How Billable Works

Multi-Signal Fallback Hierarchy

We never rely on a single signal. Client assignment uses three tiers of confidence:

• Tier 1 (highest confidence): Calendar event title contains a client name. We match your calendar event names to known clients for high-confidence assignment.
• Tier 2 (medium–high): No client in the calendar title, but meeting attendee email domains match known clients, or your active work (Slack workspace, tab domains, file paths) matches a client's patterns during that time. We combine calendar + email domain + file path signals so one weak signal is backed up by others.
• Tier 3 (low confidence): Generic calendar entries (“Meeting,” “Call”) with no other signals. We surface these for your input—e.g. “Was this 2pm meeting for Client A, B, or C?”—and learn from your answer for next time.

Confidence scoring (Tiers 1–3)

Each session gets a score (0–130 points) from calendar, email domain, file paths, and learned patterns:

• Calendar + client+40 pts — Event title contains client name (Tier 1)
• Attendee domain+30 pts — Meeting domains match known clients (Tier 2)
• Workspace / paths+25 pts — Active in client Slack, domains, or file paths (Tier 2)
• Context / history+15 pts context, +20 pts from your corrections

• >70 points: High confidence — auto-assigned to client; you can override.
• 40–70 points: Medium — shown in “Possible work” for review.
• <40 points: Low — only if you choose “Show everything”.

Adaptive Threshold

The system learns from your corrections. When you mark sessions as non-billable or assign “Unassigned” to a specific client, we store minimal patterns (time of day, apps, domains, duration). Over time we stop showing noise—repetitive low-value sessions you always dismiss—and start auto-assigning work you consistently confirm. High-confidence sessions are assigned by default; you intervene only when something looks wrong. The threshold adapts to you, not the other way around.

One Question Per Day

We deliberately minimize cognitive load. Instead of interrupting you with timers or constant prompts, Billable sends one daily summary: “Yesterday you worked ~5h 20m for clients. You billed 3h. You may have left ~$420 on the table.” You choose when to review—one focused pass to confirm or correct. No live interruptions, no start/stop buttons. One question per day: did you forget to bill for this?