Paywall Conversion Rate by Channel: Choose a Fair Cohort
Paywall conversion rate by channel defines fair cohorts and compares variants within each source, with consistent attribution, exposure and conversion windows.
A fair paywall conversion comparison by channel
- Define one eligible paywall exposure and one unique-user denominator.
- Separate trial starts, first paid conversions, and renewals into clear outcomes.
- Keep acquisition attribution rules and post-exposure conversion windows consistent.
- Compare variants within each channel before reading an overall rate.
- Use Airbridge’s Funnel Report for subscription rate and cost by channel, campaign, and creative; record paywall exposure and variant in your own analytics.
A paywall conversion rate by channel is the number of eligible users from a channel who complete a specified subscription outcome within a set period, divided by the unique eligible users from that channel who saw the paywall. The useful comparison connects an acquisition source to an actual paywall exposure, a recorded variant, and a defined subscription event.
For a small team, that definition prevents a busy dashboard from turning into a bad budget decision. A channel can send plenty of people to the app while producing fewer paywall viewers, trial starts, or first payments.
Define the paywall conversion rate
Start with one question: are you measuring a paywall view that starts a trial, or a paywall view that leads to a paid subscription? Each is a useful rate, but they answer different questions. Keep the outcome event in the metric name, such as “paywall-view-to-trial-start rate” or “paywall-view-to-first-paid rate.”
| Metric field | Definition to use | Counting rule |
|---|---|---|
| Eligible population | People who reached the same qualifying paywall entry point during the test | Include people who meet the rule for every channel and variant |
| Paywall exposure | The event that means the paywall actually rendered and was available to the person | Record the event once for the selected test exposure, with the variant label |
| Channel | The acquisition channel assigned under the chosen attribution rule | Apply the same source grouping and attribution settings to all cohorts |
| Conversion outcome | One named event, such as trial start or first paid conversion | Count a user once if the outcome occurs in the selected window |
| Denominator | Unique eligible users with a qualifying exposure, or unique users assigned to the test | Choose one denominator and report it beside every rate |
| Observation window | The time allowed after cohort entry for the outcome to happen | Give each user the same follow-up duration |
For an exposed-user rate, divide the unique users who complete the chosen outcome by the unique users with a qualifying paywall exposure. For example, if 48 of 600 exposed users from one channel start a trial within seven days, that channel’s seven-day exposed-user-to-trial-start rate is 8%: 48 divided by 600.
Use a single user as the counting unit when the decision concerns the chance that a person converts. A person who opens the paywall four times still contributes one eligible user to the denominator. A person who triggers multiple renewal events also counts once for a user-level first-paid conversion metric. A transaction-level metric can count transactions, but it needs a transaction denominator and a name that makes that unit clear.
Name the event precisely. A trial start measures willingness to begin a trial; a trial conversion measures a later transition to paid; an initial purchase can measure a direct paid start; and a renewal measures continued billing. The Airbridge RevenueCat event mapping maps TRIAL_STARTED to Start Trial, while INITIAL_PURCHASE, TRIAL_CONVERTED, and RENEWAL map to Subscribe. For a first-paid outcome, select the relevant first payment or trial-conversion event in your reporting definition, and keep renewals as a separate retention or revenue outcome.
Give every exposure a stable identity and variant label. The exposure event should include a user or device key, timestamp, paywall or placement identifier, and the variant shown. When a user sees multiple paywalls, define the specific paywall entry point and test identifier so that a product screen, onboarding flow, and return visit do not silently enter the same denominator.
Google Analytics’ traffic-source definitions define source as the referring platform, site, app, or online location, and medium as the traffic category; its campaign documentation maps source, medium, and campaign parameters to those report dimensions. Use the same channel grouping over the whole comparison, and keep campaign and creative as more detailed dimensions beneath that channel. For example, a paid social cohort may include several campaigns, while a search cohort may include several keyword campaigns.
Build channel cohorts and wait for maturity
A cohort is a set of users grouped by a shared entry event or date. For this comparison, use the first qualifying paywall exposure during the experiment as cohort entry, attach that user’s recorded acquisition channel and assigned variant, then observe the same outcome window for every cohort.
- Choose the channel assignment event. For acquisition reporting, decide whether the cohort uses the campaign attributed to the app install or another explicitly defined acquisition event. Store the attributed channel on the user or exposure record so it travels with the paywall outcome.
- Set the attribution rule. Keep the model, lookback period, channel grouping, and campaign parameters constant for every comparison period. Airbridge’s attribution model guide describes ad views and clicks as touchpoints. Its model checks eligible touchpoints within the configured lookback window, selects a winning touchpoint, and records campaign information such as channel and campaign name against the target event.
- Set a separate conversion window. Choose how long after paywall exposure a user can complete the selected outcome. For example, a seven-day trial-start rate can use a seven-day window after the first qualifying paywall exposure. A first-paid conversion following a seven-day free trial may need a longer window that includes the trial period and a defined payment-processing allowance.
- Wait until every included cohort has had the full window. A seven-day conversion window means a user who saw the paywall yesterday has six days of potential follow-up remaining. Compare weekly cohorts after the last user in each cohort has completed the full seven-day window.
A user can interact with more than one ad before installing. Airbridge’s winning-touchpoint rule assigns the target event to one eligible touchpoint, so the channel row represents the channel credited under the selected model. Keep one credited channel per user in a single-touchpoint report, and use a clearly defined multi-touch analysis when the question concerns the full path of campaign interactions.
Attribution lookback and conversion observation measure different time intervals. Airbridge’s attribution window guidance describes an attribution window as the period after a target event in which later events can be associated with that target. Its attribution model separately uses a lookback window before a target event to determine which ad touchpoint can receive credit. Your paywall experiment’s outcome window begins at the chosen exposure or assignment event, so specify it separately in the analysis.
A useful cohort row carries four dates or durations: acquisition touchpoint time, attribution target time, first eligible paywall exposure time, and outcome deadline. These fields make it easier to tell whether a paid event falls within the attribution rules and the experiment’s follow-up period. Google Analytics also documents a key-event lookback setting that controls how far back a touchpoint can receive attribution credit; its lookback-window guidance says changes apply going forward across the property. Record the setting alongside the report period when an attribution system’s configuration changes.
Use a real calendar example to set the cutoff. If the final member of a weekly cohort saw a paywall on Sunday at 8 p.m., and the selected outcome window is seven days, finish the comparison after the following Sunday at 8 p.m. If you instead read all cohorts through the same calendar end date, the newest users receive less time to convert and their rates appear lower simply because their outcome window remains open.
Airbridge Core links trial conversions, first payments, renewals, and refunds from RevenueCat, Adapty, or Superwall to the campaign that brought the user in. This links subscription outcomes to acquisition reporting across those billing or paywall stacks. For a channel cohort, pair that campaign-linked outcome with the paywall exposure and variant record that defines the analysis.
Compare variants within each channel
Keep the variant comparison inside each channel before combining results. Use the same paywall entry point, exposure rule, outcome event, user-counting rule, and observation window for variant A and variant B. A channel that supplies more eligible people affects how many conversions the app receives; its size alone does not show which paywall variant converts a larger share.
| Channel | Variant A: exposed users | Variant A: outcome users | Variant A rate | Variant B: exposed users | Variant B: outcome users | Variant B rate |
|---|---|---|---|---|---|---|
| Search | 900 | 162 | 18% | 100 | 19 | 19% |
| Paid social | 100 | 8 | 8% | 900 | 81 | 9% |
| Combined | 1,000 | 170 | 17% | 1,000 | 100 | 10% |
In this illustrative table, Variant A has a lower rate in each channel, yet its combined rate is higher because more of its users came from the high-converting search cohort. Judea Pearl’s review of Simpson’s paradox describes how an association can reverse when groups with different underlying rates are combined. Read channel-specific rates alongside the pooled rate to separate paywall performance from channel composition.
When a team needs one headline variant rate, apply the same channel weights to both variants. For example, with a 50% search and 50% paid social mix, variant A’s illustrative standardized rate is 13%: half of 18% plus half of 8%. Variant B’s standardized rate is 14%: half of 19% plus half of 9%. This common-mix calculation keeps the channel composition equal for the two variants.
Choose the weights from a reference period or user mix that matches the decision, then apply those exact weights to each variant’s within-channel rates. The live pooled rate answers what the current traffic mix experienced; a fixed-weight rate answers how variants compare at a shared channel mix. Report both when budget allocation and paywall design decisions need different views of the same test.
A controlled variant test also needs stable assignment. Statsig’s experiment overview describes random assignment at a chosen unit, such as a user, device, or session, with the same unit continuing to receive the same variant. Choose the unit that matches your product identity and paywall flow. For a subscription app, a stable user ID often gives a cleaner person-level comparison than a session ID, because the same person can return in a later session.
Keep the assigned variant attached to each eligible user, then report both assignment and exposure counts. The assignment-based rate answers what happened to the people eligible for each test arm. The exposed-user rate answers what happened among people whose paywall rendered. If the two denominators differ, the gap helps reveal delivery or rendering issues before you interpret conversion performance.
Use a table with one row per channel and variant, and include the underlying counts beside the rate. A practical report includes channel, campaign, creative, paywall placement, variant, eligible assignments, rendered exposures, trial starters, first-paid users, conversion window, and cohort end date. The figures help a small team distinguish “this channel brought fewer people” from “this variant converted a smaller share of the people who saw it.”
Airbridge Core’s Funnel Report shows subscription rate and cost by channel, campaign, and creative, and shows which campaign and creative combinations turned into subscriptions and revenue. Paywall exposure and variant are not sent to Airbridge by RevenueCat, Adapty, or Superwall, so record them in your own analytics and join them to the campaign view when deciding which combinations merit a closer cohort comparison. Define the paywall-render exposure event and record the assigned variant with each eligible user’s subscription outcome.
QA the rate before changing spend or a paywall
Run these checks before moving budget or choosing a winning variant. They protect the cohort definition, the join between systems, and the interpretation of the rate.
- Confirm one paywall entry rule. Verify that each eligible user reached the same placement and that the exposure event means the paywall rendered. Keep exposure and assignment counts available when render delivery differs.
- Check channel values at the user level. Confirm campaign source, medium, and channel fields populate for the people in the report. Review unassigned and unknown-source users as a separate row so their outcomes do not get silently redistributed to a paid channel.
- Freeze attribution settings for the comparison. Record the attribution model, touchpoint lookback, post-target attribution window, channel grouping, and report dates.
- Validate the subscription event mapping. Check that trial start, initial purchase, trial conversion, renewal, and refund reflect the outcomes in your metric name. Airbridge’s RevenueCat mapping places several billing events under Subscribe, so inspect the underlying event type when the decision depends on first payment versus renewal.
- Pass the Airbridge Device ID, align user IDs, and prevent duplicate collection. For the Airbridge–Superwall connection, the integration guide requires passing the Airbridge Device ID for device matching, using the same user ID in both SDKs when matching by user ID, and configuring the Airbridge SDK to avoid collecting events Superwall already sends. The guide notes that setup requires code changes, so include an engineer in this check.
- Check variant balance and event joins. Statsig’s experiment monitoring guidance describes exposure-balance and sample-ratio checks that can flag unexpected differences in group sizes. Inspect assignments, exposures, and outcome events using the same identity key; an event without the experiment user ID cannot join reliably to a variant.
- Compare only mature cohorts. Confirm that every user in the reporting window has had the full conversion period. Keep recent, still-open cohorts out of the mature rate or label them as preliminary.
- Read each conversion rate with its counts and uncertainty estimate. Statsig’s experiment overview reports metric lifts with confidence intervals; compare the interval and conversion counts before acting on a rate difference.
A working report review can follow a short order: inspect cohort and event definitions, verify data joins, confirm cohort maturity, compare variant rates within each channel, then review the pooled rate and conversion counts. When a channel-level result looks strong, check campaign and creative rows before changing the whole channel budget. A win concentrated in one creative-paywall pairing supports a narrower action than a channel-wide shift.
The decision rule is simple: act on a channel-by-variant rate only when its denominator, attributed channel, outcome event, and observation window mean the same thing across the rows you compare. Keep trial starts, first paid conversions, and renewals as separate outcomes, then use the mature within-channel results to choose the next test or budget adjustment.
Get Started Free
See how these criteria hold up on the real thing.
