Explain Why a Weekly Total Cannot Always Be Built by Adding Daily Cards
Prepared with AI assistance by Alactic Editorial. Examples identified as hypothetical are illustrative, not customer results.
Your daily cards add up to 84 people, but the weekly report shows 37. Another row adds up to 283 open items, although the week ended with only 36. Neither difference necessarily means the dashboard is wrong.
A weekly figure can be built by adding daily cards only when the cards count separate contributions to the same measure across non-overlapping days. Activity events can meet that condition. End-of-day snapshots do not: they repeatedly observe a changing population. Daily unique-person counts do not either: the same person can appear on several days.
The useful question is not simply, “Why do daily dashboard totals not equal the weekly total?” It is, “What does each card count, and what question is the weekly figure answering?” The filled guide below shows how to answer that before treating a mismatch as an error.
1. Identify what is being counted before doing arithmetic
Start with the unit and the time basis. A label such as “activity,” “open,” or “people” is not enough to establish whether addition is valid.
- Activity count: How many qualifying events occurred during the day? If each event belongs to exactly one day, adding the daily counts can produce the period count.
- Snapshot: How many items were in a particular state at a particular moment? Adding snapshots measures repeated observations, not a distinct inventory or a flow of work.
- Unique-person count: How many different people had qualifying activity during the day? Adding daily unique counts counts returning people again.
These measures answer different operational questions. Activity helps describe the volume of work performed. A snapshot describes the backlog at a moment. A period-wide unique count describes the number of different people reached or served during that period.
All three can be useful in the same report. The mistake is assuming that changing the date range should apply the same arithmetic to each.
2. A fictional week that produces three different answers
Fictional worked example: All figures in this guide are hypothetical and are not customer results. Imagine a report covering Monday through Sunday in one reporting timezone. It contains three measures: qualifying activity events, open items at the end of each day, and people with qualifying activity that day.
For this example, the filters remain unchanged throughout the week. Each activity event is assigned to one day by its occurrence timestamp. Daily unique-person counts and the weekly unique-person count use the same person identifier.
- Monday: 20 activity events; 40 open items at day-end; 12 unique people.
- Tuesday: 25 activity events; 43 open items at day-end; 15 unique people.
- Wednesday: 18 activity events; 39 open items at day-end; 11 unique people.
- Thursday: 22 activity events; 42 open items at day-end; 14 unique people.
- Friday: 30 activity events; 45 open items at day-end; 18 unique people.
- Saturday: 12 activity events; 38 open items at day-end; 8 unique people.
- Sunday: 8 activity events; 36 open items at day-end; 6 unique people.
The activity counts total 135 events. The snapshots total 283 daily open-item observations. The daily unique-person counts total 84 person-days, meaning 84 counted person-and-day combinations.
A separate count across the whole week's underlying activity records finds 37 different people. That weekly count cannot be recovered from the seven daily unique counts alone: the cards do not reveal which people appeared on multiple days.
3. Filled report-reading guide: what to use and what to reject
Use the following guide as a model for a report handover or weekly review. It is filled using the hypothetical week above.
Daily activity events
- What a daily card means: Qualifying events that occurred within that day's reporting boundaries.
- Daily values: 20, 25, 18, 22, 30, 12, 8.
- Valid weekly interpretation: 135 qualifying activity events occurred during the week, provided the days partition the week and use matching filters.
- Invalid weekly interpretation: 135 different people were active, or 135 items were completed. Those are different units unless the report explicitly counts them.
- Check before using: Each event is counted once, all days are present, and daily and weekly views use the same event timestamp and inclusion rules.
- Suggested report wording: “135 qualifying activity events during the reporting week.”
End-of-day open items
- What a daily card means: Items open at that day's closing snapshot.
- Daily values: 40, 43, 39, 42, 45, 38, 36.
- Valid weekly interpretation: The week ended with 36 open items. The mean of the seven day-end snapshots was approximately 40.4 open items.
- Invalid weekly interpretation: 283 different items were open during the week, or 283 items entered the queue.
- Check before using: Every snapshot represents the same cutoff time and comparable scope. The average is an average of day-end observations, not a continuous-time average.
- Suggested report wording: “36 open items at Sunday close; average day-end backlog approximately 40.4.”
Daily unique people
- What a daily card means: Different people with qualifying activity on that day.
- Daily values: 12, 15, 11, 14, 18, 8, 6.
- Valid weekly interpretation: A separate period-wide distinct count shows 37 people with qualifying activity during the week. The daily counts sum to 84 person-days.
- Invalid weekly interpretation: 84 different people were active during the week.
- Check before using: The weekly calculation counts each person once across the full period, using the same identifier and activity criteria as the daily view.
- Suggested report wording: “37 unique people during the week; daily unique counts are not additive.”
This guide deliberately keeps the measures separate. Combining them into one unlabeled “weekly total” would remove the information needed to interpret them.
4. Why snapshots need a question, not just a sum
An open item can remain open through several daily cutoffs. In a separate hypothetical illustration, one item that stays open for the entire week appears in all seven snapshots. Adding those appearances does not turn it into seven items.
Choose the weekly view according to the decision:
- What remains to be handled? Use the closing snapshot: 36 open items in the worked example.
- What did the observed day-end backlog typically look like? Use the mean of comparable daily snapshots: approximately 40.4.
- Which day had the largest closing backlog? Use the maximum daily closing snapshot: 45 on Friday.
- How many different items were open at any time during the week? Use a separately calculated period view of item histories, not a sum of closing snapshots.
The trade-off is visibility. A closing snapshot is easy to interpret but can hide a difficult middle of the week. A daily average shows more of the pattern but can conceal peaks. Even the highest daily closing value can miss a larger intraday backlog.
Likewise, the sum of snapshots can have a specialised interpretation as item-days observed at the chosen cutoff, but it should not be presented as a distinct item count. Use that measure only when repeated daily exposure is actually the question being studied.
5. Why daily unique counts need a period-wide calculation
A daily unique count removes repetition within a day. It does not remove repetition across days.
In a separate hypothetical illustration, Monday's active people are Asha, Ben, and Chen. Tuesday's are Ben, Chen, and Devi. Each daily card shows three people. Adding them gives six person-days, but the two-day period contains four different people.
The period-wide operation is to combine the qualifying people from every day and count each identifier once. It is not to add the daily counts. A weekly report therefore needs access to the underlying identifiers or a suitable period-level calculation; the seven card values alone are insufficient.
There is a useful reasonableness check. With matching scope and complete data, the weekly unique count cannot be smaller than the largest daily unique count or larger than the sum of daily unique counts. In the main hypothetical example, 37 falls between 18 and 84. That makes it plausible, not proven correct.
Do not try to repair the difference with an estimated overlap percentage. Two weeks can have identical daily counts but very different patterns of returning people. If only daily cards are available, report their sum as person-days or state that a weekly unique count is unavailable.
6. When additive activity still fails to reconcile
Even an event count will not reconcile if the daily and weekly reports describe different sets of events. Before investigating a calculation defect, compare these boundaries:
- Reporting dates and timezone. Confirm that the weekly period is exactly the same interval covered by the daily cards. A rolling period and a calendar week are not interchangeable.
- Timestamp basis. Check whether both views use occurrence time, creation time, completion time, or another timestamp. An event can fall on different dates under different choices.
- Filters and visibility. Compare team, owner, status, channel, and other active filters. Also confirm that both readers can see the same records.
- Refresh state. Check whether one view has incorporated late-arriving records or corrections while another still shows an earlier result.
- Completeness and precision. Look for missing days, a partial current day, or abbreviated card values. Rounded display values may not add to the exact underlying total.
For a hypothetical activity total of 135, a weekly display of 133 warrants investigation only after those conditions match. The fact that event counts are additive does not prove that two screens use the same population.
7. Recover from a mismatch without forcing agreement
Preserve the original values and record the date range, filters, timezone, and retrieval time. Avoid changing filters repeatedly until the numbers happen to match; that can erase the evidence explaining the difference.
Then classify the issue:
- Expected non-additivity: Keep the figures, replace the invalid interpretation, and label the period measure clearly.
- Scope mismatch: Align the boundaries and rerun the comparison.
- Unavailable period calculation: Mark the weekly result as unavailable rather than inventing it from cards.
- Unresolved additive mismatch: If authorised record-level detail is available, compare event identifiers for one disputed day first. Look for records present in only one view.
For snapshots, request historical state information if the question concerns items open at any point in the week. A current list of open items does not necessarily reconstruct earlier states. For unique people, request a period-wide distinct calculation rather than exporting more daily totals.
If the report has already been shared, issue a narrow correction. For the hypothetical example: “The previously stated 84 weekly people was the sum of daily unique counts. The period-wide unique count is 37.” Explain the changed interpretation without implying that the daily cards were wrong.
8. Make the next weekly review testable
Before the next review, choose one card from each category and complete the guide with your actual labels, time boundaries, filters, and valid weekly interpretation. Keep unavailable answers explicitly unavailable.
The observable outcome should be a report in which each selected measure has a named unit, a time basis, and a defensible weekly calculation. For an additive measure, record whether the reconciliation difference is zero after scope and refresh alignment. For a snapshot or unique-person measure, record which separate period view answers the business question.
Success is not making every weekly number equal the sum of its daily cards. It is making every weekly statement mean what the reader thinks it means.
Put the next conversation to work.
Explore Alactic Communications for WhatsApp conversations, campaigns and AI-assisted workflows. Choose the product and plan that fit your team.
Explore Communications