Blog

GA4 sessions run lower than UA and yesterday's conversions keep changing — not bugs, but session definitions, processing lag, thresholds, and attribution.

Meta, GA4, MMP, and your order DB all report different conversion counts — because of attribution windows, view-through, credit date, and last-touch overlap.

Auto-bidding and AI are said to replace marketers, but what machines took over and what's still yours are different things. Here's what machines structurally can't do.

Misread the learning phase and your budget leaks. Why CPA bounces during learning, why touching it resets everything, and why you shouldn't trust the numbers at face value.

Raising the budget on a winning campaign and watching CPA climb usually means you've hit the flat part of the response curve — saturation. Here's how to break the budget-up → frequency-up → CTR-down → CPA-up domino.

Cutting budget the moment ROAS drops usually backfires. Here's how to split the channel-level gap and reallocate based on marginal ROAS — the efficiency of the 'next won' — with examples.

Stop timing creative swaps by feel. When CTR and frequency pull apart into a 'scissors' shape, that's your fatigue signal — plus how to tell fatigue from other causes.

When CPA spikes, splitting the rise into mix effect and efficiency effect shows you the real cause — walked through with a $8/$12 channel example.

Neither narrow targeting nor broad is always right. Here's the reach/CPM/conversion-rate/audience-exhaustion tradeoff, and why narrow targets break when you scale budget.

Don't reach for creative changes when ad performance suddenly drops. A 4-step diagnosis order — check tracking, split by channel, separate mix from efficiency, trace the funnel — with a symptom map.

'What should I learn to become a performance marketer?' Here's the growth path — operations, data, decision-making, causal inference — and why tools should be learned by function, not name.

Skip memorizing the full glossary. Here are the 4 metrics — CTR, CVR, CPA, ROAS — a junior performance marketer should actually watch, and how to read them together.

Instead of memorizing CPI, CPA, ROAS, and LTV:CAC separately, read them as one connected chain and you'll see exactly where performance is leaking. A metrics primer for junior marketers.

Marketing Mix Modeling (MMM), the channel-contribution method built for a world of cookie loss and iOS tracking limits. Regression, adstock, saturation, contribution decomposition, and where it breaks down.

Why does pouring budget into your most efficient channel lose you money? The trap of average efficiency, and how to split channel budgets using 'marginal utility,' explained with practical examples.

Dashboard ROAS should only be half-trusted. A breakdown of incrementality measurement and three holdout experiment designs for measuring what ads actually create.

The difference between correlation and causation, explained through ice cream and drowning. Why confounders and reverse causation create illusions, and why an experiment is the only way to confirm causation.

A/B testing done right — from random assignment, sample size, and statistical significance, to the early-stopping (peeking) trap that catches nine out of ten people. A practical guide.