The Ecomm Analyst

The first-touch versus last-touch debate eats more operator hours than it deserves. People treat it like a deep methodological choice with a right answer, when for most stores it is closer to picking which end of the customer journey you want to flatter. Both models are simplifications. Both are wrong in predictable ways. Knowing exactly how each one is wrong is worth more than picking a side. Last-touch attribution gives all the credit for a sale to the final click before purchase. Someone discovers you through a Meta ad, thinks about it for a week, then searches your brand name…

Ask ten operators what their ROAS is and you will get ten confident answers and very little agreement about what the word means. Some are reading the number Meta reports. Some mean blended return across everything. Some mean a target they were told to hit by an agency three years ago. The confusion matters because the metric you steer by determines the decisions you make, and the two most common choices, platform ROAS and MER, pull in different directions. Platform-reported ROAS is revenue a channel claims it generated divided by what you spent on that channel, as measured by the…

For a few years now every attribution conversation has started with the same eulogy. The pixel got worse, iOS made opt-out the default, third-party cookies finally went, and the clean deterministic tracking everyone built their reporting on stopped being clean or deterministic. That part is true. What gets lost is that the eulogy is for one specific method, user-level click tracking, not for measurement itself. Plenty still works in 2026. It just works differently, and the operators who adapted are making better decisions than the ones still refreshing a Meta dashboard hoping the numbers come back. Start with what actually…

Most attribution problems are not modeling problems. They are plumbing problems. Before you argue about first-touch versus last-touch or whether to buy a measurement tool, it is worth spending one afternoon confirming that the data feeding every report is actually correct. Here is the pass I run when I sit down with a store whose numbers feel off. Start by writing down your sources of truth. Open Shopify and note total revenue for the last 30 days, the version before refunds and the version after. Open each ad platform and note spend and platform-reported revenue for the same window. Do…

You open Shopify and it says you did $84,000 last week. You open Meta and it claims $61,000 from its campaigns. Google says $28,000. Add the platforms up and they’ve sold more than your entire store did. Nobody is lying and nothing is broken. The numbers disagree because each system is answering a different question, and once you see which question each one is answering the gap stops being mysterious. Start with Shopify. Shopify counts an order when the money lands. It does not care where the customer came from or what they clicked three days ago. One order, one…

I look at a lot of attribution setups on stores in the one to ten million range, and the same handful of mistakes come up again and again. None of them require a data team to fix. Most are about discipline and mindset more than tooling. Treating platform ROAS as truth The most common one is reading the number Meta or TikTok reports as if it were real revenue. It is the platform grading its own homework on its own attribution window, and it is almost always inflated. Use it to compare campaigns inside one account, not to decide what…

Cohort curves are one of the most seductive charts in e-commerce analytics, and one of the easiest to over-read. At small scale especially, people pull a cohort report, see a line bending nicely upward, and build a CAC ceiling on top of it that the data cannot actually support. Here is how I try to read these without fooling myself. What a cohort curve actually shows A cohort curve groups customers by when they first bought, usually by month, then tracks cumulative revenue per customer as each group ages. The January cohort, the February cohort, and so on, each followed…

Three numbers get used interchangeably in operator conversations, and treating them as the same thing leads to some genuinely bad decisions. Platform ROAS, blended ROAS, and MER measure different things, answer different questions, and break in different ways. Here is how I keep them straight. Platform ROAS This is the number Meta, TikTok, or Google reports inside the ad account. It is the platform grading its own homework. It uses the platform attribution window and its own definition of a conversion, and it has no idea the other platforms exist, so when several channels all take credit for the same…

Most brands under a few million in revenue do not have anyone whose job is measurement. The founder is doing it at night, or a generalist marketer is squeezing it in between launching campaigns and answering support tickets. That constraint should shape the entire framework. If your measurement setup cannot be run in about an hour a week by someone who has four other jobs, it will quietly stop being run, and then you are flying blind with a very expensive dashboard nobody opens. The first mistake I see is copying enterprise. People read how a brand with a five-person…

Decile is a customer intelligence platform for mid-market e-commerce brands. It starts from your actual transaction data, enriches it with hundreds of demographic and psychographic attributes, and turns that into personas, predictive LTV and lifecycle segments, and audiences you can sync straight to Klaviyo, Attentive, and Meta. There is an AI analyst on top that answers customer questions without a data team. The positioning is explicitly enterprise CDP capability without the enterprise CDP price, and the pricing reflects that, with a Growth plan around $900 a month for brands up to $20M in revenue and a Scale plan around $1,500…