Skip to main content
Beton
posthog_web posthog_mvp_v1

PostHog Web (baseline)

The minimum-viable PostHog scenario. A generic event stream with a two-step conversion funnel and corresponding negative path. Good starting point for signal-detection benchmarking before moving to business-model-specific variants.

Value metric

Generic SaaS activation (baseline)

Success event

purchase

entity_type: account

Scale

  • 500 accounts
  • 4 users/account (mean)
  • 10 sessions/user (mean)
  • 30 days of history

Research metrics proxied

  • — Baseline conversion rate
  • — Funnel drop-off

Signal paths

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

Positive signals (2)

billing_then_purchase ×80
page_billing → purchase
cohorts: high_intent, medium_intent, power_user
invite_then_api_key_then_purchase ×35
invite_teammate → api_key_created → purchase
cohorts: high_intent, power_user

Negative signals (2)

billing_dropout ×70
page_billing → page_pricing
cohorts: low_intent, lurker, medium_intent
pricing_only_stall ×110
page_pricing → page_pricing
cohorts: low_intent, lurker, noisy_bot_like

Generate this dataset

Config file: configs/posthog_mvp.yaml

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

uv run dryfit \
  -c configs/posthog_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_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
6.0%
duplicate event probability
2.0%
out of order probability
3.0%
null property probability
5.0%
anonymous actor probability
2.0%
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 PostHog Web (baseline)

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

View on GitHub