Pinned Tweet
Hi 👋 I'm David. I'll be writing here about digital marketing, AI, agentic analytics and running ad campaigns with AI.
Mostly lessons from the field, few hot takes, zero hype.
Legitimate interest is not a back door for tracking.
On November 10, the @EU_EDPB adopted the final version of its guidelines on Article 6(1)(f) GDPR. Not a new law: how EU regulators read the one that already applies.
What changes for marketing 🧵
2
1
65
So a generic legitimate interest assessment won't carry ad tracking. Consent stays the basis that matters.
The question becomes whether your tracking respects it. Every tag that fires before consent, or after a rejection, gets harder to explain.
1
6
The EDPB guidelines, version 2.0:
edpb.europa.eu/documents/gui…
How @trackingplanco checks consent on every hit:
trackingplan.com/landings/pr…
6
Ask an agent about a tool it can't read, and plenty will answer anyway.
Ask @trackingplanco about a destination it doesn't receive, and it says so in one line, before anything else.
"I can't see that" is part of answering right.
trackingplan.com/changelog/c…
11
"Not being able to detect when something had broken, sometimes even weeks later."
That was On Top Media before automating analytics QA. Now a lost metric shows up the next day, and they save 25 to 30 hours a month.
trackingplan.com/customer-st…
10
Two @GoogleAds changes this fall say the same thing: control is moving from campaign settings to the data you feed the machine 🧵
1
1
16
Fewer knobs to turn means the inputs matter more: conversion values, consent signals, the events you mark as conversions.
If those are wrong, AI Max just finds more of the wrong thing, faster.
1
1
Server-side doesn't save you either. Consent has to travel with the conversion:
trackingplan.com/guides/cons…
Google's post: blog.google/products/ads-com…
1
8
A few months ago, people we know in analytics and data started worrying about the job market.
So with our friends at @TheirStackApp we launched TrackingJobs: analytics and data jobs, every Monday, by role and country, salary shown when shared. Free.
trackingjobs.com/
Your ad platform counts clicks. Your analytics counts visits. The gap between them is where tracking breaks.
@trackingplanco now reads spend, clicks and tagging from 11 ad platforms and follows each click to your pixels.
trackingplan.com/changelog/s…
4
72
Who's responsible when an AI feature you didn't build changes your ad?
@Meta's revised commercial terms tried to answer that. They're on hold now, but worth reading 🧵
1
28
Meanwhile, advertisers were already pulling Meta ads that came back with unwanted AI "glow-ups".
That's the real lesson: automated creative needs the same QA as automated tracking. If you didn't check what went live, you don't know what you're paying for.
1
15
Whatever the final terms say, the operating rule is the same.
Automate the work. Keep the checks.
What do you review before an AI-edited ad goes live?
Adweek's report: adweek.com/media/advertisers…
13
Cinco de cinco 🎉
🌊 La Quinta, la gran fiesta de los datos · 📍 A Coruña, 23 y 24 de abril · 🍽️ con cena
¿Te subes a la ola? mi.datola.es/v/WyJEYXZpZCBQb…
@datolaes #Datolada
A data agent always answers. That's its strength and its risk.
It will return a confident number from a query that counted the wrong thing.
These are the six errors agents make most often on GA4 data in BigQuery, and how to catch each one 🧵
1
38
6. Hiding missing values in one big unlabelled bucket.
(not set) and NULL end up in a row nobody reads. If that row is large, it's telling you something about your tracking.
1
5
How to catch them: open the SQL behind the answer and check the table, the dates, the filters, what is counted, and whether a join or UNNEST duplicates rows.
Then compare with a number you already trust.
The full guide: trackingplan.com/guides/chec…
5


