← All projects

Warehouse monitoring

Making scattered availability readable

Role Research & data collection
Focus Parsing, monitoring
Status Ongoing research
▦
Availability watch Example monitoring view
Illustrative
Storage types Latest observation
Compact Available Available Unavailable
Medium Unavailable Available Available
Large Available Unavailable Available

Example categories and observations

Changes over time
Source checked Availability recorded
Page changed Parser review needed
Collection resumed Observations retained
Public listings → Collection → Human review

A business question before a business decision

I created a monitoring tool while exploring a possible future storage business. Public sites advertised storage availability for offerings of different sizes, but those observations were scattered across separate pages. I wanted to retain a record over time rather than make a decision from a single visit to a listing.

I organised the parsing and collection around that research objective. The system gathers the publicly displayed availability of the offerings being watched. At the time of the interview it had been running for a couple of months, with a longer collection period planned before drawing conclusions.

Collection needs maintenance

The source websites are outside my control. When their pages change, a parser can stop extracting the intended information or interpret a changed layout incorrectly. I use Codex periodically to inspect the collection and correct parsing problems.

That makes maintenance part of the deliverable. A research record is useful only if the repeated observations still mean what the collector expects them to mean. The work includes following the extraction over time, rather than treating the first successful run as the end of the assignment.

Illustrative workflow

From public listings to research

Collect Public availability → parser → observations over time
Maintain Source changes → check extraction → correct parsing
Interpret Listed availability is a research signal, not measured occupancy

Keeping the observation separate from the conclusion

An availability listing describes what a site advertises. It does not directly establish real occupancy, demand or the commercial attraction of a market. The collection phase and the later analysis are separate parts of the project; the full statistical review had not yet been completed at the interview date.

The public case is anonymised. The workflow is straightforward: identify a research signal, collect it repeatedly, maintain the extraction and preserve the boundary between observed data and a future decision. I can build a similarly scoped monitoring tool, with the research question and the meaning of each observation agreed at the start.

Next project Voice routing ↗