eCPM

What eCPM measures

Effective cost per mille expresses advertising revenue per thousand impressions:

eCPM = (Ad revenue ÷ Impressions) × 1,000

The "effective" matters: CPM is what a buyer pays per thousand; eCPM is what a seller earns per thousand, computed after the fact across whatever mix of pricing models actually delivered — CPM, CPC, CPA deals all collapse into one comparable revenue rate. That makes eCPM the standard unit for a publisher or app developer comparing demand sources, placements, formats, and geographies.

If 240,000 interstitial impressions earned $1,920, that placement ran a $8.00 eCPM. Whether that's good depends entirely on what an alternative network, format, or floor setting would have earned for the same impressions — eCPM is a comparison tool, not a scoreboard.

What moves eCPM

• Format. Rewarded video consistently clears the highest eCPMs (opt-in completion is valuable), interstitials sit in the middle, banners at the bottom — often by an order of magnitude between extremes.
• Geography. Advertiser budgets concentrate in tier-one markets; the same placement can clear at several times the rate in the US versus emerging markets.
• Targeting signal. Impressions with usable identifiers and consent monetize meaningfully higher than signal-less ones — the iOS ATT era made this visible as a persistent iOS/Android eCPM gap in many categories.
• Auction pressure. More qualified bidders per impression raises clearing prices; this is the entire premise of ad mediation and in-app bidding.
• Seasonality. Q4 budgets inflate eCPMs; January deflates them. Year-over-year comparisons beat month-over-month for trend judgment.
• Placement quality. Viewable, expected, non-accidental impressions train buyers' models to bid; misclick-prone or instantly-closed placements train them to avoid you.

The trap: eCPM without volume

eCPM is a rate. Revenue is rate × volume:

Revenue = eCPM × Filled impressions ÷ 1,000

Almost every eCPM mistake is a volume blindspot wearing rate-colored glasses:

• A price floor that lifts eCPM 20% while fill rate drops 40% is a revenue cut.
• A network showing spectacular eCPM on tiny volume is cherry-picking premium impressions; route everything to it and the average collapses.
• An aggressive interstitial cadence lifts session eCPM while shortening sessions and retention — the ARPDAU view exposes what the eCPM view hides.

This is why the operating metric hierarchy in app monetization runs ARPDAU (user-level truth) → revenue per placement → eCPM and fill (diagnostics underneath).

Working with eCPM day to day

1. Compare like with like. Same format, same geography, same period. A blended eCPM across formats and countries is an accounting average with no decision content.
2. Diagnose by decomposition. When eCPM falls: did demand weaken (bids down), did mix shift (more low-value geo/placement traffic), or did signal degrade (consent rates, identifier loss)? Each has a different fix.
3. Test floors against holdouts. In platforms like AdMob, AppLovin MAX, and Unity LevelPlay, floor experiments are cheap to run and routinely surprising — measure revenue per thousand requests, not per thousand filled impressions, to see the real effect.
4. Watch eCPM by network honestly. In bidding setups, per-network eCPM differences mostly reflect which impressions each network wins, not which network "pays better" — re-routing rarely replicates the rate.
5. Track the iOS/Android and consent splits. Signal-driven eCPM gaps tell you how much your inventory's value depends on targeting data, and how exposed it is to privacy shifts.

Common mistakes

• Choosing a network by eCPM alone. The network with the highest rate on 2% of your traffic is not your best network; marginal revenue at realistic volume is the test.
• Comparing different countries or formats directly. "Banner eCPM is terrible compared to rewarded" and "Brazil monetizes worse than the US" are category facts, not optimization findings.
• Ignoring unfilled requests. eCPM is computed on filled impressions; a placement can show a rising eCPM while serving fewer and fewer ads. Revenue per request is the honest denominator for placement health.
• Chasing eCPM into user experience damage. More aggressive formats and frequencies lift today's rate and tax tomorrow's retention — the cost shows up in ARPDAU and cohort curves weeks later.
• Treating eCPM forecasts as commitments. Network and mediation dashboards show recent averages; actual clearing prices swing with seasonality, demand shifts, and your own traffic mix.

FAQ

What's the difference between eCPM and CPM? CPM is a buying price (advertiser pays per thousand impressions). eCPM is a selling yield (publisher revenue per thousand, whatever the underlying pricing). Buyers optimize CPM down; sellers optimize eCPM up — same unit, opposite sides.

What's a good eCPM? Unanswerable without format, geography, platform, and season. Rewarded video in tier-one markets can clear tens of dollars; run-of-site banners in low-CPM regions clear cents. The useful benchmark is your own placement's history at constant mix, and a split test against an alternative setup.

Why did my eCPM drop suddenly? Check in order: seasonality (post-Q4 cliff), traffic mix shift (geo, format, new placements), consent/identifier changes (signal loss), a demand source dropping out, and policy issues limiting bid eligibility. The cause is usually visible in segmented reporting before it's explicable in the blended number.

Should I optimize eCPM or fill rate? Neither in isolation — optimize revenue per thousand requests (and ultimately ARPDAU). eCPM and fill are the two levers underneath; floors and demand changes trade one against the other.

How does eCPM relate to web RPM? RPM values a thousand page views; eCPM values a thousand ad impressions. RPM ≈ eCPM × ads per page × fill ÷ 1,000. App monetization uses the eCPM/ARPDAU pair the way web publishing uses eCPM/RPM — the app monetization path covers the full stack.