. Ask for them. A provider doing the work has them in some form, even if the form is untidy. A provider who does not will offer a dashboard instead.

Weak work leaves only activity. Tickets closed, posts published, pages “optimised”. Activity is real, but on a catalogue it is also the cheapest thing to produce, because the same effort can be spent on 2,000 product pages via a template or on eleven pages by hand, and both appear in a report as work done.

Conventional wisdom says you should judge an SEO engagement on rankings and traffic. On a store that advice frequently backfires, and for a mechanical reason: both metrics drift upward on their own as a catalogue ages and accumulates URLs, so a programme can post improving numbers while the commercially important pages sit still or decline. Impressions in particular rise when thin filter URLs get indexed, which is the opposite of progress.

The thing you are looking for is a decision you can disagree with. Good work involves choices with costs attached: this category gets merged, this range gets no attention this quarter, these 400 URLs get removed. If nothing in six months of work has been a judgement call that somebody had to defend, then nothing consequential has been decided.

The Five Artefacts That Separate Good Store Work From Ordinary Work

These are the specific outputs catalogue work should generate. Their absence is more diagnostic than any metric.

One: an index plan with a decision per URL template. Home, category, subcategory, product, variant, each filter type, sort, pagination, search results, tag archives, legacy patterns. Four columns: crawl, canonical, internal linking, sitemap. If this does not exist in any form, the most consequential decisions on your store are being made by your platform’s defaults.

Two: a create, merge or filter list. One row per demand cluster with an architecture decision attached and a reason. This is the document that shows somebody looked at your catalogue rather than at a keyword export.

Three: a revenue-ranked product tier. Because nobody rewrites a full catalogue, the honest version of product work names which products get original writing this quarter and which are handled at template level. If your provider claims to be optimising all of them, the arithmetic does not work.

Four: a template change log. Title patterns, heading logic, schema fields, related-product rules, breadcrumb changes, with dates. Template changes are the highest-leverage work on a store and the easiest to under-report, because one afternoon can alter 3,000 pages and looks like a single line item. We have seen change logs that fit on a single page and still explained a year of movement better than any quarterly deck attached to them.

Five: a written interpretation attached to every report. Not a summary of the numbers, an argument about them. Revenue fell 9 per cent because two categories were deindexed after a theme update, here is the fix and here is when it lands. An automated dashboard is not a report, it is a data feed with a logo on it. The interpretation is the part you are actually paying a professional for, and it is the part most commonly missing. The same standard applies to how we present technical SEO findings, because a crawl export without an ordering is just 400 problems.

Read Your Last Three Reports Like an Auditor

Put the last three months side by side. Nine tests, none of which require any tooling.

Has the tracked keyword set changed? Compare the lists. If terms have quietly appeared or disappeared between months, the improvement you are looking at may be selection rather than progress. A frozen core set plus a clearly separate additions list is the honest structure.

Is organic revenue reported by landing page? Not sessions, not conversions in aggregate. Revenue, attributed to the page that brought the visit in. If this is absent on a store, the reporting cannot distinguish a good quarter from a lucky one.

Is indexation reported by template type? Aggregate index coverage is nearly useless on a catalogue. Knowing that 94 per cent of products are indexed while 38 per cent of categories are not is the whole insight.

Does anything in the report go down? Three consecutive months where every number improves is a reporting selection, not a business. Real quarters contain declines, and a provider who surfaces them before you notice is showing you how they will behave when something breaks.

Are the categories named? Generic reporting talks about “key pages”. Useful reporting names the eleven categories that carry your margin and reports each one.

Is there a crawl allocation figure? What proportion of crawl requests land on URLs you intended to publish. On an untreated store this is frequently under half, and it is one of the few numbers that improves quickly when the work is real.

Does the work planned match the work completed from last month? Drift between the two is normal once. A pattern of it means the plan is decorative.

Is the SGD value of anything stated? Revenue, average order value, cost per acquired order. A report with no currency in it is describing search rather than your business.

Would a new hire understand it? The report should survive a change of contact on either side. In our experience this is the test that most reporting fails, and it is why we publish the way we work on our about page rather than keeping method knowledge inside individual relationships.

Seven Things You Can Check on Your Own Store This Afternoon

None of these require access to anything your provider controls.

Search for your five biggest categories by name plus a qualifier. If a filtered or parameterised version of your own site appears instead of the clean category URL, your signals are split and the index plan is either missing or unenforced.

Count your indexable URLs against your product count. A rough proxy is the URL count in your sitemap files versus the number of sellable items in your admin. A ratio above roughly 5 to 1 on a store under 2,000 products deserves an explanation. When we audited local catalogues on this single check, we found it predicted the state of everything else more reliably than any other one-minute test.

Open three category pages and read what is above the grid. If there is nothing, or if there are 800 words below the grid instead, the category layer has not been worked on. The category layer is where the type demand lands.

Check one product that went out of stock last quarter. Does the page still exist, is it still indexable, does it say when the item returns, does it link to alternatives. Four scenarios, four correct handlings, and the default behaviour of most platforms is wrong on at least two.

Compare a displayed price with the structured data. Use any rich results testing tool on a product page and check the declared price and availability against what the page shows. Mismatches are the most frequent cause of lost rich results we encounter on Singapore stores, and they almost always appear after a promotion feature is added.

Look at a seasonal range from last year. If the URL was deleted and recreated, you are starting from zero every cycle and inheriting nothing.

Click from the home page to your deepest profitable subcategory. More than three clicks on a mid-sized catalogue means the navigation is organised for the warehouse. The same check applies to any listing-driven site, which is why the handling in our car dealer SEO results write-up transfers almost directly to a product catalogue.

Weak, Adequate and Strong Versions of the Same Deliverable

Every line item below appears in almost every store engagement. What separates the best work is not whether the line exists but which column it sits in.

Most agencies will happily sit in the middle column for years, and the middle column is not fraud. It is competent, defensible and roughly what the market price buys. The gap between middle and right is not effort, it is whether anyone has looked at your specific catalogue closely enough to make a decision they would have to justify. That gap is also where the pricing difference between bands is genuinely earned rather than assumed.

The Evidence Timeline You Are Entitled To Expect

If month three still contains no shipped template change, something is wrong regardless of how good the reporting looks. Front-loaded technical work is the defining shape of a catalogue engagement, and a programme that is still in analysis at the end of quarter one has either hit an implementation blocker nobody escalated or is not doing the work. An independent second opinion is cheap at that point, and it is a common reason stores commission a standalone SEO audit and consulting piece while a retainer is running.

Singapore-Specific Tells

Some failure patterns are local, and they show up in the work rather than in the numbers.

No position on the marketplaces. If nothing in a year of work has addressed what your own domain should do differently from your Shopee and Lazada listings, the strategy has not engaged with your actual competitive situation. Most local retailers of any size sell on both, and that fact should change the content plan.

Advice imported from large-market playbooks. Crawl budget projects on 400 product stores, publishing cadences designed for markets with ten times the query volume, and head terms chased at a regional level when the delivery footprint is local. A cluster with 180 monthly searches here can be worth more than a 2,000 volume term elsewhere, and a provider who does not price that in is optimising for the wrong scoreboard.

Prices, currency and shipping handled as a design detail. SGD on the page, in the markup and in the feed, with shipping cost and delivery windows visible. Currency switchers added post-launch are the most common source of structured data mismatches we see locally.

No sector re-weighting. A fashion catalogue, an electronics catalogue and an industrial parts catalogue need the same framework weighted three different ways, which is the point of the sector work on our industry SEO pages. Identical treatment across very different ranges is a template being applied rather than a store being read.

Our ecommerce case study shows what a report worth reading tracks. Alongside traffic, it reported product page indexation (68 of 200 product pages, or 34%, at baseline and 95% by Month 9), Core Web Vitals status, category pages with content (0 to 15), page 1 keywords (12 to 74) and organic monthly revenue (S$8,400 to S$28,600). If your reports show impressions and rankings but never indexation by template or revenue by page type, you cannot tell whether the work is reaching the pages that sell, and rising impressions can hide filter URLs entering the index.

What To Do When the Work Is Falling Short

The default reaction is to change provider. That is often the wrong first move, because the replacement inherits the same undocumented store and the clock restarts.

Ask for the artefacts in writing. Index plan, architecture decision list, template change log, revenue by landing page. Give a date. What comes back, and how fast, tells you most of what you need.

Change the reporting definition before changing the provider. Freeze the tracked set, require revenue by landing page and coverage by template. In our experience this single change fixes the work more often than replacing the people does, because a programme that has to report revenue stops optimising for the metrics that flatter it.

Re-scope rather than re-hire where the gap is capacity. Many engagements are failing because the technical enforcement sits with a developer who was never briefed, not because the search thinking is weak. That is a scoping problem with a cheap fix.

Get an independent read before deciding. A short paid diagnostic from a second party, SGD 800 to SGD 2,000 for a focused fortnight on a mid-sized catalogue, is a fraction of a quarter’s retainer and it resolves the argument with evidence. What you are looking for is disagreement between the two views, because convergence tells you a problem is obvious and divergence tells you where judgement lives.

Field notes: In our B2B ecommerce case study, the reporting measured what the business cared about, not just visibility. For a wholesale kitchenware supplier, the headline figure was monthly trade enquiries from organic, which moved from 5 to 28 over eight months, and by the end organic generated 19% of all new trade relationships. Traffic (470 to 1,476 monthly visitors) and page 1 keywords (5 to 32) were reported too, but as supporting numbers. When we review someone else’s work, that is the first thing we check: whether the reporting connects pages to commercial outcomes. If it does not, neither side can say which pages are earning, and the argument about the work cannot be settled with evidence.

Our Take

The best ecommerce seo work is boringly identifiable once you know what to look for. It produces documents with decisions in them, it reports revenue rather than inputs, it changes templates rather than individual pages, and it tells you about the bad month before you find it yourself.

So judge the work, not the relationship. Put three reports side by side and run the nine tests. Spend an afternoon on the seven on-site checks. Ask for the five artefacts by name and see what comes back. None of that requires technical knowledge and all of it is more informative than another review site.

The contrarian point is that most underperforming store engagements are not staffed by bad practitioners. They are competent work aimed at the wrong layer, usually because content is the easiest line to sell and the easiest to show, and nobody ever forced the reporting to answer a commercial question. Fix the measurement and a surprising proportion of the work fixes itself. If you want a sense of what the corrected shape looks like in practice, our e-commerce SEO results case study is written around the baseline as much as the outcome.

Frequently Asked Questions

How do I tell whether my current ecommerce SEO is actually working?

Look at three horizons rather than one number. At six to ten weeks, indexation coverage by template and crawl allocation should be moving, because those respond fastest. At four to six months, non-branded entrances to category pages should be rising. At nine to twelve months, organic revenue by landing page and branded demand are the honest measures. If the only thing improving after two quarters is impressions or tracked rankings, check whether thin filter URLs have entered the index, because that inflates both while earning nothing.

What should a monthly store report contain at minimum?

Organic revenue and transactions by landing page, indexation coverage broken out by template type, crawl allocation across URL patterns, ranking movement on a frozen tracked set, the work completed against the work planned, and a written interpretation of anything unusual. The interpretation is the part that matters. A dashboard export tells you what happened; a report tells you why, what it costs, and what is being done about it. If nothing in the report ever declines, the reporting is selecting rather than measuring.

Is a lot of published content a sign of good work?

Not on its own, and on a catalogue it is frequently a sign of misplaced effort. Content is the easiest line item to sell and the most visible to a client, which means it tends to expand to fill whatever budget is not firmly allocated elsewhere. Four guides a month that link properly into categories will usually outperform twelve general posts. Check whether the content connects to the ranges that carry your margin, and whether the category layer was fixed before the publishing started.

My rankings improved but revenue did not. What does that mean?

Usually one of three things. The ranking gains are on terms with little commercial intent, which happens when the tracked set was chosen for winnability. The gains are on pages that convert poorly, which is a merchandising question rather than a search one. Or the tracked set has changed composition over the period. Check each in that order. The underlying fix is almost always to attribute revenue by landing page, which converts the argument from opinion into arithmetic.

How much of the work should be technical versus content?

It shifts across the engagement rather than sitting at a fixed ratio. A reasonable shape for a mid-sized Singapore catalogue is roughly half technical and architecture work in the first quarter, dropping to a quarter or less by month six as content and authority take over. What should worry you is a constant ratio. A programme that reports the same split every month for a year is running a template rather than responding to what the store needs.

Should I expect to see actual documents, or just reports?

You should expect working documents to exist and to be shareable on request. An index plan, an architecture decision list and a template change log are internal tools rather than polished deliverables, so they may be untidy, and that is fine. What matters is that they exist and that decisions in them are dated and reasoned. A provider who can only produce client-facing decks is either doing the thinking somewhere you cannot see it or not doing it.

How do I judge work on a very small catalogue?

The artefacts shrink but do not disappear. Under about 200 products you should still see an architecture decision list and a revenue-ranked product tier, though the index plan may be a single page because there are few filter permutations to control. Judge small-catalogue work mainly on category differentiation, product tiering and local trust signals. If a provider is selling a crawl budget project on a store this size, they are solving a problem you do not have.

Is it worth paying a second party to review the work?

For retainers above roughly SGD 2,500 a month, usually yes, and it costs far less than a change of provider. Budget SGD 800 to SGD 2,000 for a focused two-week diagnostic on a mid-sized catalogue. Brief the reviewer identically to how you would brief a new provider and look for disagreement with your incumbent. Convergence means the priorities are obvious and your provider is on them. Divergence tells you exactly where to ask harder questions.

What are the fastest signals that something has broken?

A sudden drop in indexed category pages, a spike in indexed URLs with parameters in them, structured data warnings appearing after a promotion or theme update, and revenue falling on a single landing page while the rest hold. All four are visible in Search Console and your analytics without specialist tooling. Set alerts on category page coverage specifically. Theme and plugin updates break category indexability more often than anything else on a store.

Should I switch provider if the work has been flat for six months?

Not as the first move. Ask for the five artefacts in writing with a date, change the reporting definition to revenue by landing page and coverage by template, and get an independent paid read. Flat quarters have three common causes: work aimed at the wrong layer, technical enforcement blocked at the developer queue, and a commercially weak keyword set. Only the first is a reason to change provider, and the other two follow you to the next one.