291: Subtle Conversion Killers You Aren’t Accounting For With Luke Carthy

Podcast: Download (Duration: 45:45 — 52.7MB)

The biggest quick win most ecommerce stores ignore is their own internal search engine. Luke Carthy finds broken searches at scale by pulling high-volume search terms from Google Analytics, then crawling those search URLs and extracting the HTML to see which ones return no results.

Luke is an ecommerce growth consultant specializing in strategy, search, and conversion, with over 12 years helping brands grow. He has spoken at events including MozCon.

This episode covers how to audit your site search, what to do with empty and thin categories, how to handle faceted navigation for SEO, and why hosting your own coupon code page is worth doing.

Key takeaways

• Broken searches return a 200 status code, so a normal crawl will never flag them.
• Custom extraction finds them at scale by pulling the no-results text from crawled search URLs.
• Screaming Frog costs under 200 pounds and handles this, which puts it within reach of small stores.
• Non-product searches matter: online account, returns policy, customer services all deserve real destinations.
• Search volume is higher on desktop than mobile, contrary to what most people assume.
• Do not canonicalize every filter away. Some facets capture long-tail keywords worth ranking for.
• Structured data does not boost rankings, and price plus stock status in the SERP raises click-through.
• Host your own coupon code page to rank for it, attract links, and reduce cart abandonment.

Why internal site search is a neglected goldmine

Luke sees this ignored at every scale, from national brands to small independent stores.

His example was a major UK retailer where searching Black Friday returned a no results page, days before Black Friday.

Misspellings and stray spaces produce the same failure quietly, across many terms at once.

Fixing it costs essentially nothing and directly affects whether customers can find what they came for.

How to find your broken searches in Google Analytics

Site search reports live under the Behavior tab, provided you have configured Analytics to recognize your search query string in the URL.

That data tells you which terms people search and at what volume.

Volume is what determines priority. A term searched once a month matters far less than one searched 50 times a week, so fix in order of frequency rather than in order of discovery.

What Analytics will not tell you is whether the results were any good.

What custom extraction is and why it matters

Custom extraction pulls any element you specify out of a page’s HTML during a crawl.

The reason it is necessary is that a broken internal search returns a 200 status code rather than a 404. Nothing is technically broken, so a standard crawl reports everything as fine.

Since every store has a predictable search URL structure, you build a list of search URLs from your Analytics terms and crawl them.

What you extract is the no-results text, whatever your store displays when a search fails.

The output is a spreadsheet of URLs alongside the extracted content, which makes hundreds or thousands of broken searches visible at once.

Luke uses Screaming Frog because the barrier to entry is low, at under 200 pounds for a license.

What broken searches actually look like

Enterprise search platforms report failing search terms for you. Shopify and WooCommerce stores get no such report, which is why this matters most for smaller merchants.

One client in the DIY sector had customers searching for measurements using the inch symbol rather than the word inch, and the search engine could not parse it. Fixing that was close to an overnight win.

Part codes cause the same problem in resale businesses. Long manufacturer codes with slashes and unusual characters break searches, since a slash usually denotes a subfolder in a URL.

Accented characters cause it on multilingual stores, where an accented e and a plain e behave differently.

Case sensitivity affects the reporting rather than the search. The same term in different capitalizations appears as separate rows in Analytics.

Why non-product searches deserve attention

Searches like online account, customer services, and returns policy are common and rarely handled.

Luke found a major retailer receiving monthly volume on online account with no results, which is a straightforward fix that improves the experience.

The failure mode worse than no results is bad results. A software seller might return three products containing the word return when someone searches returns policy.

The second-order question is navigational. If people search for these things, your site structure may be failing them, since they would not need search if they could find it.

How to fix search without rebuilding your cart

Plugins for default WooCommerce search were not much help when Luke last checked.

The practical fix is redirects. When a particular search URL fires with a term you know fails, redirect it to a search term that works or to the right landing page.

That gets you the outcome without the dashboards an enterprise search platform would provide.

What threshold to use for fixing broken searches

The number depends entirely on your total search volume.

Ten searches a month means something completely different on a small store than on a large one.

Luke’s suggestion is thinking in percentages instead. If a broken term represents more than 1% of total queries, fix it, and review anything below that manually.

The point is avoiding fixing problems that were not problems while 50 higher-volume searches stay broken.

How to find and fix empty categories

Custom extraction solves this too, and it is a real problem at scale.

Luke has found many categories with no products on large retail sites, which cannot be spotted manually across thousands of categories.

An empty category either needs products added or needs to be redirected to the next best match.

Thin categories with one or two products raise the same merchandising question. Whether the category should exist at all, or whether it should be rolled into a better-suited parent.

Enriching that extracted data with traffic and conversions per landing page turns it into a bird’s eye view of the whole store.

How to handle faceted navigation for SEO

Filters multiply URLs enormously. Televisions filtered by size, brand, colour, and technology produce a large crawl surface with heavy duplication.

The instinct is canonicalizing every filter permutation back to the parameter-free version.

That is a mistake, because it forfeits long-tail keyword opportunities. If colour only exists as a filter, you cannot rank for red men’s t-shirts.

The workable answer is choosing which facets stay crawlable. Colour is usually worth it, brand often is, and size usually is not unless you serve a market where size is the search, like tall or plus sizing.

How to rank a deep category page

Internal link structure is the main lever.

A category four levels deep, like clothing to men’s to t-shirts to designer, will struggle on its own because of its depth.

Linking to those deeper categories from the homepage builds the structure that lets them rank.

Breadcrumbs matter for the same reason, passing equity up and down the category hierarchy.

Landing pages are an underused source of internal links. A Black Friday deals page sits near the top level and can link to categories four or five levels down, lifting them.

What to link to from your homepage

Feature your main departments plus current best sellers and seasonally relevant categories.

Seasonality is the part people neglect. Linking to summer clothing in the middle of a cold season wastes the position.

Google itself treats seasonal terms differently. Turkey near Thanksgiving returns a completely different result set than turkey in August.

Department pages are the strongest internal asset. Men’s attracts far more links and visits than men’s t-shirts, and that equity can be routed onward to the categories you want ranking.

What structured data actually does for ecommerce

Structured data is additional metadata passed to search engines describing your product, category, or page.

It has little to no direct impact on rankings.

What it improves is click-through rate. Correctly implemented product markup passes price, brand, and stock information to Google and Bing.

Stock status is the underrated one. When a page of results all sell the same television, showing price and in-stock status in the SERP reduces abandonment and gives customers what they need before they click.

There are no guarantees. Perfectly valid markup can still fail to display, since Google tightened this after widespread abuse.

Plugins handle the implementation on most carts, though you often need to populate the fields beyond your normal product descriptions.

Why policy pages convert better than product pages

Luke’s view is that returns, reviews, and policy content moves the needle more than product copy.

The reason is anxiety. That content confirms why someone should buy from you rather than a competitor, and that you are a real store rather than two people in a bedroom.

The buying process supports it. Shoppers research the product, find a retailer, then search that retailer’s name plus reviews before committing.

Structure those pages for the questions people actually ask, in an FAQ style. How long do I have to return something, can I return a Christmas gift in the new year, will I get my money back.

A wall of size-8 terms and conditions that nobody can read produces the opposite effect. Clear headings, broken-up sections, and responsive layout signal that you want people to read it.

Why you should host your own coupon code page

Everyone reaching a checkout with a coupon field opens a new tab and searches for your brand plus discount code.

What they usually find is a fake code on an affiliate site, which wastes their time and yours.

Hosting your own page means you rank for that query. People link to it because they feel they have found something others have not.

It also satisfies the impulse. The reward can be free shipping you already offer or 5% off, and the effect is reduced abandonment plus incoming links.

The page can also pass equity onward by specifying which departments a coupon applies to and linking to them.

I built one of these for Bumblebee Linens after fake coupon sites generated a flood of failed checkout attempts, and ranking it top mitigated the problem.

Why coupon codes should be case insensitive

Making a code case sensitive after spending real money promoting it is how you lose the sale.

Someone types it in uppercase, it fails, and they leave.

Beyond case, people mistype codes generally, so choose codes that are extremely easy to spell.

The better version removes typing entirely. A click that applies the coupon automatically.

Frequently asked questions

How do you find broken searches on your ecommerce site?

Pull your search terms from Google Analytics site search reports, build search URLs from them, then crawl those URLs and use custom extraction to capture your no-results message.

Why do broken searches not show up in a normal crawl?

Because a no-results search page returns a 200 status code rather than a 404. Nothing is technically broken, so you have to inspect the returned content instead of the status code.

What tool do you use for custom extraction?

Screaming Frog, which Luke Carthy prefers because a license costs under 200 pounds, putting the technique within reach of small stores.

Should you canonicalize all filter URLs?

No. Canonicalizing every facet away forfeits long-tail rankings like red men’s t-shirts. Keep useful facets such as colour and brand crawlable, and restrict ones like size.

Does structured data improve rankings?

Not directly. It improves click-through rate by displaying price, brand, and stock status in search results, which informs shoppers before they click and reduces abandonment.

How do you rank a deep category page?

Internal linking. Link to deeper categories from the homepage and from high-level landing pages, and use breadcrumbs to pass equity up and down the hierarchy.

Should you create a coupon code page?

Yes. It ranks for your brand plus discount code, attracts links, displaces fake affiliate codes, and reduces cart abandonment from people leaving to hunt for a code.

Do more people use site search on mobile or desktop?

Desktop, in Luke Carthy’s experience. His hypothesis is that mobile users have shorter attention spans and more distractions, so they abandon after one or two failed searches.