Work Coaching

For baseline comparisons, query the same aggregations over the baseline period and compute averages. Only show trend comparisons when baseline data exists.

Data Access

Data stream: logs-toggltrackreceiver.otel-default

Query using the es CLI. Always pipe through tail -1 to strip the config file line:

es search "logs-toggltrackreceiver.otel-default" -q '<json-query>' 2>&1 | tail -1 | python3 -m json.tool

Important Notes

• @timestamp is the entry end time, not the start. Use attributes.start for the start time.
• attributes.description is a text field — no .keyword sub-field. Cannot use terms aggregation on it. Fetch raw entries and aggregate in Python.
• Duration is in seconds. Divide by 3600 for hours.
• Filter Elastic work with: {"term": {"attributes.project.name": "Elastic"}}

Data Model

Each document is a single completed Toggl Track time entry:

{
  "@timestamp": "2026-02-06T17:59:56.000Z",
  "attributes": {
    "id": "4280455213",
    "description": "sync & plan",
    "start": "2026-02-06T17:07:12.000Z",
    "end": "2026-02-06T17:59:56.000Z",
    "duration": 3164,
    "project.id": "178435728",
    "project.name": "Elastic",
    "tags": ["type:sync/planning", "stream:ecf"],
    "billable": "false",
    "workspace.id": "1815018",
    "workspace.name": "Maurizio Branca's workspace"
  },
  "work": {
    "type": "sync",
    "activity": "planning",
    "stream": "ecf",
    "project": "ecf"
  }
}

Tag System

Entries are classified along two dimensions using structured tags:

Work Type (type: tag → work.type + work.activity):

Any type value not in this table is unexpected and should be flagged as a warning in Process Mechanics.

Work Stream (stream: tag → work.stream + work.project):

Any stream value not in this table is unexpected and should be flagged as a warning in Process Mechanics.

Goals File

Read the goals file from the Obsidian vault root:

~/obsidian/vaults/Elastic/Goals.md

If the file does not exist, print a note saying goal alignment analysis is skipped and continue with the other sections. Never fail because of a missing goals file.

The goals file contains sections for different time horizons. Each goal has a name, optional description, and associated stream:/type: tags that map time entries to that goal. Parse the file to build a goal → tag mapping, then match time entries against it.

Analysis Procedure

Run these steps in order. Fetch data efficiently — combine queries where possible, use aggregations for summaries, fetch raw entries only when needed (e.g., for description grouping or per-day analysis).

Step 1: Fetch Data

Run these queries for the target period:

1. Aggregations by work type — terms on work.type with sub-aggregation on work.activity, plus sum on attributes.duration
2. Aggregations by work stream — terms on work.stream with sub-aggregation on work.project, plus sum on attributes.duration
3. Daily breakdown — date_histogram on @timestamp with calendar_interval: day, plus sum on attributes.duration
4. Raw entries — all entries with _source: ["@timestamp", "attributes.start", "attributes.description", "attributes.duration", "attributes.tags", "work"], sorted by attributes.start ascending
5. Baseline period — same aggregations (1, 2, 3) for the baseline period

Step 2: Analyze and Report

Produce the following sections in order.

Output Sections

1. Summary

A brief overview: total hours, number of entries, number of working days, average hours/day.

2. Process Mechanics

Check the health of the tracking process itself:

• Tag coverage: count entries missing a stream: tag, missing a type: tag, or missing both. List the descriptions of untagged entries so the user can fix them.
• Unknown tags: any type: or stream: values not in the documented taxonomy above. List them with entry count.
• Unexpected streams: any stream: value not in {ecf, motlp, synthetics, cloud, hr} is a warning — list entries and suggest re-tagging or consolidation (e.g., legacy → cloud).
• Hours check: flag days significantly above 10h or below 4h (for working days). Flag if total weekly hours are notably above 45h or below 30h.
• Entry granularity: flag entries longer than 3h (consider splitting) or shorter than 5min (noise?).

3. Time Allocation

Show the breakdown in a table:

| Work Type  | Hours | %   | Baseline % | Delta |
|------------|-------|-----|------------|-------|
| goal       | 10.0h | 25% | 30%        | -5%   |
| meeting    |  9.9h | 25% | 20%        | +5%   |
| ...        |       |     |            |       |

Do the same for work streams. Highlight any type or stream where the delta from baseline exceeds ±10 percentage points.

4. Focus Patterns

Analyze how fragmented or focused the work was:

• Streams per day: for each working day, count distinct work.stream values. Flag days with 4+ streams as high context-switching.
• Deep work blocks: count entries tagged type:goal that are >= 1h. These are valuable focus blocks.
• Fragmentation score: for each day, compute (number of distinct streams) / (number of entries). Lower is more focused. Show per-day and average.
• Longest uninterrupted goal block per day: the longest single type:goal entry each day.

5. Trends vs Baseline

Compare the current period against the baseline average. Only show this section if baseline data exists.

• Work type % shifts (highlight changes > 5 percentage points)
• Work stream % shifts
• Total hours trend (increasing/decreasing/stable)
• Meeting load trend

Frame changes neutrally — a shift isn't automatically bad. Provide context: "meeting load increased from 20% to 25%, which may reflect the iteration planning cycle."

6. Goal Alignment

Read Goals.md and map entries to goals:

• For each goal, show hours spent on matching entries
• Show what % of total work time is goal-aligned
• Flag goals that received zero hours this period
• For quarter scope: show a month-by-month breakdown per goal

If Goals.md is missing, skip this section with a note.

7. Coaching Suggestions

Provide 3–5 concrete, actionable suggestions based on everything above. Prioritize by impact. Frame as:

• "Next [week/month], consider..."
• "Watch out for..."
• "Keep doing..."

Ground suggestions in the data: "Your support load was 20% this week (vs 12% baseline). If this continues, consider discussing SDH rotation with your manager."

Include at least one positive observation — something the user is doing well.

Formatting Guidelines

• Use markdown tables for structured data
• Use bold for key numbers and alerts
• Keep each section concise — the full output should be scannable in under 2 minutes
• Use --- horizontal rules between sections
• For quarterly scope, use ### sub-headers for each month within sections