Data-driven strategy · thistletavern.com
Our first review looked at your site from the outside. This one starts from your own Search Console and GA4: where you already win, where you lose clicks, what search earns today, and the order we would fix things in. Your stated aim is for half of the business to be online. This is how search gets it there.
Google clicks to the site each month, November 2025 to October 2026. Clicks more than tripled, with the biggest step in July 2026.
01 · What the data shows
Four views of the same site: how Google sees it, how it performs, what organic search earns and where its authority sits. Two terms to know: click-through rate (CTR) is the share of people who see you in Google and click, and position is where you appear, with 1 at the top.
Internal links to the broken events page (92) and room booking page (11), from the 10 November proposal.
Pickup availability fragments our crawler found, with no title or content. The theme requests them when a product page loads.
Googlebot fetches the first 2MB of a page’s HTML, uncompressed. Anything past that point is not read.
Extra addresses created by the rarity filters.
The addresses Google has not indexed are about 6.6 times the sitemap, which suggests most come from links, filters and variants rather than from pages you list on purpose. On a Shopify store of this size that is common. What matters is which of them are normal and which are costing you.
Pages Google has indexed. More than the sitemap lists, so some indexed pages are not in it.
Exactly 100,919, across 11 reasons, grouped into seven below.
20,600 of the 121,500 or so addresses Google knows of. Many of the rest are duplicates that are meant to be ignored.
Indexed pages fell from about 32,000 in early September to 20,600 (read from the chart). Was that a clean-up, or pages being lost?
Source: Search Console, Indexing, Pages, “Why pages aren’t indexed”, screenshot of 17 November 2026. Google’s labels in order: alternate page with proper canonical tag, crawled currently not indexed, not found (404), blocked by robots.txt, discovered currently not indexed, excluded by noindex, other. The seven groups add up to 100,919.
39,615. On Shopify this is mostly the same product reached through a collection address or a variant link, pointing back to the main page. Normal. The fix is for the theme to link to the main product address (remove the collection-scoped product link). They are already canonicalised, so we expect little harm.
36,126. Usually thin or near-identical pages. The screenshot shows validation as failed, which would mean Google re-checked and still left them out. We would confirm that in Search Console.
18,688. About 18% of Googlebot’s 1.89 million requests in 90 days hit a 404, roughly 340,000 visits. That is about 18 visits for each of the 18,688 listed addresses, so we would also look for addresses made by scripts or old links that are not listed.
5,732. Your rules keep Google out of these. A blocked address can still be listed without being read, so we check that none are pages meant to sell.
The URLs Google holds speed data for (371 on mobile) are good, the product details it has read are valid, and Merchant Centre has your delivery policy, returns policy and store ratings set up. We do not know whether the 76 oversized collections are in that sample, which is why move 3 still matters.
Good URLs (Core Web Vitals), with 0 poor and 0 needing improvement.
Good URLs, with 0 poor and 11 needing improvement.
Valid items, 0 invalid. 257 have no ratings or reviews markup, which should only be added from genuine reviews.
Valid items, 1 invalid (missing image). 136 have a barcode (GTIN) problem (130 invalid, 6 not globally valid). 497 lack shipping and returns fields in the markup, which is optional when Merchant Centre already has them.
About 33,200 items approved and visible (likely variants, not products), 132 not approved for reasons we have not yet seen. Delivery policy, returns policy and store ratings are all set.
The main sitemap was read on 16 November and lists 15,233 pages. A second entry, /a/llms, has not been fetched since 9 July and shows 0 pages. It is not a sitemap, so it should be removed from that list.
1.89 million requests in 90 days: 66% for HTML pages, 78% answered OK, 18% not found, 4% other client errors. Average response time is 328 ms. On this volume, crawl budget is probably not the limit, so we say the clean-up may help rather than will.
Source: the first review crawl behind the 10 November proposal, carried over unchanged. Everything under “What Google reports” is from your Search Console and Merchant Centre screenshots on 17 November 2026. Structured data counts are the items Google has read, not the whole catalogue.
Google clicks per month
Source: Search Console, thistletavern.com, Web search, 17 July 2025 to 15 November 2026. Open dots are partial months (July 2025 and November 2026 each have 15 days of data). We have not yet identified what changed in July 2026. Section 02 sets out our theories.
Google clicks in the period, from 1.8 million impressions. Two thirds of the clicks (67%) come from mobile.
Of clicks land on the homepage (22,690 of 77,165). No other page tops 5%.
Of clicks, at least, come from searches for your own name (14,966 across the 27 “thistle” queries in the top 1,000). The rest is searches for what you sell, plus anonymised queries Google does not show.
Palworld TCG sealed (3,106 clicks), Gundam singles (1,977), Riftbound (including two Japanese-language pages, 1,683 and 1,024), Riftbound starter decks (951) and Shadowverse Evolve singles (1,052) sit alongside the homepage and Commander decks (1,946). Your team credits optimised product naming for the Magic gains. Collection pages for newer games are now where much of the growth comes from, and that is the pattern to repeat.
The UK is 69% of clicks. The rest, 31%, is already overseas.
| Country | Clicks | Impressions | CTR | Avg position |
|---|
Source: Search Console, 17 July 2025 to 15 November 2026. Query tables omit anonymised queries, so real totals are slightly higher than the sums of the listed queries.
GA4 total across every channel, 17 Nov 2025 to 16 Nov 2026, not yet reconciled to Shopify. Late-2025 months may be inflated (section 02).
£133,046, or 12.8% of the total. Organic is the third-largest channel by revenue.
Free product listings are counted as a separate channel by GA4, so the organic total could be read as higher.
Revenue from visits sent by AI tools, across 16 purchases. A floor, because AI tools often hide where a visit came from. Worth tracking from now.
Purchase revenue by channel, 12 months
The nine largest channels. Four smaller ones (organic social, paid social, AI assistants and organic video) add about £9k, which brings the total to £1,038k.
Treat the split as provisional until section 02 is settled. Paid shopping and paid search together bring in £204k, more than organic, with cross-network (usually Performance Max) adding £137k. All of it is exposed to rising ad costs. Closing part of the gap between clicks and purchases on the pages that already rank is the most direct use of this work.
The channel split is a guide, not a ledger. GA4 gives £303k to Direct (no referrer was passed) and £111k to Unassigned (no channel rule matched). Section 02 covers why.
Search Console clicks (17 July 2025 to 15 November 2026, 16 months) beside purchases GA4 credits to organic visits that started on the page (12 months). Not a conversion rate: clicks and GA4 sessions are measured differently and the windows differ. Use it to see where revenue lands.
| Landing page | Google clicks | Purchases | Revenue |
|---|
Top items by revenue over the 12 months. £31,646 of revenue has no item name, so it is left out.
| Item | Units | Revenue |
|---|
Source: GA4 property 402616116, 17 November 2025 to 16 November 2026, GBP. No sampling reported. Event values were unusable and are not quoted. Purchase tracking appears to have changed in June 2025, so earlier comparisons are not reliable.
Referring domains per page, Ahrefs
Among the 25 pages Ahrefs ranks highest in organic search, the homepage has 1,095 referring domains (other sites linking to it). The next-best has 2, and most have none. The Commander collection ranks with no links of its own, which suggests it is carried by the strength of the site around it.
That makes internal linking the cheapest lever you have: it moves the strength you already hold down to the pages that sell. It also makes the question “are those 1,095 links good ones?” worth answering before anyone builds more.
Source: Ahrefs Site Explorer, top pages, UK, 15 November 2026. Link quality has not been reviewed.
02 · Does this look right?
Some figures in your data look odd. Each has two possible explanations: it is right and there is a reason, or it is wrong and we fix it. For each we give our lead theory and how we would settle it, usually in one call with your developer.
Every priority and every result we show you rests on these numbers, so settling them early protects the plan.
03 · The core range
We understand Flesh and Blood is the game that drives your business, and, by your account, that its 5,000-product collection breaks the standard filters. Here is what Search Console shows for it, and where we would start.
That is a 0.89% click-through at a blended average position of 4.2. A blend can hide a top spot for a few searches and a low one for the main “flesh and blood” search, so we cannot yet say it is low. We would split it by search first. We would also decide which address should own the head term: this page, or /collections/flesh-and-blood and the singles collection. It is your most-seen Flesh and Blood page, so it is where we begin.
9.5% of clicks to your top 1,000 pages go to Flesh and Blood pages. That is 176 pages and 6,553 clicks. One tile is one percent. It is a floor: it only matches addresses containing flesh-and-blood, fab or armory (excluding some Lorcana “fabled” pages), and players also search for hero and card names.
Flesh and Blood searches in the top-query data, grouped by what the player wants. Anonymised searches are not included.
Source: Search Console, Web, 17 July 2025 to 15 November 2026. Flesh and Blood pages are those with flesh-and-blood, fab, or armory in the address, within the top 1,000 pages. Searches come from the Flesh and Blood search export plus the “fab” searches in the top-queries export, and the three lanes are our grouping. Counts are approximate for these reasons.
04 · Opportunities
The proposal called these “beyond the fixes”. With your own data in hand, each one now has evidence behind it, and a fifth has appeared.
Pages and searches where Google shows you often but few people click. Two tables below: pages first, then searches.
Proposal: Rankings audit ↗“pokemon cards” is searched 148,000 times a month in the UK. Search Console shows 654 impressions for it across 16 months. Board games (24,000 a month) and Pokémon booster boxes (16,000) are in the same position.
Proposal: Content gaps ↗By your account, Flesh and Blood can run to around 5,000 products in one collection, which breaks Shopify’s filters. Its two playmat searches alone show 7,126 impressions, so there is steady demand at the edges.
Proposal: Fixes 5 and 6 ↗New-game collections already drive six of your top ten pages. A plan that runs two to three months ahead of releases makes that deliberate.
Proposal: Opportunities, continuously ↗31% of clicks come from outside the UK: the US (5,187), Japan (5,158), Canada, France, Germany, Australia and Singapore. Your 70 Japanese-language pages in the top 1,000 earn 4,243 clicks.
Pages with at least 20,000 impressions and a click-through rate under 3%, excluding the contact, events, store and catalogue pages that people reach by searching for your name. Some of these are snippet problems and some are position problems, so we split them by search before changing anything.
| Page | Impressions | Clicks | CTR | Avg position |
|---|
Source: Search Console, Web, 17 July 2025 to 15 November 2026.
A selection of 14 of the 27 searches with at least 1,700 impressions where you average position 4 to 11. The others are mainly Riftbound, Palworld, Gundam, Commander and Japanese-language variants. Combined impressions for the selection below: . Filter by theme.
| Search | Impressions | Clicks | CTR | Avg position |
|---|
Source: Search Console, Web, 17 July 2025 to 15 November 2026. Average position over the whole period.
“thistle tavern edinburgh” already earns 1,281 clicks, and Pokémon searches for Edinburgh sit around position 7 to 8. We need to confirm whether you have a presence in Edinburgh. If so, a page per shop, built for its own local searches, is the usual fix.
“pokemon cards aberdeen” averages position 3.3 with a 10% click-through rate, which is what a well-matched page looks like. The search table below starts at position 4, so this search is not listed in it. Calls, directions and map views do not show in Search Console; Business Profile Insights has them. It is worth repeating for Edinburgh if that shop exists.
Searches in Japanese, such as the Riftbound terms, bring 1,854 clicks from just 44 queries. The pages exist and already earn clicks. Whether those visitors can buy and have orders shipped to them is worth a closer look.
05 · The strategy
Grouped by when each can start and what it unlocks. Open any card for the reasoning and the link back to the proposal. The impact and effort ratings are our judgement from the first review and your data, and clean GA4 revenue will re-rank them.
06 · Roadmap
The same rhythm as the proposal: clear specs for your developers, a re-scan of the site to check the work, then the next group. Numbers refer to the twelve moves above.
The baseline is set before any work starts, so every result can be shown against it.
| Measure | Source | Baseline |
|---|
07 · Next steps
Thank you for the Search Console and GA4 access. This page runs on your own figures. The checks in section 02 are the one thing still to settle, and the order of work will be weighted again by what each fix turns out to be worth.
Moves 1, 2 and 10 can start straight away. They need no design change and no new content, only clear specs for your developers.
One person we can send redirect lists, robots rules and template changes to, and who can tell us what is easy and what is not.
They shape the local pages, the overseas work, the release calendar and the GA4 fix. They are below.
Indexed pages fell by about a third from early September. We do not yet know whether that was a clean-up or a loss, and settling it is the first job in move 2. 29% of your clicks ride on one page. Every new set that launches without a ready page is a window you do not get back.
Every move starts from Search Console, GA4 and Ahrefs. We re-scan the site after each step to confirm it worked, fix a baseline before any work starts, and send one monthly report you can hold us to. We do not forecast revenue; we report against the baseline.
By week 2, the indexing drop explained, redirect and robots specs delivered and blank pages closed. By week 6, collections readable and new titles live. By week 12, shop and overseas pages live, first guides published and a monthly report running.
We will walk through this page, settle the checks in section 02, and agree the order of work.