Methodology

How every figure in this report is produced, what it excludes, and where it should not be trusted.

What is counted

One row of the timings table is one group of completed applications, grouped by service and by the VFS Global centre the application was attended at. A national row — labelled All UK centres — aggregates every centre together for that service.

For each completed application the dataset holds two dates and nothing else that reaches this report: the date the application was submitted, and the date the outcome arrived. The elapsed time is the difference between them in whole days. We do not adjust, trim or smooth it.

The statistics

Median: The middle value: half the applications in the group completed faster, half slower. The median is used rather than the mean because a single application stuck for months would drag a mean somewhere no family will recognise.
90th percentile (p90): Nine applications in ten completed at least this fast. This is the number to plan against if you have a fixed date to meet. The median tells you the typical case; the p90 tells you roughly how bad the tail is.
Sample size (n): How many completed applications sit behind the figure. It is printed next to every number, always.

The rolling window

Figures cover a rolling 90-day window ending on the date the file was last updated. Applications completed before that window fall out of the calculation. This keeps the report describing how things are working now rather than how they worked last year — consular processing changes, and an average over all time would hide that.

One exception is described under backfill below.

The minimum sample threshold

No median is published for a group with fewer than five completed applications. Below that the number is noise, and it is also a privacy risk — the smaller the group, the closer a published figure comes to describing one identifiable family.

Below the threshold, the row does not disappear. It shows a counter:

OCI, Birmingham — collecting. First median publishes at 5 completed applications (currently 3).

Hiding the row would make the report look fuller. Showing the counter is what makes the figures beside it worth believing, and it tells you honestly how much data exists.

Because a national group reaches five completions sooner than any single centre, national rows publish first and per-centre rows follow.

The accuracy figure

Separately from timings, we publish per service how many applications we completed and how many were rejected because of something we filled or prepared. Rejections caused by a government decision, or by a fact the applicant gave us incorrectly, are not in that column — they are real, but they are not what the figure claims to measure.

This figure publishes only once at least 20 completed applications sit behind it for that service. It is the number that would substantiate the rejection guarantee offered at homeberry.ai, so it is never estimated.

Anonymisation

No personal data is ever published on this site, and none is stored on it. This is designed to be enforced structurally rather than promised — and until the aggregation is running, nothing writes the file at all:

The public file holds counts and day-figures only. Its field list is fixed by the schema recorded in the file itself: service, centre, window, median days, p90 days, sample size, status, threshold, source, totals and a timestamp (plus, before launch only, a flag marking the layout-preview rows). There is no field a name, reference number or document could occupy.
The aggregation is designed to run on a separate machine, in a separate private repository that is not connected to any hosting provider. Raw job records never cross into the repository this site is served from.
The aggregation is to validate its own output against that field list before writing, and to abort without writing if anything unexpected appears — a malformed aggregate fails loudly rather than leaking quietly. The page enforces the five-application rule for itself as well: a row below the threshold renders as “collecting” whatever the file says.
The five-application threshold prevents a figure from describing an identifiable individual.

This site sets no cookies and runs no third-party analytics. Besides its data files it loads its stylesheet and icons from homeberry.ai and its web fonts from Google Fonts, and nothing else.

Backfill and provenance

Automated logging will begin part-way through the life of the business. Applications completed before then will be reconstructed from our own records as four fields only — service, centre, submission date, outcome date — and included in the dataset tagged backfill, outside the 90-day window, and counted separately.

We say so plainly because early figures will rest partly on them, and because provenance you can see is what makes a dataset citable. A backfilled row is marked as such in the table.

Limitations

Read these before quoting a figure.

This is not a random sample of all UK applicants. It measures applications prepared by one company. Our applications are checked before submission, so they may complete faster than applications submitted with errors. Treat these as times achievable with a correctly prepared application, not as a national average.
Small samples move. A median over six applications can shift substantially when the seventh lands. That is why n is printed beside every figure.
We measure elapsed time, not government workload. The clock includes appointment availability, postal time and anything else that happened in between. It is what a family experiences, which is the useful thing, but it is not a measure of any authority's internal processing speed.
Consular processing changes. Fee schedules are revised, centres move, requirements shift. A figure is only as good as its date, which is why every one carries one.
Nothing here is legal or immigration advice. It is a description of how long things took.

Corrections

If a figure here looks wrong, tell us and we will check it and correct the source file: WhatsApp +44 7352 674636 or info@homeberry.ai. Corrections change the data file, which changes the report page automatically — the figures are never copied by hand.