Performance Max, running on default settings, counts a first-time buyer and a fourth-time returner as the same thing: a conversion.
That assumption is not neutral. Returning customers convert at a lower cost because they already know the brand. When PMax optimizes toward the cheapest conversion, it will favor return traffic over new buyers - not because it is broken, but because that is precisely what it was asked to do. High value new customer mode pmax exists to change that instruction.
The result without it: PMax becomes an efficient retention engine dressed as an acquisition campaign, and new customer acquisition cost ecommerce quietly rises as a share of real acquisition spend while the dashboard shows healthy conversion volume.
The finding: PMax's default optimization treats acquisition and retention identically. High-Value New Customer Mode rewires the signal so the algorithm actively seeks buyers it has never seen - and bids accordingly.
What we actually measured
This analysis draws from DTC/ecommerce and apparel accounts running PMax as the primary campaign type, audited across 2025-2026, where brand exclusions were already in place. In each account, the Insights tab showed a consistent pattern before any new-customer mode was active: a meaningful share of "conversions" were attributable to audience segments overlapping with existing customer lists.
PMax was not underperforming. It was doing exactly what the optimization goal asked of it. The problem was the goal itself.
High-Value New Customer Mode in PMax: Two Settings, Not One
Most operators who have heard of high value new customer mode pmax assume it is a single toggle. It is not. It sits inside the Campaign-level Customer Acquisition Goal and has two distinct operating modes with different bidding logic and different deployment contexts.
| Mode | What it does | When to use |
|---|---|---|
| Bid higher for new customers | Applies a value premium on new-customer conversions while still capturing return traffic at a lower bid | Accounts wanting to grow acquisition and retain simultaneously |
| New customers only | Restricts delivery entirely to audiences not identified as existing customers | Accounts where every ad dollar must target genuinely new buyers |
"Bid higher for new customers" is the more flexible entry point. It does not exclude returning buyers - it tells the algorithm to bid more aggressively when a potential impression matches someone outside your customer list. Return traffic still converts at a lower ceiling. New customers are prioritized in the auction.
"New customers only" is a harder constraint. It excludes return purchase intent from the campaign entirely. That is the right call in specific situations - a new-segment launch or an account with a separate retention campaign - but it is not the correct default for most ecommerce accounts.
The distinction matters because most accounts that enable HVNC mode choose the wrong version for their context, then attribute the CPA increase to the feature rather than the configuration.
The machine has no concept of whether it is acquiring or retaining by default. High-Value New Customer Mode is how you give it one.
Setting the New Customer Value: The LTV Gap Method
The setting most operators guess at is the new customer value increment - the additional value you assign to a new-customer conversion on top of your standard conversion value.
Getting this wrong in either direction is expensive. Too high, and the algorithm chases new-customer signals at any new customer acquisition cost ecommerce, burning CAC on a bet that lifetime value will justify it. Too low, and the bid premium is insufficient to actually shift acquisition behaviour.
The LTV Gap Method builds the increment from two real numbers in your business:
- Calculate average 12-18 month LTV for a customer first acquired through paid channels.
- Calculate average 12-18 month LTV for a customer who arrived through email re-engagement, direct, or repeat purchase.
- The difference between those LTV figures is your ceiling for the new customer value increment. Set the increment at 50-60% of that gap to avoid overcorrecting tROAS into unprofitable territory on the acquisition side.
For an account where new paid-channel buyers carry an 18-month LTV of $320 versus returning customers at $180, the LTV gap is $140 and the correct increment ceiling sits around $70-85 - not the full $140, which would overcorrect tROAS on the acquisition side.
If you do not yet have reliable LTV cohort data by channel, start conservative - a $20-40 increment on a $50-70 CPA target - and widen it as cohort data confirms the lifetime value premium is real.
Audience Signals Are the Prerequisite, Not the Add-On
High value new customer mode pmax only functions at the precision it promises when the customer list defining "existing customer" is clean, current, and large enough for Google to model from.
The performance max audience signals at the campaign level serve two roles here: they tell the machine who counts as an existing customer to exclude or de-prioritize in bidding, and they provide a directional pattern toward the buyer type that reliably converts. Without a strong customer list, the exclusion logic loses precision and the bid premium fires in the wrong direction.
Three requirements before enabling either mode:
- Upload your full purchaser list - matched users only - and refresh it every 30 days at minimum.
- Apply the list explicitly as the "existing customers" definition inside the Customer Acquisition Goal, not just as a general audience signal in the asset group.
- Reach 50 conversions in the campaign before reading new-customer CPA as a meaningful trend. The same learning window rule that governs standard PMax applies here - and new-customer mode adds a learning layer on top of it.
For the underlying asset group signal structure that supports effective HVNC optimization, the PMax asset groups guide covers how to organise intent-based groups around the customer signals that feed accurate new-customer identification.
Where This Falls Apart
HVNC mode has clear failure conditions worth naming before committing budget to it.
Low-volume accounts generating fewer than two conversions per day. New-customer conversions are a subset of total conversions. On a campaign generating ten to fifteen conversions per week, the new-customer subset is too thin for the algorithm to learn from in a reasonable window. The performance max audience signals mechanism fires on too few impressions to develop a pattern worth trusting.
Accounts without clean first-party data. If you have not collected a purchaser list, or are not in a position to upload it to Google, you cannot define "existing customer" at the precision HVNC mode requires. The machine falls back to probabilistic modelling with no clean reference point to exclude against.
In both cases, focus on tightening audience signals in existing asset groups first. The Performance Max best practices guide covers the readiness criteria for each additional layer of PMax control.
What to Do With This on Monday
The operator move here is a campaign-level change, not a rebuild:
- Open the PMax campaign driving your primary acquisition volume.
- Navigate to Campaign Settings > Additional settings > Customer Acquisition Goal.
- Upload your most recent purchaser list and designate it as the existing customer definition.
- Enable "Bid higher for new customers" - not "new customers only."
- Apply the LTV Gap Method to calculate the new customer value increment. Start at 50% of the LTV gap, not 100%.
- Set the next review at 50 new-customer conversions, not a calendar date.
BAVai tracks the new-customer conversion split across every Google account we manage each morning, flagging when new-customer CPA climbs past the kill threshold before it accumulates into a week of wasted premium bidding. The real-time ad alerts framework explains why daily detection frequency beats any weekly manual review.
The economics of new customer acquisition cost ecommerce are expensive by design - you are paying to earn a buyer with no prior relationship against the much lower resistance of returning customers. High-Value New Customer Mode does not change that equation. It makes PMax intentional about which side it is solving for.
When you look at your current PMax conversion data, do you actually know what share of those conversions came from people who had never bought from you before - or are you scaling a campaign that is quietly optimising for return traffic dressed as acquisition?