Every marketing intelligence tool will tell you which Meta ad set to scale and which to kill. Almost none of them will tell you to move US$5,000 from Meta to Google. The within-channel decisions are obvious and safe to surface. The cross-channel move is the unexplored lever, and for most Shopify merchants it is the highest-margin decision they could make this quarter.
Triple Whale, Polar, Northbeam, and Lifesight all stop at within-channel because the cross-channel call is harder to model and easier to blame the tool when it goes wrong. That is a real risk, but it also leaves the lever on the table.
Why cross-channel matters
Channels are not interchangeable, but they are substitutable at the margin. A dollar of paid spend has a job: deliver a qualified customer. If channel A is delivering customers at CAC US$42 and channel B is delivering them at CAC US$19 with comparable customer quality, leaving the budget split static is a choice. It is usually the wrong choice.
Most Shopify merchants run static channel mixes set at the start of the year, tuned within each channel, and never reweighted across channels. The mix that worked in Q4 holiday rarely works in Q2. Audience saturation, seasonality, and competitor moves all break a static mix within months.
The classification model
The cross-channel decision needs three buckets per channel:
- Over-performing. Top-quintile blended ROAS for the channel, healthy CAC, growing or stable conversion volume. Candidate to receive shifted budget.
- Under-performing. Bottom-quintile blended ROAS, rising CAC, declining conversion volume. Candidate to shift budget out.
- Neutral. Middle quintiles, stable trend. No action this cycle.
The classification runs on 14-day windows compared to the prior 14-day window. Anything shorter is noise. Anything longer misses the move.
A worked example
Multi-brand operator running a baby apparel store on Shopify, 30-day window:
- Meta: US$8,000 spend, blended ROAS 1.6, CAC US$42, conversion volume declining 18 percent week-on-week. Classification: under-performing.
- Google Shopping: US$4,000 spend, blended ROAS 4.2, CAC US$19, conversion volume growing 8 percent week-on-week. Classification: over-performing.
- TikTok: US$1,500 spend, blended ROAS 2.1, CAC US$31, conversion volume flat. Classification: neutral.
The cross-channel move: shift 20 percent of Meta spend (US$1,600) to Google Shopping. Hold TikTok at current spend. Re-evaluate in 14 days.
Typical outcome on a move like this over the next 30 days:
- Blended ROAS lifts from 2.5 to 2.9 (+16 percent)
- Total conversion volume holds or increases (Google scales up faster than Meta scales down)
- Total spend stays flat at US$13,500
The numbers vary by brand and category. The direction is consistent: when one channel is decisively under-performing and another is decisively over-performing, moving 15 to 25 percent of the under-performer's budget to the over-performer almost always lifts blended ROAS.
Why you cannot just shift everything
Two reasons.
Audience saturation in the receiving channel. Google Shopping at US$4,000 a month is hitting one set of in-market shoppers. Doubling spend to US$8,000 in a single month does not double the conversions. The new dollars chase lower-intent users at higher CAC, and the over-performer regresses to neutral or worse.
Algorithm reset risk. Large budget changes on Meta and Google both reset the platform's learning phase. A 50 percent budget cut on Meta in week one followed by another 50 percent cut in week three teaches the algorithm nothing useful. Smaller cuts (15 to 25 percent at a time) preserve the learning.
The right cadence is 15 to 25 percent moves every 14 days until the classification changes. A channel that was under-performing at week zero might be neutral by week six and not need another cut.
Confidence scoring
Three checks before recommending the shift:
- Volume floor. The receiving channel needs to support at least 30 conversions per week at the new budget level. Below that, the post-move ROAS is too noisy to evaluate.
- Trend persistence. The under-performing channel's decline has to hold for at least 14 days. A single bad week does not justify a structural budget shift.
- Margin sanity. Receiving channel CAC plus a 10 percent safety buffer must stay under your contribution margin per order. Otherwise scaling lifts ROAS but burns cash.
When not to reallocate
Three situations where the cross-channel move is the wrong call even when the classification flags it.
You are inside a launch window. A new product, a new audience expansion, or a seasonality ramp. The within-channel learning is not stable yet, and pulling budget out of a channel that has not finished learning resets the algorithm at the worst possible moment.
The under-performer is your only brand-awareness channel. Cross-channel ROAS comparisons assume comparable customer quality. A Meta prospecting campaign that builds brand awareness for a Google Shopping retargeting campaign is not actually under-performing; it is feeding the over-performer. Cut it and the over-performer regresses inside two months.
Total spend is small. Below US$3,000 a month in blended spend, channel-level statistical significance is too noisy to act on. Tune within channels at that scale and revisit cross-channel reallocation when total spend supports it.
What oddly does about this
oddly's cross-channel reallocation lives in the Steer tier. The classification runs weekly across Meta, Google, TikTok, and any other connected channel. The dashboard surfaces the recommended move with a one-paragraph rationale, the expected blended ROAS lift, and the confidence score. The shift waits for one tap. The platform proposes; you decide. The platform never increases your ad spend without you.