When our Meta cost per purchase or ROAS gets worse week over week, how do we find which lever moved it: CPM, click-through rate, conversion rate or order value?
Our answer
First confirm that store revenue and orders fell too. Then put CPM, click-through rate, conversion rate and order value side by side, and fix the lever that moved: auction, ad, site or product mix.
- Confidence
- Medium
- Strongest proof
- 3 datasets
- 4 practitioners, 1 opinion or claim
- Sources
- 7 sources
- 8 statements
MIAGOKBH
New here? Take the 1-minute tour
Why this is the answer
3 reasons and our judgmentOnly four levers move cost per purchase and ROAS: CPM, click-through rate, conversion rate and order value (Marin Istvanic, a creative strategist). A 20% weekly ROAS drop can be four 5% moves or two bigger ones, so look at all four together.
Which levers moved shows what broke. Higher CPM with steady click-through is the auction. Lower click-through is the ad. Lower conversion with steady click-through is the site. Steady cost per purchase with lower ROAS is order value (Affect Group). Flighted agrees on CPM and conversion.
In 2026, CPM, click-through and conversion rates moved different ways. Northbeam found Meta CPMs up 10 to 20% from January, click-through at 12-month highs and conversion rates falling. Haus brands saw CPMs about 10% above a year earlier and no offsetting conversion rise (Olivia Kory).
Our judgment
The order is our own: store revenue first, then all four levers at once, joining Marin Istvanic's split, Affect Group's patterns and Flighted's advice to confirm the drop against store revenue. Read conversion rate beside cost per purchase, not alone. Northbeam shows conversion rate falling while new-customer acquisition cost still fell until February 2026, and Affect Group notes that a higher conversion rate can outweigh a CPM rise. Act on a lever only once its move has held for days, not after one bad day (Moving budget mid-month). Jetfuel Agency acts on a click-through drop of 15 to 20% or more held for 5 days, or cost per purchase up 25% for 3 days, and Affect Group uses a 7-day or 14-day view.
Our judgment, built from Marin Istvanic, Affect Group, Flighted, Northbeam and Jetfuel: our reading of these sources, not a finding from any one of them.
What the sources say
8 statements from 7 sourcesHaus says Meta CPMs this fall rose about 10% year over year in the typical market, not only election-contested ones, and conversion rates did not rise to offset it. Haus sells incrementality testing.
“Conversion rates did NOT increase to compensate higher CPMs, resulting in ~8% worse pixel-attributed efficiency relative to August.”
Marin Istvanic says only four metrics move performance: CPM, CTR, AOV and conversion rate, so check all four to see what changed ROAS or CPA. His splits are examples, not data.
“always look at those 4 metrics to find what impacted performance and to what extent”
Affect Group, an agency, says to find which metric caused a drop before changing a campaign: higher CPM points to costlier inventory, lower CTR to weaker ads, lower conversion rate to site problems.
“Before changing an advertising campaign, it is useful to determine which metric is responsible for the deterioration in performance”
Ben Heath says higher CPMs need not mean worse results, since Meta may be better at turning prospects into customers and get the same sales from fewer impressions.
“Meta is now able to do a better job of turning your prospects into customers. So you can get the same amount of sales from fewer ad impressions.”
Jetfuel, an agency, runs a fixed weekly rhythm of audit, creative decision and brief, and acts only on signals that hold: CTR down for 5 days, CPA up 25% for 3 days, or frequency above 3.5 for a week.
“Hold on acting until CPA confirms.”
Northbeam's H1 2026 review of thousands of advertisers says Meta conversion rates fell by as much as 50% up to January 2026 as click-through rates rose, which it reads as more low-intent traffic.
“This is suggestive of increased low-intent traffic dragging down normal conversion rate numbers.”
Flighted, an agency, says to confirm a Meta ROAS decline before fixing it: if platform ROAS fell but Shopify revenue held flat on stable spend, it is an attribution problem. It shows no data.
“If platform ROAS fell 20% but backend Shopify revenue held flat against stable total spend, you have an attribution problem, not a performance problem”
Northbeam's data on 950+ advertisers shows Meta CPMs up 10% to 20% from January 2026 and CTRs at 12-month highs, while conversion rate fell in step with cost per click. Northbeam sells attribution.
“although conversion rate is dropping, this appears to be a function of increased traffic coming through ads”
In the full answer
For readers who sign in- What to do this week3 steps
- The numbers behind it3 figures from cited sources
- The strongest disagreementThe best case against the answer, from a cited source
- How sure we areWhat its confidence rests on, and what would change the answer