Explain the number before comparing it

M-12 reports 60 people and 140 attendance entries. Without definitions, a network dashboard might display two measures of reach or add them together. With definitions, it can show unique people separately from repeated participation.

Write a short dictionary entry for each important measure. Include what is counted, who is included, the period, how it is obtained and which source supports it. State whether the value is exact, estimated, unknown or awaiting review.

Keep the dictionary small enough to maintain. Start with the measures used in the pilot and expand when a new question needs them. A long catalogue that contributors cannot interpret is not an improvement over inconsistent spreadsheets.

How different instruments support a common comparison

Start with a question the participating teams genuinely share. For each required field, define the unit, eligible population, period, response options, missing states and calculation. Record how each local field maps to that definition, who reviewed the mapping and which version applies.

Member A asks about workshop participation; Member B asks about mentoring. Both can supply an organization identifier, reporting period and a carefully defined count of unique participants. Their activity questions can differ. Attendance occasions remain a separate measure.

For example, a local field called visits can map to attendance occasions when both count each attendance. It cannot map to unique people without evidence that identifies and removes repeated attendance. If that evidence was never collected, label the unique-person result unavailable rather than asking AI to reconstruct it.

Before combining records, check coverage and overlap as well as definitions. State which teams supplied usable evidence, which are pending and whether the same person or transaction appears in more than one source. Report separate results when populations, methods or permissions make combination inappropriate.

In a Sopact workflow, the shared definitions provide context for analysis while the connected record retains local responses, files and history. Your team can review a source-to-field mapping instead of redesigning every local survey. Test that configuration with sample records before relying on a combined result; a data dictionary cannot supply missing evidence or make incompatible measures equivalent.

Use a common core without erasing local work

Northstar’s members may run different activities. Agree a common core only where it supports a real network question. Keep activity-specific information when it helps the member or explains the result, rather than forcing every program into one outcome measure.

The voluntary study has a different dictionary: question wording, response scale, respondent role, wave and interpretation notes. An organizer’s expenditure document uses financial periods and units. These sources can share context without pretending to measure the same thing.

A theory of change or other framework can explain why evidence matters. It does not automatically validate a measure or permit aggregation. Record the reasoning that connects the measure to the question, including what the evidence cannot establish.

Check who is represented

Suppose only part of the network submitted a complete return. A total from those members is not automatically the total for the whole network. Show invited or eligible members, received submissions, usable values and pending review as appropriate.

The same applies to themes. If a group of comments awaits language review, they are not confirmed absences of the theme. State the reporting base and pending count. If people can report more than one theme, explain why percentages may not add to a single whole.

Before comparing members, inspect differences in period, activity, coverage and method. A benchmark can be inappropriate even when both values are valid. Returning each member’s own reviewed history may be more useful than an unsupported ranking.

Decide what happens when definitions change

If a measure changes, retain the old and new definitions with effective dates. Decide whether a comparison is valid, needs qualification or should stop at the change. Do not silently relabel old results under a new meaning.

Ask two people to use the dictionary on the same fictional evidence. If they choose different units or populations, revise the entry. A worked example and a non-example often clarify meaning better than a lengthy definition.

The dictionary becomes part of the analysis context, including any instructions given to an assistant. A reviewer should be able to see which version informed a finding. The goal is more reliable interpretation, not a promise that a dictionary eliminates every error.

Work through the Northstar example

This is a fictional teaching example. Adapt the fields and rules to the question your own network needs to answer.

Add this part to your plan

Fill workbook page 5. Define people reached and attendance entries. Add one rule for incomplete network coverage and one for a changed study question. Map one local field to a common definition. Give one example that must remain separate.

Self-check: Could two colleagues apply your definitions to the same evidence and explain any remaining disagreement?

Questions you may have

Does a common framework mean all members report identical measures?

No. Use shared measures where they answer a valid common question and retain useful member-specific evidence. Explain which results can be combined.

Can we compare an estimate with an exact count?

Sometimes with appropriate context, but do not hide the difference. Record the method and uncertainty and decide whether the intended comparison is defensible.

Apply the method to your work

Use the five-part plan to assess the sources, analysis and permissions your team needs. The solution page shows where a connected platform can support that workflow.