Skip to content

:hedgehog: PostHog is the leading platform for building self-driving products. Our developer tools – AI observability, analytics, session replay, flags, experiments, error tracking, logs, and more – capture all the context agents need to diagnose problems, uncover opportunities, and ship fixes. Steer it all from Slack, web, desktop, or the MCP.

fix(signals): let projects take turns in anomaly scoring picks (#122402)

posthog[bot]f34836541ae0
diff --git a/products/signals/backend/anomaly_scores/queue.py b/products/signals/backend/anomaly_scores/queue.py
index fdbdd41a9ed3..dad7b1dc851c 100644
--- a/products/signals/backend/anomaly_scores/queue.py
+++ b/products/signals/backend/anomaly_scores/queue.py
@@ -5,6 +5,7 @@ from the `$anomaly_score` events. `selection.pick_order` then orders the due ins
"""
import datetime
+import itertools
from collections.abc import Collection
from typing import Any, cast
@@ -223,10 +224,10 @@ def select_insights(
for team_id in sorted({insight.team_id for insight in due_by_id.values()}):
due_ids = [insight.insight_id for insight in due_by_id.values() if insight.team_id == team_id]
usage.update(insight_usage(team_id=team_id, insight_ids=due_ids, now=now))
- candidates = []
+ candidates_by_team: dict[int, list[Candidate]] = {}
for insight in due_by_id.values():
newest = newest_buckets[insight.insight_id]
- candidates.append(
+ candidates_by_team.setdefault(insight.team_id, []).append(
Candidate(
insight_id=insight.insight_id,
interval=insight.interval,
@@ -237,9 +238,19 @@ def select_insights(
usage=usage[insight.insight_id],
)
)
- order = pick_order(
- candidates, config=selection_config, tick_start=now, max_staleness=config.limits.max_backfill_buckets
- )
+ team_orders = [
+ pick_order(
+ candidates_by_team[team_id],
+ config=selection_config,
+ tick_start=now,
+ max_staleness=config.limits.max_backfill_buckets,
+ )
+ for team_id in sorted(candidates_by_team)
+ ]
+ # Projects take turns, so one large project cannot take every slot of the tick budget.
+ order = [
+ insight_id for turn in itertools.zip_longest(*team_orders) for insight_id in turn if insight_id is not None
+ ]
picks = [
InsightPick(
team_id=due_by_id[insight_id].team_id, insight_id=insight_id, scored_through=state.get(insight_id, {})
diff --git a/products/signals/backend/test/test_anomaly_scoring_tick.py b/products/signals/backend/test/test_anomaly_scoring_tick.py
index a6aaa3317e0e..42fbb44e22e6 100644
--- a/products/signals/backend/test/test_anomaly_scoring_tick.py
+++ b/products/signals/backend/test/test_anomaly_scoring_tick.py
@@ -13,7 +13,7 @@ from parameterized import parameterized
from posthog.schema import IntervalType
-from posthog.models import Organization, Team
+from posthog.models import Organization, Team, User
from products.dashboards.backend.models.dashboard import Dashboard
from products.dashboards.backend.models.dashboard_tile import DashboardTile
@@ -339,3 +339,35 @@ class TestScoredBuckets(ClickhouseTestMixin, BaseTest):
assert [pick.insight_id for pick in selection.picks] == [insights["never_scored"].id, insights["stale"].id]
assert selection.picks[1].scored_through == {"stale": today - timedelta(days=3)}
+
+ @time_machine.travel(NOW, tick=False)
+ def test_select_insights_lets_each_project_pick_before_any_project_picks_twice(self) -> None:
+ small_team = Team.objects.create(organization=self.organization, name="Small")
+ viewers = [
+ User.objects.create_and_join(self.organization, f"viewer{index}@example.com", "password")
+ for index in range(5)
+ ]
+ large_insights = []
+ for index in range(3):
+ insight = Insight.objects.create(team=self.team, name=f"large_{index}", query=_trends(interval="day"))
+ for viewer in viewers:
+ record_insight_view(insight_id=insight.id, team_id=self.team.id, user_id=viewer.id)
+ large_insights.append(insight)
+ small_insight = Insight.objects.create(team=small_team, name="small", query=_trends(interval="day"))
+ small_dashboard = Dashboard.objects.create(team=small_team, name="Main")
+ DashboardTile.objects.create(dashboard=small_dashboard, insight=small_insight)
+ for insight in [*large_insights, small_insight]:
+ cache.delete(_cooldown_key(insight.id))
+ config = AnomalyScoringConfig(
+ enabled=True,
+ organization_ids=(str(self.organization.id),),
+ allow_team_ids=(self.team.id, small_team.id),
+ )
+
+ with patch("products.signals.backend.anomaly_scores.queue.list_limited_team_attributes", return_value=[]):
+ selection = select_insights(config, SelectionConfig(weighting=Weighting.WEIGHTED), NOW, limit=2)
+
+ assert [pick.team_id for pick in selection.picks] == [self.team.id, small_team.id]
+ assert selection.picks[0].insight_id in {insight.id for insight in large_insights}
+ assert selection.picks[1].insight_id == small_insight.id
+ assert selection.candidates == 4