Build a weekly spreadsheet digest for Slack with Make

Fix the reporting boundary

Use Monday 00:00 inclusive through the next Monday 00:00 exclusive in Asia/Tokyo. For a controlled trial, fix the period at 2026-10-19 through before 2026-10-26. A rolling “last seven days” window changes on replay. Include the fixed period and a version in report_id, and verify the sheet’s time zone and timestamp format.

Prepare records with known inclusion rules

Create request_id, received_at, category, status and summary columns. Include two records inside the week, one exactly at its lower boundary, one at its upper boundary and one outside. Expect only the three eligible records. Preserve source IDs in the digest so the recipient can reconcile the report without relying on a paraphrased total.

Search the relevant rows

Configure Google Sheets > Search Rows with the intended date types and period conditions. A prevalidated report_week text column is another option when date comparisons are difficult to maintain. Choose a limit appropriate for the expected maximum and treat reaching that limit as a possible incomplete result, not proof that every matching row was retrieved.

Aggregate the lines before sending

Add Tools > Text aggregator, select the search module as Source Module and leave Group by empty for one weekly digest. Map a line such as “TEST-W001 | Information request | New,” separated by a newline. Include required IDs and categories in the aggregated text; earlier per-row values are not automatically available after aggregation.

Place Slack after the aggregator

Connect Slack > Send a Message after aggregation. Use report_id and the reporting period as the heading, then map the aggregated text. Sending immediately after the search can create one post per row. For large reports, choose a concise digest and a permission-controlled source link rather than relying on an unlimited message length.

Define the zero-record outcome

To notify on an empty week, disable Stop processing after an empty aggregation and use ifempty to replace an empty aggregate with “No records.” Run a zero-row case and verify one aggregate and one post. If the search stops the route before aggregation, a downstream Router alone cannot recover it; use a supported continue-on-no-results setting or a separate reporting path. Check missing dates separately so excluded bad data is not mistaken for no activity.

Track report versions and delivery

Store report_id, fixed period, retrieved count, message ID and sent_at. Decide whether a corrected report gets a new version or updates the recorded post. If delivery succeeded but the ledger write failed, search the channel for the report ID before sending again. A scheduled execution is not itself delivery evidence.

Validate before enabling the weekly schedule

Verify the known three-row digest, empty results, a reached limit, boundary timestamps and replay. Then set the weekly day and time and inspect the organization time zone and next run. Review acquisition, delivery and bookkeeping in history, and assign an operator to compare the first scheduled report with its source.

• Estimate row-dependent credit usage
• Manage replay with a report ID
• Inspect and recover partial execution