Fleet holdout
The platform default holdout is currently 10%. An admin can set one whole percentage, from 0% through 100%, for your account. That one percentage applies to all four recovery workflows: cancellation, trial end, access loss, and abandoned signup. Your account inherits the current platform default until an admin sets a custom percentage. Your Dashboard and Analytics show one effective percentage and whether it comes from the platform default or a merchant override. For example, a 20% override shows 20%, even when the platform default remains 10%. At 0%, no new members enter the control group. All otherwise-eligible members can receive outreach. At 100%, all otherwise-eligible new assignments enter the holdout and receive no outreach. Changes affect future assignments only, including older enrollments that never received an assignment. Existing holdouts remain held out. Previously assigned or contacted members and retries keep their assignments. Only admins can change this percentage, from the user account’s Revenue Recovery tab in the admin app. The admin recovery report shows a summary with a Manage holdout settings link. Use platform default restores inheritance. Admins can review who changed the percentage and when. Eligibility checks and the platform emergency switch still apply. The holdout creates a comparison between contacted members and members who did not receive outbound recovery. You do not start, stop, or size this measurement.Continuous lift is different from an offer experiment. An offer experiment
compares live scripts or offers. The holdout measures outreach against no
outbound recovery.
Measurement window
The platform uses a rolling 90-day window. It recomputes the pooled result every six hours as new recovery and payment outcomes mature. Cancellation-family measurement pools evidence across supported hyzl merchants. Signup recovery uses your own merchant’s holdout evidence. Neither is a custom experiment you start yourself. Comparisons account for different percentages and policy periods. Periods need both outreach and holdout members to contribute causal evidence. Assignments at 0% or 100% cannot create a new comparison. Analytics keeps observed recovery totals separate from comparable evidence. It uses qualifying lift to estimate net-caused revenue.Measurement states
The continuous lift card uses three evidence states:- Not measured yet means the current data cannot support a lift estimate.
- Early signal shows a directional comparison without a proven causal factor.
- Proven means the relevant measurement supports the displayed causal factor.
Read the card
Signup measurement compares conversion rates for contacted and holdout groups. The pooled cancellation-family card compares revenue for the same number of members in each group. A proven result also shows the causal factor used by the commission calculation. Adjusted comparisons account for allocation differences. Observed counts describe actual outcomes, while the effective sample describes the evidence available for a comparison. A partial or insufficient recomputation can return the card to an earlier evidence state. Treat the current label as part of the result. When the platform result is proven, the Dashboard applies its factor across all recovery workflows. Both split boxes say Estimated. The estimate uses your chosen date range and does not establish account-specific causal proof. Before proof, the Dashboard can show numeric estimates in gray Preview from a fresh, complete, usable platform comparison. Preview may use an earlier measurement method with unverified cohort quality. This display does not authorize commission or change the evidence states above. Missing, malformed, outdated, or partial comparison data cannot supply Preview amounts. Known holdout contact or missing payment evidence also prevents the estimate. A zero organic estimate does not prove that nobody returns without outreach. Caused and organic estimates add up to observed revenue for all time, chosen dates, and chart selections. Member rows keep actual payment amounts. The full recovery list, chart, and recovered count describe the same observed cohort, without assigning individual returns to either causal group. Use continuous lift to understand the estimate’s evidence. Use Dashboard totals and trigger funnels for observed activity.When holdout is off
At 0%, the policy shows Holdout disabled. No new control-group members are assigned. This applies to every recovery workflow. The same label also shows at any percentage when the platform emergency switch is off. Valid lift from earlier cohorts remains visible with its measurement window and last measurement date. Historical cohort percentages describe that period, separately from today’s policy. Without qualifying evidence, the view showsLift unavailable—holdout disabled. Expired evidence is unavailable until it is recomputed. Missing evidence is not a measured zero conversion rate or proof that hyzl caused every recovery.
More outreach alone cannot restore lift while no new control members enter the measurement. Outreach, conversions, and recovered revenue remain visible.
Signup commission can still use qualifying historical evidence under the existing rules. Once that evidence is unavailable, there is no signup lift commission. A 0% setting does not erase history or change finalized invoices.
Cancellation-family billing continues to use qualifying pooled platform evidence. hyzl does not replace signup evidence with external controls or pooled lift.