Choose a Singapore analytics workshop for mismatched report totals

The problem

A Singapore spreadsheet user creates a new dashboard whose total differs from the monthly report and cannot tell whether the rows, filters or calculation changed.

Name what the total counts

Write the unit beside each number: orders, line items, people, visits or another defined record. Then list the time window and exclusions. Two reports can be internally correct while counting different populations. Start with that comparison before changing a formula. It gives a workshop facilitator a precise question instead of a request to make one unexplained number match another. Put the intended question beside the two descriptions. A count of sign-ups, confirmed sign-ups and participating people can have different definitions even if the dashboard gives them similar names. Keep what one input row represents separate from the unit you want to report. A displayed number does not supply either meaning. Identify whether the total is a count, a sum or another stated operation before selecting an explanation. If the definition is unavailable, record that missing condition rather than inventing a measure that makes the reports agree. This guide prepares one spreadsheet learner’s inspection question; it does not decide the source owner’s reporting rules.

Make one comparison card for each report

Use seven entries: input origin and version; question the result should answer; row meaning and reported unit; observation period; inclusion and exclusion conditions; calculation or aggregation actually used; and result with the steps supporting it. Keep unknown entries visible. A retrieval time and the period described by the rows are different fields; the snapshot guide develops that separate input-version question. Beside each condition, identify its source or the response still needed. Do not replace an unexplained formula with its output label. The two cards should let a peer locate a specific difference without receiving your entire workbook. Use the cards to locate the missing definition while keeping private input within its authorised process.

Keep an exclusion trail

Copy a small practice table and count its rows before and after each filter. Retain the excluded rows in a separate view with a reason, such as an incomplete category or a cancelled transaction. Data.gov.sg's search guide documents filter and field selection on public datasets. Those features can support an exercise, but the meaning of an exclusion remains your analytical decision. State whether a condition concerns the task, a supplied practice requirement or an unresolved choice. Keep the untouched input and the rows examined at each stage. Record a count next to the operation that produced it, so a later person can distinguish a period selection from a status filter. Do not delete uncertain rows to make an attractive final number. A field with an unknown meaning needs clarification; an excluded row needs the reason actually applied. Keep duplicate headers and missing units with their existing specific guides rather than treating every unfamiliar value as the same population problem.

• Data.gov.sg search and filter within dataset ↗ · Source checked 20 October 2026

Compare two wholly invented totals

Every record, report and requirement here is fictional. Five practice sign-up rows have distinct identifiers. R1 is Period Amber, East team, confirmed; R2 is Period Amber, West team, confirmed; R3 is Period Amber, East team, withdrawn; R4 is Period Blue, East team, confirmed; R5 is Period Amber, East team, confirmed. Each row represents one sign-up in this defined exercise. Report A counts all Period Amber sign-ups: R1, R2, R3 and R5 give four. Report B counts confirmed Period Amber sign-ups for East: R1 and R5 give two. Both reported counts follow their stated definitions. Their difference alone establishes no broken formula, actual volunteer turnout or error in a real dataset. The input and exclusion trail make the differing questions inspectable.

Investigate one difference at a time

Align the time window first, then the record unit, then each exclusion. Avoid changing the query, chart and spreadsheet together because you lose evidence about which change mattered. Use invented shop records if your business data is private. Include an example that falls exactly on a boundary so the group can inspect how the rule handles it. Continue the fictional case with an explicit new target: count confirmed sign-ups for both teams in Period Amber. Its rows are R1, R2 and R5, giving three. A’s four includes withdrawn R3; applying the target status condition to A’s original period selection leaves three. B’s two excludes West R2; removing that team restriction while keeping its period and confirmed conditions also gives three. Keep both starting definitions and these separate changes in the record. This completed comparison answers the fictional target on the stated five rows. It is not a general rule that either larger or smaller totals are right, and matching counts do not establish that a real source is complete or accurate.

Keep an unexamined input outside the conclusion

A correct calculation can still concern a different source version, an incomplete extraction or a row unit changed by matching tables. Establish the applicable input before carrying the fictional explanation into your real work. Use the existing first-page retrieval guide for an incomplete API result, the join guide for a measure repeated at another row unit, and the snapshot guide for changing inputs. A public API’s filter capability cannot determine which population your question needs. If you cannot establish one source’s definition or permitted input, state the comparison still missing rather than filling that card from the other report. Keep a proposed check unperformed until it actually happens; no toy arithmetic here is an executed software test.

Pick a session that can examine the calculation

Read whether participants practise cleaning and querying or only learn a chart interface. Ask if the exercise exposes intermediate row counts and definitions. Afterward, write the reconciled definition in plain language next to the report. Search Boardzip for the required exercise, then check current software requirements and access on the organiser's page. Ask about the actual participant action: inspecting supplied rows, tracing a filter, examining your permitted small example or watching a finished dashboard. An unsent enquiry could say, “I can calculate a spreadsheet total, but these two practice reports use different inclusion conditions. Can participants inspect the intermediate rows and explain those definitions? Which setup and material route apply?” Adapt it to your own truthful starting point. The data.gov.sg documentation supplies public-data context, not evidence that a particular class accepts workbooks or allocates individual review. If the offer provides only presentation advice, consider it for communicating an understood result while leaving this population question unresolved.

Retain the working definition and the next missing answer

After a permitted exercise, keep the question, input origin and version, row meaning, period, conditions, calculation and actual observation together. In the invented case, three describes confirmed Period Amber sign-ups across both teams. It says nothing about attendance, unique people, source accuracy or the reason a sign-up was withdrawn. Record any explanation or solution supplied by the facilitator and the part you checked yourself. A chart can then communicate the established question without hiding how the count arose. Use Boardzip’s free data-event search for the remaining task, then verify the particular session’s activity and preparation. Search does not provide an assessment, a repaired workbook or a current place. A clearer definition and an honest unresolved input can be useful learning without an employment or business-impact claim.