Defining the Automated Weekly Self-Review
Treat the review as a weekly data product rather than a journal entry. This systematic aggregation of behavioral data bridges a critical gap—the space between intention and reality. The calendar records what was intended. The task tracker records what was finished. The activity logger records where computer attention actually went.
Establish explicit boundaries for this data product. Monday 00:00 through Sunday 23:59 in the local time zone provides a stable frame. A rolling seven-day window changes during inspection and corrupts the baseline.
Automation removes the friction of manual data entry. The machine handles the collection, leaving only the high-level analysis for the human. Reserve 20 to 35 minutes for this human review. Allocate 5 minutes for data-quality checks, 10 to 20 minutes for pattern interpretation, and 5 to 10 minutes for planning changes.
The Architecture of Personal Analytics
The three-stream model captures a complete picture of a week because each source answers a different question. Calendar records represent commitments. Completed-task records represent declared output. Activity records represent actual attention.
Store timestamps in UTC plus a separate source-time-zone field. Convert them to local time in the review layer. This prevents daylight-saving transitions from creating duplicated or missing local-hour labels. Use three raw tables: calendar_events, task_transitions, and activity_intervals. Retain at least the previous 8 to 12 weekly windows. Recurring Monday or month-end patterns become distinguishable from a single unusual week.
Automated pipelines are strictly dependent on tracking hygiene. Missing calendar blocks, inconsistently closed tasks, or untracked devices remain missing evidence. A perfectly automated system cannot fix inconsistent manual inputs.
Extracting and Routing Calendar Data
Create a scheduled calendar sweep. Assuming a webhook will fire when time passes often fails. Change notifications usually report that an event record changed. An event merely reaching its end time may not trigger a push.
Calculate duration from end minus start after time-zone normalization. Store all-day entries separately. Their date-based boundaries should never be interpreted as 24 hours of work. Fetch records in pages until the response has no continuation token. For an initial backfill, request 28 to 42 days at a time. Checkpoint the final page token so a timeout does not restart the entire extraction.
Map calendar categories through a small lookup table. Use focus, meeting, administration, personal, and unclassified. Store attendee count or salted attendee hashes instead of names when identity is unnecessary.
Historical data extraction is often limited by standard API pagination rules. Forward-looking automation proves more reliable than retroactive scraping. Review the Google Calendar API documentation for specific endpoint behaviors.
Syncing Task Manager Outputs
Subscribe to completed-state changes or poll the task source every 10 to 20 minutes. Write each state transition to an append-only table before deriving a completed-task view. A task might be marked complete, reopened, and completed again.
Capture the task identifier, sanitized title, project identifier, labels, creation timestamp, due timestamp, completion timestamp, estimated minutes, actual minutes, and source revision timestamp. Represent absent effort values as null. Zero means measured work with no elapsed effort. Null means the source never supplied a value.
Deduplicate retries with a key composed of the task identifier, transition type, and transition timestamp. Reconcile the previous 48 to 72 hours during each nightly run to catch delayed mobile synchronization.
Thursday Bram's Task Sync Rule: Task completion metrics measure volume, not necessarily impact. The data must be reviewed contextually rather than as an absolute productivity score.
Capturing Screen Time and Attention Metrics
Install a local or account-based activity logger on each work device. Export timestamped application or window intervals. Categorize records before aggregation. Recoding a weekly total after the fact destroys the granular timeline.
Configure collection intervals in the 5 to 15 second range where the logger permits it. Mark the user idle after 4 to 7 minutes without keyboard or pointer input. Keep the chosen idle threshold stable across comparison weeks.
Use productive, neutral, distracting, communication, and unclassified as mutually exclusive attention categories. Communication is separated because it can support work while still fragmenting a scheduled focus block. Aggregate raw activity into 1-minute intervals for storage and 15-minute bins for the dashboard. Strip query strings and redact document titles before synchronization. They often contain account numbers, search terms, or private project names.
Native OS screen time features restrict direct API exports for privacy reasons. Third-party tracking software is necessary for true automation.
Building the Unified Review Dashboard
Create separate raw-data tabs or base tables for the three streams. Add lookup tables for category mappings and a generated interval table. Expand the review week into 15-minute bins and assign every calendar, task, and activity record to its corresponding bins.
Recommended calendar columns are event_id, start_utc, end_utc, local_date, planned_category, attendee_count, status, and updated_at. Task columns require task_id, completed_at_utc, project, labels, estimate_minutes, actual_minutes, and transition_id. Activity columns need interval_start_utc, interval_end_utc, device, application, domain, attention_category, and idle_flag.
Define planned deep-work minutes as the sum of non-overlapping bins labeled focus. Define observed deep-work minutes as bins in which the dominant non-idle activity belongs to the approved focus-tool set. Calculate deep-work share as observed deep-work minutes divided by tracked non-idle minutes. Calculate administrative-overhead share as observed admin minutes divided by tracked non-idle minutes. Display blank when the denominator is zero instead of returning an artificial zero.
Run the weekly aggregation 30 to 60 minutes after the review window closes. This allows late device and task synchronization to settle.
Identifying Drift and Time Allocation Patterns
Compare planned and observed category minutes within the same 15-minute bins. Label a bin aligned when the categories agree. Label it redirected when another work category dominates. Flag a focus block for inspection when at least two consecutive 15-minute bins are dominated by communication, administration, or distracting activity.
Estimate context switching by counting changes between attention categories after merging excursions shorter than 30 seconds. Keep device unlocks and idle-to-active transitions separate. Returning from lunch is not a work-context switch.
Role-specific mappings prevent universal productive-app lists from skewing the data — practitioner accounts of Kevin Rexroat indicate as much. Feeding this data into a lifelogging application like Saga provides a searchable history of these behavioral shifts.
Kitty Ireland's Drift Analysis: A batching failure appears when the same administrative category occurs in three or more separated clusters during one day. Consolidate those clusters into one or two calendar blocks for the next week.
Change one scheduling variable for the next 7-day window. Adjust block length, start time, notification state, or meeting placement. The following review can then attribute improvement or deterioration to a specific intervention.
Your First Automation Step
Begin with calendar output. Event records already have bounded start and end times. They do not require device-level permissions. Create a blank spreadsheet with one header row. Connect a calendar trigger to append a new row whenever an event concludes.
Create these initial columns: event_id, title_or_tag, start_time, end_time, duration_minutes, category, attendee_count, status, and synced_at. Test with events lasting 10 to 20 minutes. The full trigger, delay, and write cycle can be observed within one session.
Inspect the duplicate rate, missing categories, canceled-event handling, and time-zone conversion after 5 to 7 days. Add completed tasks only after the calendar pipeline is stable. Add screen activity after another 5 to 7 days.
Build the calendar routing rule today to establish your baseline data stream.
