Can you add daily GA4 users to get a weekly total?
Why daily user counts do not produce weekly unique users, with a synthetic example and a practical agency reporting workflow.
Do not add daily GA4 user counts to report weekly unique users. The same user can appear on several days. Request the chosen user metric for the whole reporting period, and keep the daily breakdown for the trend chart. Google defines Total users as unique users who triggered an event within the specified date range. The distinct-count definition is why summing daily values answers a different question. [1]
Name the number your client actually needs
An agency account lead may need two useful views: how recorded audience activity varied during the week, and how many distinct users the property reported across that week. A daily chart serves the first question. A period-level user figure serves the second. The total tile should not inherit a spreadsheet SUM simply because the rows contain numbers.
Choose the exact metric before building either view. Total users and Active users have different definitions; Google describes Active users using engagement criteria. Preserve that distinction in the export and client-facing label. [1] Croissant recommends labels such as GA4 Total users, agreed reporting week rather than a bare audience label. An analytics user count also does not establish the number of paying customers or accepted enquiries.
Worked example: one returning user, several daily counts
This is a synthetic counting example, not a customer result or a prediction of exact GA4 output. Imagine an idealized dataset with three consistently recognized user identifiers: A, B and C. On Monday A and B appear, giving two daily users. On Tuesday A and C appear, giving two. On Wednesday only A appears, giving one. Assume there is no activity on the remaining four days.
Adding the daily counts gives 2 + 2 + 1 = 5. The distinct set across the week is A, B and C, so its size is three. The sum of five counts repeated appearances across daily user sets. It is not five distinct weekly users. Repeated presence is enough to explain why these two calculations differ; it is not evidence of broken tracking.
Now remove the identifiers and retain only the daily totals 2, 2 and 1. You cannot recover the weekly distinct count from those totals alone. The overlap could differ while the daily numbers stay identical. Do not ask AI to estimate the missing overlap or subtract an assumed repeat-visitor percentage. Request the period-level measure from the authorized reporting source.
Collect the period figure separately from the trend
For one client, write down the approved property, exact user metric, reporting start and end, governing time zone, filters and collection time. Specify two outputs: a date breakdown for the chart and a whole-period figure for the summary. Give each output an explicit place in the report template.
For an API implementation, Google's runReport method supports requested metrics, dimensions, date ranges and filters. [2] Ask the implementer to retrieve the agreed user metric over the complete period without a date breakdown for the summary, and to retrieve a separate date breakdown for the chart. This is Croissant's proposed collection pattern, not a claim that every connector exposes identical controls.
Store the source request or export reference with each result. A connector that supplies only daily rows has not supplied enough information to construct the distinct weekly figure. Leave that tile unavailable until the appropriate source result exists. A small team may handle this with two checked exports and a spreadsheet; a custom AI workflow is not required to fix a SUM formula.
Keep a mismatch from becoming an invented diagnosis
The toy example uses exact sets to explain the arithmetic. Real Analytics reports have additional limitations. Google documents unique-count approximation, possible sampling and thresholding, reporting identity, and differences introduced by query settings. Its guidance recommends matching dates, dimensions, metrics and filters when comparing API and interface results. [3] A source-reported period count is therefore not a census of identifiable people.
When two whole-period views disagree, compare their saved settings and collection times before explaining the gap. Do not force the daily sum to match by changing individual days. Do not change reporting identity or tracking settings merely to obtain a neater total. A responsible analytics specialist should assess any configuration change separately.
If the source marks a result as limited, retain that limitation for the reviewer. If the period query fails, distinguish collection failure from a reported zero. AI may draft a description of checked figures; it should not turn an unresolved discrepancy into a claim about returning visitors, acquisition performance or lost leads.
Test the report before it reaches the client
Use the fictional A/B/C example as an acceptance case for your template: the trend retains 2, 2 and 1, while the period tile uses the supplied whole-period value of three. Then test a missing period result. The tile must become unavailable rather than silently falling back to five. Also test that an Active users export cannot replace a Total users input without review.
Assign a reporting owner to approve labels, source settings and unresolved differences. Keep client distribution as a separate reviewed step. The existing agency reporting workflow guide covers collection and review more broadly; the conversion-rate guide addresses a different calculation, combining compatible numerators and denominators.
If repeated collection and exception handling justify a connected workflow, use the reporting demo to discuss the intended inputs and review boundary. Bring Croissant the report question and fictional sample rows through the contact form. Leave client records and credentials out of the initial enquiry.
Implementation checklist
- Choose Total users or Active users deliberately and retain the exact label.
- Record property, period, time zone, filters and collection time.
- Collect a whole-period user figure separately from the daily trend.
- Prevent SUM of daily user counts from populating a unique-user tile.
- Test repeated users, a missing period result and a wrong-metric export.
- Retain source limitations and route unexplained differences to the reporting owner.
- Require human review before client distribution or tracking changes.
The sources behind the sauce.
Primary research and official documentation behind the cited points. The workflow examples and checklists are Croissant’s own guidance.