12,462 posts across 38 brands. 223 of them are ours.
Hairy Housemates sells corduroy dog beds into Spain and Canada. It is a Shopify shop, it takes money from real customers, and it belongs to us. It is also where Proofloop, the content engine we sell to other people, runs against a live competitor set.
Two figures rather than one, because one on its own would mislead you. The big number is the whole tracked set, 38 brands. The small one is this shop's own posts inside it. The engine started indexing in July 2026, so the evidence base is large and only weeks old, and you can ask us to open it on a screen.
PROOFLOOP 38 BRANDS, ONE OF THEM OURS INDEXED SINCE JULY 2026 COUNTS AS AT 2026-08-09
On this page
- What these numbers measure
- The business
- What we built
- What the engine does, with the counts
- The review gate
- What a growth figure would have to survive
- What this proves
- Where this stops
- What running it taught us
- How to check any of this
- If you want it on your own business
READ
What these numbers measure
OPERATIONAL SCALE NOT CLIENT RESULTS INDEXED SINCE JULY 2026
Every figure on this page measures what the system processed. We label them operational scale wherever they appear, because they count work done and not money made.
Three things about the counts before you read them. We print all three, because a figure you can take apart is worth more than one you have to believe.
The big number belongs to the set, not to this shop. The 12,462 posts in the index span all 38 brands the engine tracks. Hairy Housemates is one of those 38, and 223 of the posts are its own. A reader who takes 12,462 as one small shop's output has been misled, and the only reliable way to prevent that is to print both figures next to each other.
Most of the index arrived at once. 9,047 of the 12,462, which is 72.6% of it, came in as a backfill of older posts over two days at the end of July rather than accumulating while the engine ran. The database marks those rows as historical, which is how we can tell you the split rather than guess at it.
The engine is weeks old. The first post was indexed on 2026-07-09, and the first on this shop's own account on 2026-07-20. So the number measures the size of an evidence base. We use it as scale, which is the one thing it can carry: not a rate, and not an operating history.
"This engine has read 12,462 posts across 38 brands" is a fact, and we will open the record it comes from. "This engine grew the shop" would be an inference dressed as a measurement, so we keep the two apart.
We build it, we run it, then we sell it. Every count on this page comes out of a system we own and run, which is why we can hand you the record rather than a summary of it.
The business
Hairy Housemates makes corduroy beds for dogs and sells them into two countries, with two languages, two logistics realities and two different competitive sets. It is small, it is real, and the people running it have other jobs.
That makes it an honest test bed. Everything that goes wrong for a small brand goes wrong here too. The category is loud: pet brands post constantly and plenty of them post well. Content is somebody's third job. And the failure mode is the week the operator gets busy and everything stops.
The question that stalls a shop like this is not what to say. It is what to make next, and how anyone would know. That is the question the engine was built to answer.
Figure: the Hairy Housemates shop, live. Caption: The shop is ours, which is the entire reason this study can show you the counts instead of describing them.
What we built
Three things, and they are not the same kind of thing.
The brand and the shop. Identity, art direction and the store a customer touches. The studio's older discipline doing work that is still load bearing: a content engine pointed at a shop nobody wants to buy from produces expensive nothing. See web design and branding.
Proofloop, the content engine. It watches the competitive set, reads what those brands publish, works out what is landing, and turns that into briefs specific enough for someone to pick up and make the thing. This is the product now sold as Proofloop.
The review gate. A person reads everything before it ships. It gets its own section below, because it is the step most systems drop.
We describe what these do for the business. How they are wired is not on this page and will not be on any page.
What the engine does, with the counts
SOURCE: PROOFLOOP RECORD, THE 38 BRAND SET AS AT 2026-08-09
POSTS INDEXED, ALL 38 BRANDS
12,462
Thirty seven competitors and our own shop, indexed since July 2026.
OF THOSE, ON OUR OWN ACCOUNT
223
The gap between the two is roughly what one small brand looks like next to the category it competes in.
OPERATIONAL SCALE · NOT CLIENT RESULTS 1 MARK = 1 POST
-
It watches
38 brands. 122 social accounts. 186 snapshots of the websites behind them.
Thirty seven competitors and our own shop. The set is chosen deliberately and it is public. These are ordinary brands with ordinary accounts, so the raw material is available to anyone willing to spend their life reading it. Choosing which 37 to watch mattered more than anything downstream of it.
-
It reads
12,462 posts ingested and analysed, across all 38 brands. 223 of them from our own shop. 31,070 post performance metrics captured. 1,344 media assets processed.
Those two post counts are the same measure at two scales, and the gap between them is roughly what one small brand looks like next to the category it is competing in.
Reading is the expensive part, and the part nobody does by hand past the first fortnight. A person can watch a competitor. A person cannot watch 122 accounts and hold a year of them in their head.
-
It remembers
10,487 indexed evidence passages.
The least glamorous number on the page, and the one we would defend hardest. Every observation the engine makes traces back to the post it came from. A recommendation you cannot trace is an opinion with a chart on it.
-
It briefs
29 creator briefs produced.
A brief here is a shot list, a script and an edit plan, specific enough that someone can pick it up and make the thing without a meeting first. Output runs in English and Spanish, both handled as first-class rather than one translated off the other.
MEASURE · COUNT · LOG SCALE 12,462 Posts indexed, across all 38 brands- 223 Of those, posts from our own shop
- 9,047 Of those, older posts brought in by backfill
- 31,070 Post performance metrics captured
- 10,487 Indexed evidence passages
- 1,344 Media assets processed
- 122 Social accounts monitored
- 186 Website snapshots across the set
- 38 Brands tracked, ours among them
- 29 Creator briefs generated
- 2 Languages, both systematic
Records by kind1 to 100K
OPERATIONAL SCALE NOT CLIENT RESULTS INDEXED SINCE 2026-07-09 COUNTS AS AT 2026-08-09
Every count on this page is a total, with the date the index opened printed next to it. A monthly rate needs a run long enough to have one, and a rate calculated from a few weeks of a mostly backfilled index would describe the import rather than the engine.
The review gate
Nothing the engine produces ships without a person reading it first.
The tell in machine written content is that nobody had an opinion. Grammar is fine everywhere now. What a system will do, given the chance, is produce the average of what the category already says, forever, and the average of a loud category is invisible.
So a writer signs off. They cut what is correct and dead, they put back the joke the model smoothed out, and now and then they throw the whole brief away because the evidence supports it and their gut does not. That last one is why the step exists.
It is also the step that does not scale, which is why most content systems quietly drop it. We kept it, and it is inside the price rather than sold as an upgrade.
Figure: one of the 29 briefs. Caption: A brief is a shot list, a script and an edit plan. A person signs it off before anyone films anything.
What a growth figure would have to survive
NO GROWTH FIGURE NO ATTRIBUTION THREE REASONS
You came here wanting a growth figure. Here is the bar any number has to clear before it goes on a page of ours.
We would be crediting the engine with what several things did. The brand, the product, the price, the season and the platform all moved in the same window. Any single percentage pulled out of that is a story about attribution dressed up as a measurement.
We built the engine and we run the shop. A number produced under those conditions is us marking our own homework, and you should discount it to roughly zero. So should we.
A result counts against a test signed in advance. Nothing was signed before this engine started, so what came after it is history rather than measurement. That is the argument of our approach page, and it would be strange to abandon it on the one page where a flattering number was available.
So the page publishes the processed record, and we will open it on a screen with you. That is the part that carries a source, and the part that tells you what an engine like this would read in a category like yours.
What this proves
- It proves the system runs on a live business. Money moves through the shop, the competitors did not agree to be studied and do not behave conveniently, and none of it is a demo dataset or a sandbox.
- It proves the evidence base is real and traceable. 10,487 passages, each tied to a public post you could go and read.
- It proves the human step exists. A real cost on a real calendar, not a line in a deck.
- It proves the two languages are systematic. English and Spanish, both handled properly, which is a different thing from running the English through a translator at the end.
READ
Where this stops
Four boundaries, named here because they change what a proof month should test.
- Every count here is a total, not a weekly rate. The index was built since July 2026 and most of it arrived in one import, so the counts tell you how much evidence sits behind a brief. A weekly figure pulled from a run this short would describe the import, so we publish the total and the start date instead.
- Your sector sets the ceiling. Your competitive set may be smaller, quieter, more private, or worse at posting. Fewer competitors means less evidence, and less evidence means the engine has less to say. We map your set on the call and tell you whether the category is dense enough for this to be worth doing.
- Content produces attention, and attention is where this hands over. Turning attention into sales is a different system, anywhere and for anyone. If that part is broken, this will not fix it, and it is the part worth looking at first.
- Your review gate is the one that decides. Ours holds because one named person is accountable for it. Put the same person at your end and the volume stays worth reading. Leave it unread and volume is all you get.
What running it taught us
Three things, all of which changed the product other people now buy.
The set is the strategy. Choosing which 37 competitors to watch mattered more than anything the engine did afterwards. Pick a comfortable set and you get a system that confirms what you already believed, at scale, in two languages.
A brief a human will not pick up is a wasted brief. The early ones were thorough and unusable. They got shorter and more opinionated until people started making things from them. Thoroughness was the problem.
Two languages doubles the review, not the production. Producing Spanish is cheap. Checking that the Spanish says the same thing, in language a native speaker would use, is where the hours go. That is why the review gate is priced in rather than discovered by the client in month two.
How to check any of this
The shop is public. hairyhousemates.com is open, it takes orders, and you can read everything it has published.
The competitors are public. Ordinary brands with ordinary accounts. We will walk you through how the set was chosen and what makes an account worth watching.
The counts come from the engine's own record, and we will put them on a screen and go through them with you, live, on a call. That offer covers every figure on this page, including the three that shrink it: the start date, the 9,047 backfilled posts and the 223. Those three are the reason the rest of them are worth looking at.
If you want it on your own business
The product is Proofloop. It opens with a 30 Day Proof Month: defined output, a pass or fail test signed before the month starts, and at the end you either start the term or stop with nothing owed, keeping the brand voice profile either way.
We will look at your category with you and map what an engine like this would read in it. The contact page also carries a short form if you would rather write first.
Read next
- Case studies, including the four earlier projects, named for the work rather than for a number.
- Proofloop, this engine sold as a product.
- Our method, with the timelines, the named risks and what happens when a test fails.
- Web design and branding, the discipline behind the shop this runs on.