A quick primer
The key distinction is internal (a channel-mix shift, driven by the system's own behavior) versus external (a real market change), and Google's own troubleshooting guide for the channel performance report speaks directly to this: PMax optimizes across channels in real time to maximize overall campaign return against your core conversion goals, so it's important to consider both the overall result and each channel's numbers together, not in isolation. If overall ROAS fell but ROAS within each individual channel stayed stable, that points to a mix effect (budget shifted toward a channel with a naturally lower ROAS). If ROAS fell inside every channel at the same time, that looks more like an external factor hitting the whole market rather than PMax's own internal allocation.
A documented external driver worth checking directly is seasonality and competitive dynamics: marginal cost optimization means the system constantly hunts for the most cost-efficient conversion at any given moment, and conversion cost naturally shifts with the dynamic auction environment, which is affected by competitor activity, seasonality, and swings in user demand. In other words, a ROAS dip can reflect tougher competition or genuinely softer demand rather than a "wrong" reallocation by PMax.
It's also worth accounting for an attribution effect that can mask one cause as the other: with last-click attribution, a channel can look like it has low ROI when it's actually playing an important role earlier in the path to conversion; last-click often fails to capture the full contribution of channels that engage users earlier in their journey. So a channel's apparent ROAS drop is sometimes an artifact of the attribution model, not a real drop in its value.
What to check before you touch anything
- ROAS inside each individual channel over time (through the channel distribution table). Is it stable channel by channel, or has it dropped everywhere at once.
- How spend share shifted between channels over the same period. Growth in the share going to a channel with a naturally lower ROAS mechanically lowers the campaign-wide figure without any decline inside that channel.
- Whether there are external signals of seasonality or softer demand: year-over-year comparisons, industry trends, changes in competitor behavior.
- What attribution model is in use. With last-click, channels that engage users earlier in the funnel can look systematically weaker than they really are.
- Whether the ROAS drop coincided with a change in goals, bid strategy, or new assets. The time series chart shows this clearly.
Possible approaches
- Split the ROAS change into a mix effect (share shifted between channels) and a within effect (ROAS changed inside each channel), comparing the channel distribution table over two comparable periods. If within is stable and only the overall number moved, that's a mix effect, not a demand problem.
- If ROAS dropped inside every channel at once, check the period against seasonal patterns and external demand signals before treating this as a campaign setup issue.
- If you're on last-click attribution, try switching to data-driven attribution to see whether the model itself is masking a channel's real contribution.
- Don't artificially "pin" channel mix in response to a ROAS dip. The distribution across channels is a result of the optimization, not the cause requiring manual intervention.
- Evaluate the trend over a long enough window (two or more weeks, or a full conversion cycle), accounting for conversion delay, before concluding anything about the nature of the drop.