: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.
Python ab-testingai-analyticsanalyticscdpdata-warehouseexperimentsfeature-flagsjavascriptproduct-analyticspythonreactsession-replaysurveystypescriptweb-analytics
import { dayjs } from 'lib/dayjs'
import { setLatestVersionsOnQuery } from '~/queries/schema/query-versions'
import { EventsNode, InsightVizNode, NodeKind, TrendsQuery } from '~/queries/schema/schema-general'
import { BaseMathType, PropertyFilterType, PropertyOperator } from '~/types'
import { SchemaPropertyGroupProperty } from '../schema/schemaManagementLogic'
const MAX_PROPERTIES = 25
const LOOKBACK_DAYS = 90
export interface PropertyGroupTrendsQueryResult {
query: InsightVizNode
isTruncated: boolean
totalProperties: number
displayedProperties: number
/** Human-readable description of the charted window, e.g. "last 90 days" or "since Jun 15, 2026". */
dateRangeLabel: string
}
export function buildPropertyGroupTrendsQuery(
eventName: string,
properties: SchemaPropertyGroupProperty[],
eventFirstSeen?: string | null
): PropertyGroupTrendsQueryResult {
// Limit to 25 properties since we use letters B-Z for series labels (A is reserved for the base series)
const isTruncated = properties.length > MAX_PROPERTIES
const limitedProperties = isTruncated ? properties.slice(0, MAX_PROPERTIES) : properties
const baseSeries: EventsNode = {
kind: NodeKind.EventsNode,
event: eventName,
name: eventName,
math: BaseMathType.TotalCount,
}
const series: EventsNode[] = [
baseSeries,
...limitedProperties.map(
(property): EventsNode => ({
kind: NodeKind.EventsNode,
event: eventName,
name: eventName,
properties: [
{
key: property.name,
type: PropertyFilterType.Event,
value: 'is_set',
operator: PropertyOperator.IsSet,
},
],
math: BaseMathType.TotalCount,
})
),
]
const formulaNodes = limitedProperties.map((property, index) => {
// Generate series labels B, C, D, ... Z (ASCII 66 = 'B', 67 = 'C', etc.)
const seriesLetter = String.fromCharCode(66 + index)
return {
formula: `${seriesLetter}/A * 100`,
custom_name: property.name,
}
})
// Coverage is `propertyCount / eventCount`, so weeks where the event never fired evaluate to 0/0,
// which the formula engine renders as 0% rather than a gap. For an event younger than the lookback
// window that paints a misleading flat 0% line across every week before the event existed, hiding
// the real coverage in the final buckets. Never start the window before the event first appeared.
const defaultFrom = dayjs().subtract(LOOKBACK_DAYS, 'day')
const firstSeen = eventFirstSeen ? dayjs(eventFirstSeen) : null
const clampedFrom = firstSeen && firstSeen.isValid() && firstSeen.isAfter(defaultFrom) ? firstSeen : null
const dateFrom = clampedFrom ? clampedFrom.format('YYYY-MM-DD') : `-${LOOKBACK_DAYS}d`
const dateRangeLabel = clampedFrom ? `since ${clampedFrom.format('MMM D, YYYY')}` : `last ${LOOKBACK_DAYS} days`
const trendsQuery: TrendsQuery = {
kind: NodeKind.TrendsQuery,
series,
interval: 'week',
dateRange: {
date_to: null,
date_from: dateFrom,
explicitDate: false,
},
trendsFilter: {
formulaNodes,
showLegend: true,
aggregationAxisFormat: 'percentage',
},
}
return {
query: setLatestVersionsOnQuery<InsightVizNode>({
kind: NodeKind.InsightVizNode,
source: trendsQuery,
showHeader: false,
showTable: false,
showFilters: false,
embedded: true,
}),
isTruncated,
totalProperties: properties.length,
displayedProperties: limitedProperties.length,
dateRangeLabel,
}
}