Start with the next question your team must answer
A nonprofit rarely begins with a shortage of information. It begins with information in inconvenient places: a program description in one document, photographs on someone’s phone, attendance records in a spreadsheet and a useful story in an email. The immediate temptation is to upload everything. That may centralize storage, but it does not automatically make the material usable. A better starting point is a question someone needs to answer: What does this program do? Who is it for? What happened during the reporting period? What can we responsibly tell a donor?
Choose one upcoming task, such as a small donation campaign or an application narrative. List the statements that task will need. For each statement, identify the source and the person who can explain it. This creates a practical evidence boundary. You are organizing information for a decision, rather than building an archive that no one knows how to search.
In EvidofOS, the organization profile, programs and Evidence workspace play different roles. The profile establishes identity and purpose. A program record explains a body of work. Evidence supports particular facts. Keeping those roles clear helps colleagues understand why a document belongs in a workflow and what it can actually establish.
Separate activities, outputs and outcomes
Consider a fictional community reading program. A description says that it offers weekly sessions. A schedule supports when sessions were planned. Attendance records may support how many sessions occurred and how many people attended. None of those documents automatically proves an improvement in reading ability. That would require suitable outcome evidence and an explanation of how it was measured.
This distinction matters because compressed campaign writing can quietly change a claim. “We offered reading practice” becomes “we improved literacy,” even though the source has not changed. The second sentence may sound stronger, but it carries a different factual burden. Evidence review should examine meaning, not simply check whether the same words or numbers appear somewhere in an attachment.
Label goals as goals, completed activities as activities and measured outcomes as outcomes. Include the reporting period. When a source contains a count, ask what the count means: individual people, visits, households or registrations. Repeated visits should not silently become distinct participants. A carefully defined, modest statement is more useful than an impressive statement your team cannot explain when a funder asks a follow-up question.
Create records people can recognize and reuse
Use descriptive source titles. A title such as “Reading sessions attendance, spring reporting period” tells a colleague more than “Final document 3.” Include the available date and context fields, connect the source to the correct program, and keep a clear distinction between internal material and information suitable for publication. The goal is that someone who did not collect the evidence can understand why it matters.
Start with a small working set. Add the current program description, the records supporting the specific campaign claims and any approved public material you intend to show. Reopen each record after saving. Check that the source belongs to the selected organization, that its title is understandable and that the relevant passage can be located. Upload completion is only the beginning of review.
If Program Intelligence proposes structured facts from a document, compare those facts with the source before confirming them. A suggestion can save transcription work while still misreading a date or combining unrelated passages. Keep uncertain information visible. When a necessary fact is missing, assign a follow-up to the person who knows it instead of filling the gap with plausible language.
Handle stories and photographs as evidence with permissions
Stories and photographs can help people understand a mission, but availability is not the same as permission. An image on a staff member’s phone may contain someone who never agreed to public use. A participant’s story may be appropriate for a private service record but inappropriate for a fundraising page. Decide who can approve the intended use before selecting media for a campaign.
Add accurate descriptions and choose images that support the actual work being discussed. An unrelated photograph can mislead even when the caption contains no incorrect number. Avoid implying that a pictured person experienced an outcome that is not documented. Where permission is uncertain, use an approved image of the setting, staff or materials instead of inventing consent.
EvidofOS provides local image uploads in relevant profile and campaign editors, so a team does not need to find a public image URL first. That convenience does not remove the review step. Check the uploaded file, its description and the audience that will see it. Keep private supporting documents private, and select only suitable public evidence for donation pages or social content.
Use the same evidence differently for different audiences
A funder, a donor and a social audience may need different levels of detail. The underlying fact should stay consistent. A proposal may describe the reporting method and limitations in a paragraph. A donation page may use a short supported statement with a relevant report. A social post may explain one aspect of the program and link to the campaign. Shortening a statement should not broaden its meaning.
Before generating content, choose the sources relevant to that particular objective. Providing every document can introduce conflicting periods, unrelated programs and details that are not intended for publication. A focused selection gives reviewers a clearer basis for checking the result. When several sources disagree, keep the disagreement and resolve it with the responsible colleague; do not choose whichever number produces the most persuasive caption.
Grounding findings are useful signals in this process. If a draft cannot be validated, inspect the claim and source relationship. Correct the wording, select legitimate supporting evidence or remove the claim. Save and revalidate the changed version. Repeatedly clicking Approve does not solve an evidence problem, and a fluent AI draft is not proof that the underlying facts are correct.
Make maintenance part of the operating routine
An evidence library becomes less useful when no one owns its maintenance. Choose a practical review rhythm for the material your team uses most often. Recheck time-sensitive records before an application or major publication. Confirm whether a description still reflects the current program, whether a reporting period is still relevant and whether a public link still points to the intended document.
Preserve the distinction between updating a working record and changing an already published campaign. A profile correction should prompt review of affected public material, not an assumption that every saved page has already been republished correctly. Use the relevant review and publication workflow, keeping version history where the system provides it.
A useful first milestone is deliberately small: one program, a few reliable sources, one reviewed campaign and a colleague who can trace its claims. After that works, expand the method to the next program. You will learn which fields matter, which sources are repeatedly requested and which gaps deserve attention. The result is more than a tidy folder. It is a shared basis for explaining your work, preparing funding requests and communicating with supporters without rebuilding the facts each time.