From nine days to an afternoon

Ask an M&E officer what the monthly report takes and you get a number like nine days. Ask what the writing takes and you get a number like two. The other seven are assembly: exporting from the collection tool, reconciling in a spreadsheet, pasting into a document template, rebuilding the charts the template expects.

The assembly line

• Export. A CSV leaves the collection tool, or several do, one per form.
• Reconcile. The spreadsheet merges them, fixes the district spellings, removes duplicates, recalculates the indicators.
• Assemble. Numbers are pasted into the document, table by table, chart by chart.
• Redo. A late submission arrives, or an error is found, and the line runs again from the top.

Each step is ordinary. The problem is that they run every month in full, regardless of how much changed. A quiet month with forty new submissions takes almost as long as a busy one, because the work is driven by the format, not the data.

One dataset, three views

What changes when the indicator, the dashboard and the report read one dataset is structural. The indicator is defined once, with its formula, its disaggregation and its source form. The dashboard is a live view of it. The report is a document that reads the same numbers. There is no export, because there is nowhere to export to. A monitor fixes a submission at ten in the morning and the report draft reflects it at 10:01.

The afternoon is real. It is spent on the parts of a report a machine cannot write: why completion fell in one district, what the team is doing about it, what the numbers do not show. The mechanical layer, the exports, merges and pastes, is simply gone.

The afternoon itself, in practice: the first hour is reading, the indicator page, the districts that moved, the submissions the data quality checks sent back. The second hour is the phone calls, the district coordinator whose number fell, the partner whose attendance file arrived late. The third hour is the writing, and the report is half drafted before the writing starts, because the factual sections assembled themselves from live data. What took nine days took three hours, and the difference was not speed. The difference is that nobody moved a number by hand.

What still needs a person

Being honest about the remainder matters more than promising it away.

• Narrative judgement. Whether a dip is a story or a blip, and whether it needs a paragraph or a sentence.
• Explanation. The system can say the indicator fell four points. Why it fell is a call to the district coordinator.
• Framing. The same number is on track against a revised target or behind the original plan. A donor reads those differently, and choosing is your job.

Caveats, too. The report that goes out on the 5th covers data to the 31st, and two districts had sync delays. Someone has to decide whether to note that, and that someone is a person who knows the field, not the tool.

The test

Here is a test for any tool that promises faster reporting. Ask where the number in the headline table comes from. If the answer involves a file moving between systems, the assembly line is still there, just better lit. If the answer is a query over live submissions, the nine days are gone, and what is left is the part of the report that was always worth your time. Then ask a second question: what happens when a submission is corrected on the 4th, after the report went out. If the answer is a shrug, or a new version of the spreadsheet, you have your answer.