Why it matters
Every performance team changes campaigns constantly: new creative, audience edits, bid cap tweaks, and fresh conversion events. That is normal operation, but it is a poor scientific control. When you launch predicted lifetime value (pLTV) on Meta CAPI or shift Google campaigns to value goals, you need a frozen reference point that answers one question: what would have happened if we had not made this change?
BAU conversion is that reference. It might be purchase-only pixel and CAPI events, flat first-order value, or a trial-start proxy you already run today. The treatment arm receives the new pLTV score, timing, or value scale. The readout compares outcomes after cohort maturity, not after the first weekend of platform ROAS.
Teams that skip BAU definition often debate whether a lift was real or whether someone changed audiences mid-test. Finance and procurement increasingly ask for BAU-backed proof before renewing signal vendors or scaling spend.
Business as usual (BAU) conversion
BAU is the control arm in a typical Churney pLTV pilot:
- Document current conversion events, value parameters, and delivery paths (pixel, CAPI, MMP postbacks) as BAU.
- First-party data in your data warehouse continues to feed modeling; BAU does not turn off analytics, only the new platform-facing value signal.
- Churney activates pLTV values directly to ad networks for the treatment segment while BAU segments keep prior event design.
- Holdout or split-test routing ensures no accidental bleed of pLTV events into control.
- Compare treatment vs BAU on incremental conversions, revenue, and incremental ROAS once maturity allows.
BAU is not "do nothing." It is "keep doing what we did before the test." The data warehouse remains an input to both modeling and readout; Churney sends treatment signals to ad platforms while BAU preserves the counterfactual.
Category variants
| Model | What BAU conversion often looks like |
|---|---|
| Ecommerce / DTC | Standard Purchase events with actual first-order value; no pLTV overlay; same audiences and budgets as pre-pilot where possible. |
| Subscription app | Install or trial-start events without modeled subscription LTV; existing MMP-to-network postback path unchanged in control. |
| SaaS / PLG | Lead or signup conversion without modeled expansion value; prior offline or server-side upload schema frozen for control cells. |
Common mistakes
- Redefining BAU mid-pilot. Changing control event schema, value scale, or budgets invalidates lift readout.
- BAU that is not actually live today. Inventing a weak control that no real campaign uses produces misleading lift estimates.
- No documentation of BAU parameters. Event names, value fields, match identifiers, and campaign IDs must be frozen in writing.
- Treatment changes that also touch BAU campaigns. Accidental global CAPI updates push pLTV into control traffic.
- Comparing BAU to treatment on different maturity windows. Both arms need the same readout dates and revenue definitions.
Advertiser lens
| Role | What they ask | What good looks like |
|---|---|---|
| Head of Performance / UA | What exactly stays the same in control? | Written BAU spec: events, values, campaigns, and routing rules. |
| VP Growth / CMO | Is the test fair? | Matched spend levels or pre-agreed scale rules; single primary metric. |
| Marketing Analytics / Data Science | Is BAU stable enough to measure lift? | Stability checks on control metrics before treatment launch; leakage monitoring. |
| Data Engineering | How do we route users to BAU vs treatment? | Feature flags, holdout IDs, and deployment checklist with rollback plan. |
| Finance / Procurement | What baseline triggers success? | BAU documented in contract or SOW; incremental ROAS threshold at agreed maturity. |
FAQ
What is BAU conversion in a pLTV test?
BAU conversion is the existing conversion event setup (names, values, timing, and delivery paths) kept unchanged in a control group while a treatment group receives the new pLTV or value signal.
How is BAU different from a holdout?
BAU describes what the control experience is. A holdout is how you assign users or traffic to control vs treatment. You can run BAU in all non-test traffic while holding out a randomized segment, or split campaigns explicitly.
Can BAU change during a pilot?
Changing BAU mid-test confounds readout. If you must change control setup, end the current experiment and start a new one with a fresh baseline.
Should BAU and treatment have the same budget?
Matched spend simplifies incremental ROAS interpretation. If budgets differ by design, pre-register how you will adjust for incremental spend in readout.
What metrics do you compare BAU against treatment on?
Pre-register primary metrics: incremental conversions, incremental revenue, incremental ROAS, and secondary checks on customer quality (repeat, refund, churn) at cohort maturity.
Does BAU mean turning off server-side tracking?
No. BAU keeps your current tracking and value design. Treatment adds or replaces value signals (often pLTV) on a defined segment while control keeps prior behavior.
Who owns BAU documentation?
Joint ownership: performance marketing defines live setup; analytics validates it matches production; engineering confirms routing and event schemas in deployment.
Not the same as
| Term | Difference |
|---|---|
| Holdout test | Assignment mechanism; BAU is the control experience being held out or compared. |
| Control group (generic) | Any experiment control; BAU specifically means your pre-change conversion setup. |
| Proxy metric | Short-window event BAU may use; BAU is the full live setup, not just the metric label. |
