Where your PMax money actually goes (read the channel report)

The Performance Max campaign shows 3.5x ROAS on ₹8,00,000 a month. For a long time, that single blended number was all anyone got — Google chose where the money went and reported one result.

Open the channel performance report, though, and the picture splits. Search and Shopping turn ₹3,60,000 into ₹21,84,000 — about 6.1x. YouTube turns ₹2,00,000 into ₹1,68,000 — 0.84x. Display and Discover sit somewhere in between, and both below breakeven.

A 3.5x campaign can be a 6x engine towing three passengers. You still can't set budgets per channel inside PMax, but you can now see who's pulling their weight — and you have more levers than most accounts use.

One campaign, many storefronts

Performance Max buys inventory across Google's surfaces from a single budget. The channel performance report — under Insights & reports → Channel performance — breaks results out by Search, YouTube, Display, Discover, Gmail, Maps and Search partners. Through 2026 Google kept adding to it: downloads, splits by conversion action and ad event type, and diagnostics for issues like missing assets or bid targets limiting delivery.

Spend share vs. value share

The fastest read of the report is two columns side by side: share of spend and share of conversion value. A channel taking a quarter of the money and returning 6% of the value needs a good explanation.

Two explanations are legitimate. Upper-funnel channels can create demand that converts later through Search — real, but something to test, not assume. And their conversions may be counted differently.

Is each channel paying its way?

Judge each channel against your breakeven ROAS (1 ÷ gross margin), then again with view credit removed. A channel that fails both is spending your money to collect credit.

The channel scorecard in SQL

Download the channel report — or load it through a connector — into a daily table, and one query gives you every channel's shares, ROAS and click-only ROAS:

WITH by_channel AS (
  SELECT
    channel,
    SUM(cost)                                              AS spend,
    SUM(conv_value)                                        AS value,
    SUM(IF(ad_event_type = 'CLICK', conv_value, 0))        AS click_value
  FROM pmax_channel_daily
  WHERE campaign = 'PMax | Catalogue'
    AND day >= DATE_SUB(CURRENT_DATE(), INTERVAL 30 DAY)
  GROUP BY channel
)
SELECT
  channel,
  ROUND(spend)                                  AS spend,
  ROUND(spend / SUM(spend) OVER () * 100, 1)    AS spend_share_pct,
  ROUND(value / SUM(value) OVER () * 100, 1)    AS value_share_pct,
  ROUND(value / NULLIF(spend, 0), 2)            AS roas,
  ROUND(click_value / NULLIF(spend, 0), 2)      AS click_only_roas,
  value / NULLIF(spend, 0) >= 2.5               AS above_breakeven   -- 1 ÷ your margin
FROM by_channel
ORDER BY spend DESC;

Run it monthly. The mix shifts as PMax learns, as assets change and around sale periods — a channel that paid its way in March may not in October.

The levers you actually have

You can't type a budget next to "YouTube". You can change what PMax has to work with, and that moves where it spends.

• Assets. No video? PMax may assemble its own from your images. Weak video means weak YouTube results; strong video can turn the channel around. Read the channel diagnostics first.
• Signals and exclusions. Audience signals steer PMax toward better people; brand exclusions, negative keywords and URL exclusions stop spend on searches and pages you don't want.
• Structure. Run standard Search or Shopping alongside PMax for your best terms and products, leaving PMax less room to lean on easy conversions.

A monthly 20-minute review

1. Pull the channel reportSpend, conversion value and ad event type by channel for the last 30 days. Let the latest week's conversions mature first.
2. Compare spend share with value shareFlag any channel taking more of the budget than it returns in value.
3. Strip view credit and check breakevenRecompute flagged channels on click-only value. Below breakeven on both? It needs proof to keep its share.
4. Fix inputs before cuttingMissing or disapproved assets, or bid targets limiting delivery — fix those first.
5. Test what's leftA geo or time holdout tells you whether Search actually falls without the weak channel.

• Never report PMax as one number again — put channel ROAS and value share next to the campaign total.
• Segment by ad event type — show click and view conversions separately for YouTube and Display.
• Keep your brand out of the picture — brand exclusions stop a strong Search number from just being people searching your name.

Quick gut-check