Cross-Channel Attribution
Cross-channel attribution distributes conversion credit across all marketing touchpoints in a customer's journey — paid social, search, email, affiliate and organic.
4 min readRead →
One question per entry. Plain language, the formula, and our take. Forty entries across five clusters, plus long-form pieces on the numbers a board actually underwrites.
Incrementality testing, media mix modeling, holdout tests, multi-touch attribution, and what the signal loss after iOS 14 actually broke — and how to fix it.
Explore →LTV:CAC, blended vs incremental CAC, payback period, contribution margin, and LTV forecasting — the ratios that determine whether growth is efficient or just expensive.
Explore →Creative fatigue, ad frequency, hook rate, thumb-stop rate, UGC vs produced creative, and platform-specific principles for Meta, TikTok, and YouTube.
Explore →Growth loops, growth governance, northstar metrics, experimentation velocity, and the operational difference between an embedded team and an agency retainer.
Explore →First-party data strategy, CAPI, modeled conversions, data clean rooms, predictive LTV, and cohort analysis — what to trust when the platform signal is broken.
Explore →The board audits the CAC to the cent and waves the LTV through. The margin underneath it is where the ratio breaks.
Every company I have worked inside has the same meeting.
Marketing says the channel is profitable. Finance says the company is burning cash. Both are reading a correct number. They are reading different margins.
What your media is actually causing — vs. what it's only observing.
Cross-channel attribution distributes conversion credit across all marketing touchpoints in a customer's journey — paid social, search, email, affiliate and organic.
DTC attribution is the practice of connecting sales to the marketing channels that influenced them in direct-to-consumer businesses, where the platform and the ledger rarely agree.
Holdout testing is a causal measurement technique that withholds advertising from a defined group — a geographic market, a user segment, or a time window.
Incrementality testing measures whether your advertising actually causes conversions — or whether those customers would have bought anyway.
Last-click attribution assigns 100% of conversion credit to the final touchpoint before a purchase, and systematically overpays the channels closest to checkout.
Media mix modeling is a statistical technique that measures the contribution of each channel to total revenue using aggregate, privacy-safe data.
Multi-touch attribution distributes conversion credit across multiple touchpoints using rules or machine learning, and depends on user-level tracking to work.
Signal loss after iOS 14 refers to the structural reduction in user-level tracking data available to ad platforms following Apple's App Tracking Transparency.
View-through attribution credits an ad impression — not a click — with contributing to a conversion, on the assumption that seeing an ad changed behavior.
The ratios that determine whether growth is efficient or just expensive.
The margin a customer actually produces after variable fulfillment, processing, returns and discounts — the only basis that prices acquisition correctly.
The ratio a board treats as gospel. Its denominator gets audited to the cent; the margin inside its numerator usually gets waved through.
How many months of contribution it takes to recover fully loaded acquisition cost — the number that sizes how much acquisition the balance sheet can carry.
Blended CAC divides all spend by all new customers. Incremental CAC prices the next customer. Only one of them should set a bid ceiling.
Media plus agency fees, tools, creative production and allocated salaries. The version of CAC an investor will build their own model on.
Projecting lifetime value from order frequency, churn and margin — and the assumptions that make a forecast either useful or decorative.
Reading revenue and retention by acquisition month rather than in aggregate, so improvement and mix shift can be told apart.
Labor does not scale smoothly. It steps. The gap between steps is hidden capacity — zero-marginal-cost runway before the next hire.
The discounted lifetime contribution margin of every current and future customer, net of acquisition cost. The asset marketing builds while being measured as a cost.
What makes creative work, and what makes it stop working.
The decay in performance as an audience sees the same asset repeatedly — and how to distinguish it from auction pressure or seasonality.
Average impressions per person over a window. A diagnostic for saturation, not a target to optimize toward.
The share of impressions that survive the first three seconds. The earliest reliable signal that a concept will or will not carry spend.
A stricter cousin of hook rate used on short-form feeds, measuring genuine scroll interruption rather than passive autoplay.
When native, low-fidelity creative outperforms studio production — and the point in the funnel where that reverses.
Structuring concept, format and message tests so results attribute to a variable rather than to the algorithm.
What each platform rewards, how its auction reads early engagement, and why the same asset performs differently across the three.
How many concepts a program needs per month to hold performance, and the operational shape required to produce them.
The operating layer that makes growth repeatable rather than heroic.
Compounding acquisition mechanisms where output from one cycle becomes input to the next, in contrast to linear funnel spend.
Who decides what gets tested, who approves spend shifts, and how decisions are recorded so they survive turnover.
Choosing the single measure a team steers by, and the counter-metrics that stop it from being gamed.
Tests shipped per month at sufficient power. The rate limit on how fast a growth program can actually learn.
The operational difference between owned capacity and contracted capacity, and where each one structurally fails.
One signed page defining every classification ruling, split ratio and step threshold, changed only at a scheduled review.
Close the books, run the ledger through the contribution model, compare the dashboard on three deltas, adjust at that meeting.
The permanent layer — data, routines, definitions and reporting — left behind when an engagement ends.
What to trust when the platform signal is broken.
Collecting, consenting and structuring customer data you own, so measurement survives the next platform change.
Server-side event transmission that restores conversion signal lost to browser and app-level tracking restrictions.
Statistically estimated conversions the platform reports when it cannot observe them — and how much weight they should carry.
Privacy-preserving environments for matching advertiser and platform data without exposing user-level records to either side.
Modeling expected customer value from early behavior to bid on quality rather than on volume.
Deduplication, event-match quality, attribution windows and tracking gaps — the audit that precedes any measurement conclusion.