Skip to main content
Beton
posthog_usage_based posthog_usage_based_mvp_v1

Usage-based (metered) SaaS

Metered SaaS where revenue scales with consumption. Positive signals are completed jobs and compute cycles; negative signals are stalled usage.

Value metric

API calls, compute hours, messages, requests

Success event

job_completed

entity_type: account

Scale

  • 320 accounts
  • 4 users/account (mean)
  • 9 sessions/user (mean)
  • 30 days of history

Research metrics proxied

  • — Usage velocity
  • — Quota consumption
  • — Usage acceleration

Signal paths

Positive paths end in job_completed. Negative paths do not. Every generated event_id belonging to a path is recorded in ground_truth.json.

Positive signals (2)

request_to_job_completion ×95
api_request → job_completed
cohorts: high_intent, medium_intent, power_user
message_compute_job ×45
message_sent → compute_hours_used → job_completed
cohorts: high_intent, power_user

Negative signals (2)

request_compute_stall ×70
api_request → compute_hours_used
cohorts: medium_intent, low_intent, lurker
message_only_repeat ×65
message_sent → message_sent
cohorts: low_intent, lurker, noisy_bot_like

Generate this dataset

Config file: configs/posthog_usage_based_mvp.yaml

Quickstart 
# Dockerized Postgres (recommended for inspection)
docker compose up -d

uv run dryfit \
  -c configs/posthog_usage_based_mvp.yaml \
  --dsn postgresql://dryfit_writer:dryfit_writer@127.0.0.1:54329/dryfit \
  --print-summary

# Or local Postgres
./scripts/generate-local -c configs/posthog_usage_based_mvp.yaml --print-summary

Full setup instructions are in the repo's README — including local Postgres, Grafana inspection, and dataset restore.

Noise parameters

DryFit injects realistic noise on top of the generated signal paths. These probabilities are per-event. Noise never touches rows referenced by ground_truth.json — your scoring logic can trust the truth file is exact.

missing event probability
5.0%
duplicate event probability
2.0%
out of order probability
3.0%
null property probability
3.0%
anonymous actor probability
1.0%
weird property probability
1.0%

Other scenarios

Combined coverage (all models)

→ upgrade_clicked

Union of event types across all business models

Contact / record-based SaaS

→ contact_created

Contacts, leads, subscribers, accounts managed

Credits / token-based

→ credits_purchased

Credits consumed, tokens used, compute units

Event-volume SaaS

→ custom_event_tracked

Events tracked, data points ingested, log lines
Browse all scenarios →

Benchmark your detector against Usage-based (metered) SaaS

Clone the repo, run the config, check your agent's output against ground_truth.json.

View on GitHub