04 · Websites and software
Data products
Your dataset, or a public one, turned into a product people search for and trust.
In short
Every product in my portfolio starts from a dataset: NOAA frost records, a weather model for 380 Polish counties, 194 researched perfumes, 38 chicken breeds. The data is the centre; the pages, tools and articles orbit it.
If you hold data people would search for, or there is a public source nobody has made usable yet, this is the service that turns it into a product.
What's included
Sourcing
Finding the public datasets, APIs or research that answer the question, and checking their licence and quality.
Cleaning and modelling
Import scripts that validate every row, so updates are repeatable and errors are caught before anything is published.
Pages from data
One page per useful thing, such as a species, a breed or a region, and only where the data is good enough to deserve one.
Tools and forecasts
Calculators, scores and forecasts computed from the data, with the method written down.
Sources on the page
Every number can say where it came from, which is what makes people, and search engines, trust it.
Scheduled updates
Jobs that pull fresh data, recompute and publish, within the source's rate limits.
How I run it
The question
What people search for that this data can answer better than anyone else.
Source and licence
Where the data comes from, how fresh it is and whether it may be republished.
Model
Cleaned, validated and stored in Postgres, with the rules documented.
Pages and tools
Templates and calculations, checked against cases you can verify by hand.
Publish and refresh
Launch, then scheduled updates, with index status and traffic watched.
Content Monitor · Site Scanner · Polaris CMS
Run on my own tooling
Data products tend to have hundreds of pages. My tooling is what keeps them healthy at that scale.
Content Monitor
The dashboard every live site is run from
Index status at scale
Search Console URL inspection for every page, rechecked on a schedule, so I can see which generated pages Google accepts.
Quick wins
Pages at positions 4 to 20 with impressions: the ones a better title or a few links could lift.
A link crawl every 12 hours
Orphan pages, pages with fewer than 3 internal links and broken links, flagged across the whole site.
Site Scanner
A 98-item launch checklist, a crawler and Lighthouse
Checks on the templates
A template error repeats on every page it builds, so the crawl and the 98-item checklist catch it once, everywhere.
Polaris CMS
My own CMS, this site included
Articles around the data
Guides that explain the data and link into it, written in Polaris CMS with link suggestions.
Infrastructure
Docker on my own servers, behind Cloudflare
Scheduled jobs
Data refresh jobs run on my servers next to the site, with the database backed up.
Proof
- Project PlantAlong Frost dates for 41,195 US ZIP codes from 6,769 NOAA stations, with the source under every date.
- Project Grzybolog A forecast for 380 Polish counties computed from weather data twice a day, and an atlas of 67 species.
- Project Olfaria 194 perfumes and 125 scored pairings, loaded into Postgres through validated import scripts.
- Project Yolkday 38 breeds, 95 foods and 7 planners, built from 139 sources.
Questions
We have data but don't know what to build. Can you help?
Yes. I start from search demand: what people look for that your data answers. That usually points to a small set of pages, or one tool, to build first.
Is it allowed to publish public data?
Often, but not always. I check each source's licence and terms before building on it and show the attribution the licence asks for.
Won't hundreds of generated pages look like spam to Google?
They do if they are thin. I create a page only where the data says something useful, skip near-duplicates, and watch index status to see what Google accepts.
Related services
All services →Tell me what you need
A few lines are enough: what you want to build, fix or automate, and any date that matters. I reply in English or Polish.