Field notes
Measurement, identity and attribution, explained properly.
20 pieces on how first-party tracking, identity resolution and server-side conversions actually work - what breaks, why, and what to do about it. Every one ends with something you can go and check on your own store.
What a CDP actually is, and why the one you have probably is not one
A customer data platform is not a warehouse with a nice interface. It is the decision to treat a person as one person - and almost nothing in a normal stack does that.
Read itCDP
4 piecesWhat a CDP actually is, and why the one you have probably is not one
A customer data platform is not a warehouse with a nice interface. It is the decision to treat a person as one person - and almost nothing in a normal stack does that.
6 min readTwo ways to run Cookiebees: stream everything, or enrich the server-side setup you already paid for
You do not have to rip out Stape or your existing server-side GTM container. There is a path that keeps your tags exactly where they are and just makes the data flowing through them better.
7 min readAbandoned checkout recovery that knows who it is talking to
Every store sends abandoned cart emails. The difference between one that works and one that annoys is whether it knows anything about the person receiving it.
5 min readReactivating lapsed customers with their own history rather than a bulk blast
The list of people who bought once and never returned is usually the largest addressable audience a store owns, and the one it treats worst.
5 min readTracking
3 piecesYour "first-party" tracking is still third-party to Safari, and it is being capped
Setting a cookie from JavaScript on your own domain feels first-party. Safari disagrees, and the seven-day cap it applies is quietly eating your attribution.
6 min readFirst-party tracking on a path: how to stop losing customers to the seven-day cap
Not a subdomain, not a CNAME - a path on your own root domain. It is the only arrangement Safari treats as genuinely yours, and it is simpler to set up than it sounds.
7 min readHow the visitor id survives checkout, and why so many orders arrive anonymous
Checkout is where attribution goes to die: a different domain, a locked page, sometimes an entirely different app. Cart attributes are the copy that makes it through.
5 min readAttribution
3 piecesAttribution coverage: the number that decides whether any of your other numbers are true
Before you compare channels, ask what share of orders you can attribute at all. Everything else is a ratio of that.
5 min readMatch quality: why the same sale is worth more to your ad platform with better data
Sending a conversion is not the same as sending a conversion that can be matched to a person. The second one is what changes bidding.
6 min readShadow mode: see every conversion you would have sent, before you send one
The riskiest moment in any measurement project is the first time real data leaves for an ad platform. It should not be the first time you look at it.
4 min readSignals
4 piecesSend a conversion event per brand, and stop letting your best margin subsidise your worst
If you stock more than one brand, a single Purchase event tells the ad platform every sale is worth the same. It is not, and bidding will happily learn the wrong lesson.
5 min readCategory-level conversion events, and the seasonal buying you cannot currently bid on
Categories behave differently: different margins, different repeat rates, different seasons. One blended Purchase event flattens all of it.
5 min readProduct-level events: for the handful of SKUs that actually run your business
Most stores have a few products that bring in customers and a long tail that does not. Product-level events let you bid for the first group specifically.
5 min readYour email opens and WhatsApp replies are conversion signals you are throwing away
You have been generating high-intent engagement data for years. It dies inside your inbox instead of teaching your ad platforms who actually engages.
5 min readJourneys
6 piecesThe complete journey behind a sale: not a model, a receipt
Attribution models argue about how to divide credit. Before you have that argument, it is worth being able to simply read what happened.
6 min readChannel journey: the sequence of channels behind an order, not just the last one
Paid social finds them, search catches them, email closes them. Last-click credits one of the three and quietly tells you to stop funding the other two.
5 min readThe funnel behind a single order: viewed, carted, checked out, bought
Aggregate funnels tell you the drop-off rate. Per-order funnels tell you what the drop-off actually looked like for the people who came back.
5 min readWhat they looked at before they bought something else
The products a customer considered and rejected are some of the most useful data you own, and almost nobody stores them against a person.
5 min readTime to purchase: your retargeting window is probably wrong
Most stores pick a retargeting window out of habit. Your own data has the answer, and it is rarely the number you guessed.
5 min readTouches to purchase: how many times you have to be seen before it works
Frequency capping and budget decisions both depend on this number, and most stores are working from an industry rule of thumb rather than their own.
5 min readClose the loop on your next lead
Transform your brand with signals.Your first order, credited, in ten minutes.
Measurement first, then everything else. Spin up your workspace in minutes and watch the first signal go back on your very first order. 30-day free trial, no card required.