Defining activation before configuring PostHog funnels.
Define activation as a completed customer outcome with a clear population, entity, and time window. Instrument the evidence of that outcome, configure PostHog to match the contract, and validate labeled journeys before interpreting conversion rates.
Inside this article 5 sections +
Define activation as a completed customer outcome before configuring a PostHog funnel. Specify the entity that activates, the evidence of value, the entry population, and the time allowed. Then instrument the business events that make that definition testable.
For a B2B product, account activation often involves several people. An administrator creates a workspace, an analyst imports data, and a colleague uses the result. A person-level sequence can report failure even though the customer successfully adopted the product.
Write a value statement that excludes setup activity
Consider an illustrative reporting application. “Created an account” means the customer reached the application; “uploaded a file” means data entered it. Neither proves that someone obtained a useful result. A candidate activation definition is that a new customer workspace publishes a valid report and a second member views it within ten days of workspace creation.
This is a hypothesis about value, not a universal benchmark. Discuss it with product and customer-facing colleagues. A solo user plan may need another definition, and an automated export product may deliver value without a second human. Name the population for which the definition is intended so one convenient metric does not become an inaccurate company-wide target.
Instrument the business fact, not the button
Emit report_published only after the authoritative operation succeeds. Include the event ID, account ID, actor ID, report ID, occurrence time, environment, and schema version. Keep an explicit marker for demo data or internal workspaces so those actions can be excluded consistently.
A click event can help diagnose friction, but it should not stand in for a successful publication. Client-side retries, validation failures, and navigation away from the page create a gap between intention and result. Keep exploratory interface events alongside server-confirmed business events, with names that reveal which is which.
The second-member condition also needs evidence. A generic report-view event does not prove that its actor differs from the publisher. Derive an explicit qualifying event or implement a query that checks the relationship. Do not assume a sequence of event names expresses a relationship that the event data never establishes.
Configure the funnel to match the contract
Choose the account group as the aggregation unit when different members may complete different steps. Ensure every relevant event carries the correct group association; the B2B account identity guide provides fixtures for workspace switching and shared membership.
Use sequential order when unrelated actions may occur between milestones. Strict order can remove valid customers merely because they opened settings between publishing and viewing. Any-order analysis is useful only when the business definition truly permits the milestones in either order.
Set the conversion window deliberately. PostHog's selected date range also affects which step events are included, so an entry cohort near the end of the range needs enough observation time for later steps. Extend the analysis range and constrain the entry period where appropriate. Label recent cohorts as incomplete instead of interpreting their unfinished journeys as a sudden drop in activation.
Use exclusions without changing the question silently
Exclude development environments, staff demonstrations, imported test accounts, and other populations outside the contract. Keep the reasons in the metric definition. Removing customers who encountered an error, however, can hide the very failure the funnel should reveal. Exclusion is a population decision, not a way to make conversion look healthier.
Distinguish a permanent account attribute from a property captured at entry. If a customer changes plan during the observation window, grouping by their current plan may rewrite the apparent history. Decide whether the report describes entry conditions or current customer status, then select properties and query logic accordingly.
When a definition changes, version it. A new ten-day activation metric should not be overlaid on the old seven-day series without explanation. Recompute comparable history where feasible and retain the old definition long enough to explain the discontinuity.
Validate with labeled journeys before setting a target
Create a small fixture containing a successful two-person account, a solo account, a publication failure, a late viewer, a demo workspace, and a customer with unrelated events between milestones. Write the expected inclusion and outcome for each. Compare the funnel against those labels and trace discrepancies to events or query semantics.
This executable Python fixture is an independent oracle for the illustrative ten-day definition, not a PostHog query. Inputs are already filtered to a real customer account and committed production events; elapsed days are measured from that account’s creation.
def activated(publications, views):
return any(
p["report"] == v["report"]
and p["actor"] != v["actor"]
and 0 <= p["day"] <= v["day"] < 10
for p in publications for v in views
)
published = [{"report": "r8", "actor": "author", "day": 2}]
assert activated(published, [{"report": "r8", "actor": "reader", "day": 4}])
assert not activated(published, [{"report": "r8", "actor": "author", "day": 4}])
assert not activated(published, [{"report": "r9", "actor": "reader", "day": 4}])
assert not activated(published, [{"report": "r8", "actor": "reader", "day": 10}])
The report match matters: viewing an unrelated report should not complete this definition. Compare these expected outcomes with the configured funnel and investigate any difference in group identity, event order, or window boundaries. For real timestamps, specify whether the product means elapsed hours or local calendar days before translating this oracle into a query.
After validation, inspect real adoption paths and retention alongside activation. A high activation rate is useful only if the selected behavior represents customer value. The broader product event integration guide connects analytics to lifecycle and warehouse destinations. Software implementation can make the activation contract executable across the application, event pipeline, and reporting layer.