Operations report #1 — July 2026: the real numbers of the fleet we run
First edition of our monthly operations report. Every figure below comes from a measurable API — Google Search Console, our project registry, our server crontabs. Including what didn’t work.
This is the first edition of a monthly exercise: publishing the operating numbers of our own fleet — the sites, data platforms and AI agents we run for ourselves, on the same rails we sell to clients. Most providers publish promises; we prefer meters. Research backs the instinct: a study by Buell, Kim and Tsay published in 2017, Creating Reciprocal Value Through Operational Transparency, shows that letting customers see the work actually being done increases perceived value and trust on both sides of the counter.
The month in numbers
- →74,391 Google impressions over 28 days (June 25 – July 22) across the 14 properties of the portfolio, measured by the Search Console API — including the properties still at zero.
- →515 clicks from organic search over the same window.
- →Three sites above 16,000 impressions each: our boat comparator (20,167), our van comparator (19,394) and our US plumbing directory (16,364).
- →57 projects in the operations registry, 4 servers running the fleet around the clock, and roughly 65 scheduled AI agent sessions per day.
Shipped in July
- →A public operations page on this site — the same numbers as this report, updated at every reading, with the measurement method alongside.
- →An English twin of our Ayurveda guide, launched from zero to 332 impressions per day within two days of indexing.
- →6,900 new data-first pages on our boating and van-life platforms — 3,349 ports and 3,575 rest areas, built from OpenStreetMap data with no empty fields allowed.
- →A reply-tracking pipeline for our outreach: every prospect answer now lands in a CRM table within two hours, with an instant notification.
What didn’t work
- →Our real-estate data site is still at 0 impressions after weeks online: data-first pages only pay once Google trusts the domain, and that trust is earned in months, not days.
- →This very site sits at 480 impressions over 28 days — the cobbler’s children syndrome is real, and fixing it is exactly why this report now exists.
- →One outreach wave hit an email provider quota mid-send; 81 messages failed with an error and had to be rescheduled. The sender now ramps up progressively instead.
The lesson of the month
An unattended metric behaves exactly like an unattended agent: it degrades silently. The single change with the highest leverage this month wasn’t a new model or a new site — it was closing the loop, making sure every number we produce (impressions, sends, replies) is read by someone, every day, with a date on it. That is the operator’s job, and it is the standard we now hold ourselves to in public. Next reading: end of August.
Frequently asked questions
Where do these numbers come from?+
From measurable sources only: the Google Search Console API for search figures, our project registry for the fleet inventory, and our server crontabs for agent sessions. Nothing is estimated, and every reading is dated.
Why publish failures?+
Because a report that only contains wins is marketing, not measurement. The zeros are part of the meter — and they are what makes the growing numbers credible.
Free resource
The self-assessment grid: 20 tasks AI can automate
Sales, admin, support, operations: the 20 tasks AI agents already handle in SMEs — with, for each one, the tell-tale sign that your team is concerned.
Read next
Strategy, costs & ROI
AI and jobs: should your SMB rethink junior salaries?
19 August 2026·5 min read
Strategy, costs & ROI
ChatGPT Business Premium at $125: what OpenAI’s new tier reveals about the real cost of agentic AI for your SMB
12 August 2026·5 min read
Strategy, costs & ROI
Claude Sonnet 5’s promotional pricing ends August 31: what a 50% token price hike changes for your SME’s AI agents
8 August 2026·4 min read