A weekly analytics report site that runs itself, for $0 a month
Raw job data in, a clean bilingual report site out. Every week. No server, no manual step, no invoice.
$0 / mo
Infrastructure cost
Weekly
Fully unattended cadence
20 / 20
Scheduled runs green
13,152
Raw records rebuilt every run
Figures as of July 2026.
The problem
Most teams sit on a stream of raw data that is technically valuable and practically useless. It arrives messy, in files nobody wants to open, and it goes stale the day it lands. Turning it into something a person can actually read, a clean dataset, a chart, a report they trust, is manual work that someone has to remember to do. So it happens late, or inconsistently, or it quietly stops happening at all.
You do not want a dashboard you have to babysit. You want the numbers to be current when you look, without anyone touching them, and you want to trust that they are right.
The constraints I worked under
No infrastructure budget
No server to rent, no database to host, no monthly bill.
Fully unattended
It runs on a schedule and publishes itself. If it needs a human to start it, it drifts.
Trustworthy output
The numbers are reproducible, and the site refuses to publish if anything looks wrong.
The approach, in plain terms
Raw snapshots go in one end. A clean, queryable dataset and a published report site come out the other, on a schedule, with a gate that blocks any bad build from going live.
Under the hood it rebuilds the entire dataset from source on every run, so the result is always reproducible from scratch and there is no fragile running state to corrupt. The written summary is generated once and rendered in both English and Spanish from the same source, so the two language versions can never disagree. A strict build step checks the whole site before anything is published: one warning and the deploy is blocked.
That is the machinery. What matters to you is what it produces and what it costs to keep running, which is nothing.
The outcome, with numbers
Snapshot as of July 2026; the schedule keeps running, so these keep climbing.
- Runs entirely on free tooling. Infrastructure cost is $0 a month. There is no server.
- 20 of 20 scheduled runs have finished cleanly since March 30, 2026, about 15 weeks of fully unattended operation: 16 weekly report bundles and 4 monthly rollups.
- Each run rebuilds 13,152 raw job snapshots (covering March 22 to July 13, 2026) into 3 curated tables and distills the latest closed week to 342 clean, deduplicated records, published as a report and a downloadable CSV.
- Every report ships in English and Spanish automatically, from a single source, so the two never drift.
- Backed by 78 automated tests, with a build gate that refuses to publish a broken site.
What Offer 2 delivers for you
The same engagement, applied to your data instead of a job feed:
- A clean, queryable dataset built from your messy source data, rebuilt reproducibly so you can trace any number back to its source.
- Automated reports in the languages and format your audience actually reads.
- A published site that updates itself on a schedule, with no manual refresh.
- A build that fails loudly instead of shipping wrong numbers quietly.
- Handoff notes so your team can run it, or I keep it running. Either way it costs close to nothing to operate.
One honest note
The first version ran fully stateless, and I had wrongly assumed that rebuilding from source also preserved the published archive. It did not, so I gave the archive a deliberate home and backfilled the history. I would rather flag the sharp edge than hide it.
This is Offer 2 in practice.
If you have recurring data that should be a living report and is not, that is the engagement.
Work with me.
Have a project or idea that needs to become real software? Let's build it together!