Missing PSC timepoints & insufficient backlog in PSC streaming API

Data Issues
1

Due to the recent and unprecedented ingestion of approximately 5 million PSC records by Companies House over the period of a few days combined with a brief outage on my infrastructure, I realized I am missing about 2 million timepoints from the PSC streaming API. This gap covers timepoints 20_326_972 (published 2025-10-31) to 22_865_092 (published 2025-11-01).

(See Massively increased rate of PSC events).

Nothing prior to 22_865_092 is available from the streaming timepoint API at this time.

I note that this means there are only about 7 days worth of data available on the streaming API at this point (not really compliant with the published statements of approximately 10 days backlog available).

So two questions:

(1) To Companies House staff: Would you consider putting back online timepoints you’ve pulled more quickly than your guidance would suggest? I can’t imagine I’m the only one finding myself in this situation; API users should be able to rely on published guidance in planning their data acquisition–it’s a bit of a problem if they cannot rely on that. In this instance, putting timepoints back to just before the large burst of filings started (looks like 2025-10-29 from my logs) would be reasonable to address this issue.

(2) Failing that, does anyone else on this forum have timepoints they can share? I have an otherwise meticulous collection of PSC timepoints back to 3_600_000 and companies data timepoints back to 46_200_000 (both dating to June 2022) that I could share in exchange. Send me a PM if so.

Thanks
John

2

Hi John,

I’m having the exact same problem. We use a cron job and timepoint to fetch a batch of updates every hour from the stream rather than keeping instances open 24hrs a day. When implementing this I presumed from the docs that the timepoint param was valid for any event ID 10 days in the past.

However, after this recent increase in events due to legal requirements my hypothesis is that there is a finite number of event IDs in the past that are valid, regardless of time, so when there is a huge influx of events it no longer allows for us to fetch updates after events 10 days in the past. In fact I have run into issues between runs every hour, as the timepoint is no longer valid on the next run.

Other than keep a stream connection open 24hrs a day I don’t know how to resolve this and it kind of defeats the purpose of the timepoint param. If anything goes wrong we are out of luck because we can no longer recover lost stream data.

The ideal solution would be that CH Stream API add a new parameter which allows for a datetime in the past rather than an eventId

3

One hour expiry of timepoints? Which resource type?

4

PSC and Officers we have run into issue over the past 2 weeks. Companies have been fine

5

I don’t pull the officers data, but I haven’t had problems with PSCs in that time frame.