Pulsar
metrics playbook · efficiency & fatigue
Present: ← → Space · Esc back to Doc

Pulsar · Metrics playbook · derived from a live cockpit session. All example numbers observed on the test ad account (currency: EGP).

Efficiency & Fatigue

Reading CPM, CTR, CPA and Frequency the honest way — a Pulsar field guide.

Contents

01 The core principle · 02 The metric web · 03 CPM × Frequency · 04 CTR × Frequency · 05 How the verdict is computed · 06 Worked examples · 07 The visualization lesson · 08 Interactive diagnoser · 09 Quick reference glossary

01

The core principle

A rate or cost metric is AMBIGUOUS on its own. It is only interpretable next to its delivery-shape companion — FREQUENCY — plotted OVER TIME.

Because: CPM = Spend ÷ Impressions × 1000, and Impressions = Reach × Frequency. So a LOW CPM can mean either cheap FRESH reach (good) OR high frequency on a saturated audience (bad — fake efficiency, early fatigue). The number alone can’t tell you which; you must read it against frequency over time.

A low CPM is only good news if you’re reaching NEW people. If it’s the same people seen more often, cheap is a trap.
02

The metric web — formulas & the chain

Six formulas, one chain. Every downstream cost is an upstream cost propagated through a rate.

CPM= Spend ÷ Impressions × 1000 (auction price per 1,000 impressions)
CTR= Clicks ÷ Impressions (share who click)
CPC= CPM ÷ (1000 × CTR) = Spend ÷ Clicks (cost per click)
CVR= Conversions ÷ Clicks (share of clicks that convert)
CPA= Spend ÷ Conversions = CPM ÷ (1000 × CTR × CVR) (cost per acquisition)
Frequency= Impressions ÷ Reach (avg times each person saw it)
CPM ──÷(1000×CTR)──▶ CPC ──÷CVR──▶ CPA
CPA is CPM propagated down the funnel through CTR and CVR. A good CPM does NOT guarantee a good CPA. Rising frequency inflates CPA from BOTH sides at once — that’s why it’s the most dangerous signal.
Frequency’s double hit on CPA
MechanismEffect on the CPA fractionNet
Frequency ↑ → CPM ↑ (saturation) → numerator up CPA ↑
Frequency ↑ → CTR ↓ (fatigue) → denominator down CPA ↑
03

CPM × Frequency — the saturation / efficiency read

Pair the CPM direction with the frequency direction and the ambiguity collapses into one of five states.

The five-state decision matrixdirection pairs read over the same time window
CPM directionFrequency directionVerdictWhat it means
CPM ↓Frequency flat EFFICIENCY CPM is falling while frequency stays flat — you are reaching MORE fresh people for less. Genuine efficiency, the healthy version of a low CPM.
CPM ↓Frequency ↑ SATURATION CPM is falling while frequency rises — the cheaper impressions are the SAME people seen more often. This is saturation, not efficiency: marginal value is declining. Refresh the creative or widen the audience.
CPM ↑Frequency flat INFLATION CPM is rising at stable frequency — this is genuine auction-cost inflation (competition, audience quality or seasonality), not saturation.
CPM ↑Frequency ↑ WORST Both CPM and frequency are rising — you are paying more AND re-serving the same audience. The worst combination; rotate the creative.
CPM flatFrequency flat STABLE CPM and frequency are both stable — the CPM number is a clean, comparable auction cost this window.
When CPM and frequency are both climbing, you are paying more to re-serve the same people. That is a stop signal on its own — you don’t wait for CTR to collapse.

The live read — and the call: STOP / ROTATE

A real Pulsar read of a saturating creative. The two verdicts disagree — and knowing which one wins is the whole skill.

Pulsar Efficiency and fatigue over time: CPM and frequency both rising, flagged the worst combination, rotate the creative; CTR holding as frequency climbs, creative resilient so far.
Efficiency & fatigue over time, straight from Pulsar. The blue CPM line and the amber Frequency line are both climbing together, while the green CTR line is still holding.
×

EFFICIENCY (CPM × Frequency) — the worst combination. “Both CPM and frequency are rising — you are paying more AND re-serving the same audience. The worst combination; rotate the creative.”

≈

FATIGUE (CTR × Frequency) — holding, for now. “CTR is holding even as frequency climbs — the creative is resilient so far, but keep watching the frequency ceiling.”

!

The call: stop / rotate this creative now. The two signals disagree — and saturation wins. A holding CTR only means fatigue hasn’t shown yet; it is not a reason to keep spending. The money math has already turned against you: every extra impression costs more (CPM ↑) to reach someone who has already seen it (Frequency ↑). Waiting for CTR to fall is waiting for the autopsy — rotate the creative or widen the audience before it does.

?

Read the arc, not the last dot — two caveats before you pull the trigger. (1) The tail turns. Around 06-30 both curves bend back down; the final few days are recent and thin (and any pull-back you or Meta already made shows up right here), so weight the whole-window trend, not the last point. (2) Learning phase. This creative only started delivering around 06-24 — barely a week. While Meta is still learning (roughly the first ~50 conversions, or after any edit that resets it), CPM and frequency swing as the algorithm explores, and that early volatility looks almost exactly like saturation. Don’t rotate a creative that hasn’t exited learning on the strength of a climb that may just be it settling — confirm the trend persists once delivery stabilizes, then act.

Action by verdict: EFFICIENCY keep running · INFLATION hold, check bids/competition · SATURATION refresh the creative or widen the audience · WORST stop / rotate now · STABLE carry on.

04

CTR × Frequency — the fatigue read

FATIGUE LINK — Falling CTR while Frequency rises = the audience is worn out → refresh the creative.

The fatigue matrixthe same paired-direction read, applied to engagement
CTR directionFrequency directionVerdictWhat it means
CTR ↓Frequency ↑ FATIGUE The audience is worn out. Refresh the creative.
CTR ↓Frequency flat CREATIVE WEAKENING The ad itself is losing appeal — NOT overexposure. A refresh of the same concept won’t fix it; the concept is tired.
CTR flat or ↑any HEALTHY Engagement is holding or improving.

Fatigue gate — a frequency above ~2.5 with a falling CTR is what trips Pulsar’s Fatigue watch.

05

How the verdict is computed

Pulsar’s auto-verdict compares the FIRST-HALF average of the window against the SECOND-HALF average for each metric, with a ±5% flat band — inside the band the metric is read as ‘flat’, outside it as rising or falling. The paired direction (e.g. CPM-direction × frequency-direction) selects the verdict cell above.

Half-window averaging. The series is split at the window midpoint; each half is collapsed to its mean. The second-half mean is compared to the first-half mean — the shaded band is the ±5% flat zone around the first-half average.
06

Worked examples — real observed numbers

Two creatives from the same live session. One low CPM that is good news, one that isn’t — and only frequency tells them apart.

EXAMPLE A

“Bible Cover”

two lessons in one creative
EFFICIENCY CREATIVE WEAKENING
EFFICIENCY
CPM 46.7 → 34.9 EGP at flat frequency ≈ 1.2  ⇒  EFFICIENCY (cheaper FRESH reach; the good kind of low CPM).
CREATIVE WEAKENING
CTR 6.0% → 4.76% at flat frequency ≈ 1.18  ⇒  CREATIVE WEAKENING, not fatigue — because frequency is NOT climbing, Pulsar correctly does NOT flag it as fatigue.

Down-funnel reality check

Funnel over 31 days / 1,114 sessions — distinct-persona counts per stage

ViewContent 1,036 → AddToCart 49 → Checkout 13 → Shared details 14 → Purchase 2

EXAMPLE B

“T-shirt”

the classic bad pair
SATURATION
SATURATION
CPM 84 → 159 EGP WITH frequency 1.05 → 1.66  ⇒  SATURATION — the classic bad pair: paying more to hit the same people more often.
The frequency trace is the tell: 1.05 → 1.66 means the average person is already being re-served ½ more time than at the start of the window — the rising CPM is buying repetition, not reach.
07

The visualization lesson

Dual- and triple-axis overlays are the LEAST reliable way to compare CPM, CTR and Frequency: where the lines cross and how steep they look are ARTIFACTS of arbitrary axis scaling — they can manufacture a correlation that isn’t there.

The honest fix: INDEX every metric to 100 at its first delivering day, on ONE shared % axis. Then the DIVERGENCE is the signal. The ‘scissors’ = CTR falling below 100 while Frequency rises above it; Pulsar shades the danger zone from the day Frequency-index first exceeds CTR-index by more than 8 points.

Indexed to 100 — “Bible Cover”, first 12 delivering days

View the data (indexed values)

Product views available in Pulsar — Combined (indexed), Relations, Split, Overlay, 3D.

08

Interactive diagnoser

Set the two directions you’re seeing in the cockpit; the verdict updates live. This is the same pairing logic the matrices encode.

Read
CPM
Frequency
GOODEFFICIENCY

This tool is interactive in the HTML version. On paper, the full decision matrices in sections 03 and 04 are the equivalent cheat-sheet.

09

Quick reference glossary

The cheat-sheet. Print this page and pin it next to the cockpit.

Metric quick reference
MetricFormulaWhat it tells youGood direction
CPMSpend ÷ Impressions × 1000What 1,000 auction impressions cost you↓ only with fresh reach
CPCSpend ÷ ClicksWhat one click costs, end to end↓
CTRClicks ÷ ImpressionsHow compelling the creative is to its audience↑
CVRConversions ÷ ClicksHow well the post-click experience converts↑
CPASpend ÷ ConversionsWhat one acquisition costs — the bottom line↓
FrequencyImpressions ÷ ReachHow often the same person is re-servedflat, ≈1–2.5 watch >2.5
Reachunique people servedHow many distinct people saw the ad↑
Impressionstotal times servedRaw delivery volume (people × repeats)context

The one-line summary of this whole playbook: never read a cost or a rate alone — pair its direction with frequency’s direction over the same window, and let the pair name the verdict.

1 / 10