The method blog
One method at a time: how it works in plain words, the formula unpacked, where it lies to you, and a notebook and spreadsheet you can run on your own data. Each essay names its sources and shows its working. It is not marketing collateral. It is the maths, written to be checked.
Search budgets and bids · 20 min read · Interactive · Video
Where am I losing money right now, and can I trust Google's own estimate of the fix? Impression share and a replay of logged auctions
My search report already shows where the money leaks: the auctions the budget kept me out of and the ones I lost on bid. What it cannot say is what the fix is worth. I tested the kind of estimate Google offers, a replay of last week's auctions, on five hundred synthetic campaigns to see when it holds and when it carries last week into a different one.
Cross-platform reach · 24 min read · Interactive · Video
How many real people did I actually reach across platforms? Capture-recapture, the Petersen-Lincoln estimator
Meta and TikTok each report their reach, and adding the two counts everyone who saw both twice. I show what your own site can count exactly, how to bracket the real reach between a floor and a ceiling, why no report can give one total, and what the range lets you decide about the second platform.
Frequency and audience decay · 20 min read · Interactive · Video
How many times is too many, and how long does an audience stay worth targeting? A Hill curve and a survival half-life
I fitted two curves to reports you can get. One says where the next impression stops paying at your own prices, the other when an audience has given most of what it will give. I scored both on hundreds of synthetic audiences with the answers built in, and I say plainly what neither can see.
Forecasting · 17 min read · Interactive · Video
What will I get next month if nothing changes, and how far can I trust that number? A simplified Silverkite forecast
The finance lead wants next month's conversions by Friday. Last month's total, the recent average and a ruler through the chart all give a number, and none says how far off it might be. I built the forecast I use from LinkedIn's published method, checked its range on five hundred accounts, and found where a ruler beats it.
Seasonality · 11 min read · Interactive · Video
Why do some days always look worse, and when is it actually a problem? Seasonal indices
It is Monday, a little after ten, and somebody has posted Sunday's cost per sale at 30.14 euros next to Tuesday's at 24.14. Usually nothing went wrong. I show how to find each weekday's normal, how much history that takes, and the one case where the method invents a weekend effect that is not there.
Anomaly detection · 19 min read · Interactive · Video
Something moved. Is it a blip, a trend, or a broken thing? A median score, the Theil-Sen slope and change-point detection
Conversions are down, cost per acquisition is up, and somebody wants to know if it is a problem. I show how to tell one bad day from a slow slide or a step to a new level, how big each must be before your data can show it, and why the alarm line is a budget you choose.
Significance testing · 14 min read · Interactive · Video
Is this difference real, or am I looking at noise? A two-proportion test and the Benjamini-Hochberg correction
Mobile converts at 2.4%, desktop at 1.9%, and the meeting moves the budget. I show the two checks that gap has to pass first: that it is bigger than the clicks behind it allow, and that nobody found it by scanning forty rows. In my simulations, scanning alone produced a false finding in most tables, and correcting cost real finds too.
Creative fatigue · 20 min read · Interactive · Video
Is my ad wearing out, or did something else change? Segmented regression on exposure
Your best ad's click-through rate is falling and someone on the call says the creative is tired. I show why fatigue runs on how many times people have seen an ad rather than on days, how I fit when it started to tire and how fast, and how I tell that apart from a bid change or a competitor.
Conversion lag · 17 min read · Interactive · Video
Why do today's numbers look bad, and how long until I know the real result? Kaplan-Meier survival with a cure model
It is Monday morning and last week's conversions look down, though nothing in the account changed. Last week is not worse, only younger, because conversions arrive days after the click. I show how to measure your own filling-up curve, what to multiply a young day by, when I refuse to estimate at all, and how long a reporting window has to be.
Search terms · 16 min read · Interactive · Video
Which search terms should I block, and which deserve a chance? The Wilson score interval
It is review day, and a dozen search terms have spent real money without a single sale. Blocking them feels like discipline, but for most it is a coin toss. I show how many clicks a term needs before "no conversions" means anything, the free relevance check that comes first, and how to run both on your own report.
Interrupted time series · 13 min read · Interactive · Video
I changed something. Did it actually work? An interrupted time series
My before-and-after average said the change had failed, and it had the direction backwards, because cost per lead was already climbing. I show you how to project the line your metric was already on, how many days you can honestly read it, and how to check that a quiet week could not fake the same result.
Marketing mix modelling · 15 min read · Interactive · Video
How much should I spend on each channel, and where do the extra euros stop paying off? Marketing mix modelling
A table of cost per sale says feed Social and cut Video. On my example account that move is exactly backwards. I show how a marketing mix model reads the return on the next euro for every channel from weekly spend and revenue, how sure you can be of it, and the ways it misleads you.
Experiments · 14 min read · Interactive · Video
Is this channel actually driving sales, or would those sales have happened anyway? A geo-holdout read by difference in differences
I turned a channel up and sales rose, then a finance lead asked whether the shop would have sold those anyway. This is how I find out: switch the channel off in randomly chosen regions, compare them with the rest, and read the lift as a range, with a notebook that runs it on your own data.
Attribution · 16 min read · Interactive · Video
Which marketing channels actually drive sales, and which just show up at the close? The Markov chain removal effect
Last-click hands every sale to the final touch and starves the channels that start journeys. I score each channel by how many sales would vanish if it disappeared, give its credit as a range, and show where that method misleads and how to run it on your own journeys.