Detail behind sections one, four and seven
Why the numbers say what they say, why the order of the work matters, and what the end of it actually looks like.
The findings page says more people are visiting than ever. That's true of the number, but a good part of the increase isn't people.
In the real world, roughly four people in a hundred browse on Linux. On your site it's twenty-two. That isn't an audience — it's automated software, mostly the machines that gather text for AI. American visits rose from 22% to 33% over the same period, and March had 440 visits from 408 different people, almost none of whom came back.
Once you strip that out
Australian visits have fallen about 11%. The real audience is shrinking slightly. Machine traffic is hiding it.
But this doesn't explain the sales fall. Even measuring only Australian visitors, the share who buy dropped by about 60%. The people are still arriving; they're just not getting through. That's the finding, and it survives the correction.
Two people in a hundred, where eight to twelve would be normal. This has been the same for two years and it is the largest single loss. 874 people looked at a piece this year and didn't put it in a basket. That's a shopfront problem: sold-out work at the top, no prices, Sale badges, no visible categories.
This one is new in 2026 — halved from 30% to 16%. The abandoned-basket figures agree independently: 83% abandoned against a 70% norm. Eighteen people started, fifteen gave up. Something at the checkout got worse between last year and this, and we don't yet know what.
Ceramics are heavy. A $60 mug carrying $20 postage is a third added to the price, revealed at the last step. Sophie charges a flat $18, plus $6 for each extra item, anywhere in Australia — and says so on every product page, not in an FAQ. It's predictable, it's stated before anyone commits, and it quietly rewards buying two or three things at once.
This is why the postage test is the first thing on the plan. Everything after week two depends on whether this hypothesis is right.
Our first assumption was that the expensive work had sold and not been replaced. It hadn't. Here's what's for sale right now, against what traffic actually reaches:
The shop sends people to sold-out and inexpensive things and starves the expensive available ones. Plates is the fourth most-viewed page on the whole site and cannot be bought. This is a sort-order problem and it is free to fix.
Nothing above $180 has sold in 2026. Not because nothing above $180 exists — because nothing above $180 is being shown.
One thing we've walked back. An earlier reading of 2025 suggested sculptural work earned dramatically more per view than tableware. That rested on four products in one year and doesn't hold up in 2026. The safer claim is the simpler one: available, well-presented, fairly-priced work sells; buried work doesn't.
The finding underneath every other finding
Every symptom in this work has the same cause. Signals arrived and nothing was set up to carry them anywhere.
The cause isn't attitude, it's architecture. The page that records what people searched for is three clicks from the dashboard, appears in no summary, is never emailed, and has apparently never been opened. The page showing which products get looked at but not bought sits beside it.
Two years of your customers' requests accumulated faithfully into screens nobody had a reason to visit.
Working on instinct is how most studio potters work, and legitimately so. The distinguishing fact here is only that a machine has been quietly writing the answers down and nobody showed you the page.
The plan shouldn't end at "the website works again". That's a milestone, not an outcome. The outcome is a cycle you run yourself, in about fifteen minutes a month.
Every part of this already exists in your Squarespace account. The "notify me" form is there. Site search has been recording for two years. The mailing list exists. The sales report shows views alongside sales. None of it is connected to anything.
The most durable thing this work produces isn't the rebuilt shop. It's the fifteen minutes a month — and that has to survive after Adam stops.
Backup before anything is written
Squarespace keeps no history for products or styling. No backup, no undo.
The AI visibility check comes before any change
It's the only chance at a clean "before" reading.
Nothing paid until the site converts
Every visit currently produces about 14 cents. A click on Facebook costs 70 cents to $1.50. Every purchased click today would lose most of its cost. The gate is roughly eight times the current conversion rate.
Fix the list's leaks before the first email
A campaign to a still-leaking list wastes the one moment when hearing from you is novel. The leaks are set out here.
Photography last
Forty pieces is the biggest job in the plan, and it's wasted effort before the shop is arranged. The direction for it is here.
Don't touch the soap dish name
It's the only thing currently working. Everything else can be renamed.
The soap dish proves the mechanism and shows its ceiling in the same breath. A good name got you onto two phrases at fifty searches a month each. The phrase most people actually use — ceramic soap dish, eight times the volume — you don't appear for at all, because that page is Bunnings, Bed Bath N Table and Pillow Talk.
Free, immediate, and wins the quieter phrases. Necessary — but on its own it only ever gets the long tail.
Slow, and mostly other people's decisions. But they're the only thing that ever contests the busy phrases.
Neither substitutes for the other, which is why they run as two parallel tracks in the plan rather than one after the other.
It also tempers the butter dish. Eight thousand searches a month will behave the same way — the realistic near-term target is the qualified version ("lidded butter dish", 390 a month), not the headline phrase. Restocking it is still right; expecting the front page in month one isn't.
The soap dish is a cautionary tale as well as a proof. It took 163 views and sold two, because someone searching a $12 supermarket category met a $74 handmade object.
Pursue both, but judge them differently. And note what this does to the mailing-list argument: someone who arrives from "tea strainer" and won't pay $50 today may very well join the list — which is exactly what the longer-term plan needs.
We looked at the whole front page for "ceramics Sydney". Three individual potters rank in the top fifteen. There is no big shop dominating it. That makes it winnable by one person — and you, an actual Sydney ceramicist of fifty-two years, don't appear anywhere.
And the page tells you how to get in. What ranks is markets, associations, council listings and community threads — which is exactly the list of links you don't have.
So the Sydney opportunity and the links work are the same job. The target list is here.
One number worth holding on to: "pottery st ives" has no measurable search volume at all. "Ceramics Sydney" and "pottery Sydney" together are about 650 a month. The positioning word is Sydney.
This isn't "there's upside sitting in your existing traffic." It's that the shop has broken, the decline is still going, and almost all of the repair costs nothing but attention.
Getting back to last year's level is the first milestone, not the starting point.