Oscar's database is researched, not simply scraped
We use multiple source types — including our own research, published and open data, regulatory information, web research and direct organisational information — but records do not become useful just because they can be found online. People validate, structure, classify and maintain what enters the database.
The web has made information easier to find. It has not made public-sector research simple.
A council may publish one structure in an organisation chart, another set of names on individual service pages and different contact details elsewhere. A trust can rename a directorate without changing the underlying responsibility. A postholder can move organisation while old pages, PDFs and search results continue to exist for months. Two organisations can also use completely different job titles for essentially the same function.
So the research problem is not really “can we find a name online?”. It is “can we work out what the organisation looks like now, who is responsible for what, whether the information is current, and how that record should fit into a structured national database?”
What sources does Oscar use?
Those broad source types can include organisational websites, official directories, statutory publications, public registers, government data, organisational documents, direct confirmation from postholders or organisations, and other legitimate research sources.
The important point is that no single source is treated as infallible. Public-sector organisations change too often, publish information too differently and use too many overlapping structures for that to work reliably.
A raw source can tell us that a person or organisation exists. Research is what tells us whether the record is current, how it should be classified, what responsibility sits behind the title and how it relates to the rest of the public-service landscape.
So what does “human verified” actually mean?
Identify a change or record
Researchers find or receive evidence of a new organisation, new postholder, departure, structural change, renamed team, contact update or other relevant change.
Validate against context
The information is checked against other available evidence and the organisation's known structure rather than accepted as a standalone search result.
Classify what the record means
Organisation type, geography, hierarchy, seniority, function and responsibility are structured so users can search consistently across organisations that describe themselves differently.
Keep following the change
The database is continuously maintained, with known changes handled as they arise alongside scheduled checks and a rolling research cycle.
Why not just scrape every public-sector website?
Because that would create a lot of data very quickly — and quite a lot of problems very quickly too.
| Automated collection alone | Human-researched structured data |
|---|---|
| Can collect duplicated, historic or contradictory records. | Conflicting information can be checked before the record is accepted. |
| Usually captures the words an organisation uses. | Can map those words to standardised functions and responsibilities. |
| May treat every page or title as equally meaningful. | Researchers can interpret organisational hierarchy and decision-making relevance. |
| Can struggle when pages move, documents disappear or structures change. | Historical context helps researchers recognise that a change has happened rather than assuming a record simply vanished. |
| Produces information. | Produces information with context. |
That does not mean technology is absent from our research process. Quite the opposite. Automated tools, APIs, change detection, AI and other technical methods can all help surface information, compare records or accelerate a workflow.
But there is an important difference between using technology to support research and allowing automation to decide what the database says.
The difficult bit is not the email address
An email address is a field. The harder questions are usually:
- Is this still the correct postholder?
- Is the title current?
- Where does the role sit in the organisation?
- What function is the person actually responsible for?
- Is the role senior enough for the audience being built?
- Has responsibility moved to another team or organisation?
- Has an organisation merged, renamed, closed or changed type?
- Should this record be linked to a parent body, trust, group, partnership or wider structure?
Those questions become even more important when someone wants to target by responsibility rather than by a literal job title.
Why historical knowledge matters
Oscar has continuously researched Government and Public Services since 1984. That history matters because public-sector change rarely happens in a vacuum.
When an organisation disappears, a researcher needs to know whether it has genuinely closed, merged into another body, been renamed or transferred its functions elsewhere. When a senior post disappears, the responsibility may have moved rather than vanished.
A purely current snapshot can tell you what is visible today. Historical context helps explain how it got there.
Continuous change handling + a rolling research cycle
Known changes are researched and applied as they are identified. Alongside this, Oscar's wider postholder research operates on a rolling full-review cycle currently described across the site as approximately four to six months. That combination is designed to avoid treating data maintenance as a once-a-year clean-up exercise.
Where does AI fit?
AI can be useful for identifying likely changes, summarising large volumes of material, classifying text, highlighting anomalies and helping researchers navigate information faster. Those capabilities are increasingly useful when dealing with a public-sector landscape this large.
But AI can also confidently connect the wrong person to the wrong organisation, misunderstand an old document, flatten two different responsibilities into the same thing or invent certainty where the source material is ambiguous.
For that reason, Oscar's direction is not “AI instead of research”. It is technology and AI supporting a research system in which people remain accountable for what becomes structured data.
What happens after research?
The raw information is not the finished product. Once validated, it is structured into linked database records: organisations, postholders, functions, seniority, geography, organisation type, responsibilities and other classifications.
That structured layer is what makes it possible to ask useful questions such as:
- Who leads IT across NHS organisations in a particular region?
- Which councils have senior sustainability or estates responsibilities?
- Which academy trusts sit behind a group of individual schools?
- Which organisations and postholders have changed since a previous data delivery?
- Who performs a procurement function even when their job title does not contain the word “procurement”?
In other words, collection gives us ingredients. Research and structure turn those ingredients into something people can actually use.