Blog
Field notes on tracking.
Plain, practical writing on server-side tracking, first-party data, consent and the data underneath good marketing. No fluff, no theory for its own sake.
More field notes
Why your Shopify purchases stopped showing in Google Ads, GA4 or Tag Assistant
Orders did not vanish. The page they fire from got sandboxed, and the deadline to move off it just passed.
Stape vs Taggrs vs Addingwell: the European server-side hosting decision
Three vendors, one identical server container underneath. Here is what actually differs, and what does not.
Taggrs vs Addingwell for agencies
Two European server-side hosts, one identical container underneath. Here is what actually separates Taggrs from Addingwell.
Client-side vs server-side tracking: what changes, what does not
The real difference is cookie lifetime and request destination, not consent, ad blockers or deduplication rules.
Google tag gateway vs server-side Google Tag Manager: which one do you need?
Two Google products with confusingly similar names solve different problems, and neither is where a missing-signal problem actually starts.
Shopify tracking in 2026: the four ways to measure a store and which one to pick
Four ways to send Shopify tracking data, why inherited stores often run two at once, and which one an agency should reach for first.
The Shopify pixel sandbox: what it can measure and what it cannot
Shopify's pixel sandbox is not a configuration gap. It is a deliberate wall, and here is exactly what sits on the other side of it.
POAS on Shopify: the data you need and how profit reaches Google Ads and Meta
Shopify never sends profit to Google Ads or Meta on its own, and a refund often lands after the bidding algorithm used the wrong number.
Automatic user-data collection in Google Ads, Google Tag Manager and Meta: what the setting really does
Turn on automatic data collection and the tag scans your whole page, not just the form, in two very different failure modes.
What does a tracking specialist cost? Rates, retainers and project prices in the Netherlands
Two 2026 Dutch freelance rate surveys, and the budgeting framework that actually separates a fair quote from a padded one.
Tracking specialist job description: what to put in it and what to leave out
The buzzwords that fill a tracking specialist vacancy, and the copyable job description block that actually filters for competence.
Consent Mode v2 basic vs advanced: what each one measures and what it costs you
Consent Mode v2 has two modes, and the real difference is not compliance, it is how much a denied visitor still tells Google.
Enhanced conversions and Consent Mode v2: what is sent when, and what you may send
Two consent toggles sound the same and gate completely different things. Which one decides what enhanced conversions can send.
Google Tag Manager basics: how it works and what agencies use it for
Installing Google Tag Manager does not measure anything by itself. What the container actually does, and where agencies hand it off.
Stop asking your marketers to build the tracking
Three documented Google Tag Manager failures show why tracking needs an owner, not a marketer's side task.
Automatic advanced matching on Meta: when to turn it on, and when it quietly fails
Meta's automatic advanced matching only works on a form it can actually see. Where it can't, turning the setting on changes nothing.
What UTM naming convention should your email flows use?
Welcome, cart and winback flows all point at the same site. Without one convention, GA4 can't tell them apart.
Klaviyo and Google consent mode: what actually gates the tracking cookie
Consent mode gates Google's own tags automatically. The Klaviyo tag in the same container needs its own setting, added by hand.
HubSpot UTM tracking: what your CRM actually sees on the website
Your CRM traffic-source report looks authoritative. Here is what its tracking cookie never had the chance to see.
Taggrs, tested from the configuration side
A configuration-side look at Taggrs, not a feature list: what the free tier gets right, and where the real work still sits.
Stape, tested from the configuration side
A configuration-side Stape review: what its tiers actually include, where the claims check out, and where they don't.
Taggrs alternatives for agencies that want EU hosting
Filtered to hosts that actually prove EU data residency, and the one thing switching vendors never fixes on its own.
How Stape, Taggrs and Addingwell count requests, and how to estimate your bill
Stape, Taggrs and Addingwell each define a billable request differently, so three quotes for the same site rarely mean the same thing.
Shopify's Google and YouTube app or Google Tag Manager: which one should send your data?
Two paths reach Google Ads from Shopify, the app and a GTM pixel. Why running both doubles a conversion, and how to pick one.
WooCommerce and Google Tag Manager: the data layer, GA4, and where a purchase gets counted twice
WooCommerce needs a plugin to build a data layer, and running two at once double counts the same sale. Where that happens, and how to check.
Webflow conversion tracking: forms, GA4, Google Ads, and the two ways it double counts
A Webflow form doesn't reload the page, and that one detail is why native GA4 and GTM can both end up claiming the same lead.
What does a tracking audit actually cover?
A quick tag test and a real audit both get called an audit, but they find very different things. Here is what a proper one covers.
Does your tracking hold up when your client advertises in five markets?
One banner, five markets, five different numbers. The rule your setup has to satisfy changes at the border, and so does the data.
Why do your Looker Studio dashboards keep breaking?
Quota errors and numbers that disagree are not chart problems. They are data source problems, and they share one fix.
Why does client reporting still take days every month?
The slow part of monthly reporting is not the copying. It is deciding which of your three conversion numbers the client sees.
Are you using half of what your data setup can do?
Nobody switched these features off. They were never switched on, and the data to feed them has been sitting there for months.
Meta conversions dropped suddenly, what happened?
Spend held. The client's own order count held. Meta fell anyway. That is a measurement break wearing a performance break's clothes.
Why GA4 and Shopify never show the same numbers
Shopify is not right and GA4 is not broken. They are two ledgers, and only one of them decides how the client's budget gets bid.
Where does GA4 "(not set)" traffic actually come from?
Four different empty fields hide behind one label in GA4, and three lookalike values get blamed for it. Telling them apart is the fix.
Did your new cookie banner just kill your conversions?
The drop has a date on it and the banner is the obvious suspect. Reading the shape of the drop tells you what actually broke.
How do you know your tracking broke before the client does?
A tracking break makes no error, only a smaller number. Why silent data loss reaches the client report before it reaches yours.
How do you get client data ready for AI?
Cleaning the export is the last step, not the first. The five things that change in a client dataset, in the order that works.
Your client asks if the tracking is compliant. What do you say?
The honest answer is not a yes or a no. It is a description of what your setup does, and most agencies cannot produce one.
Is there personal data hiding in your analytics?
An email address in a URL, a phone number in a site search term. How personal data reaches analytics, and what it costs you.
Is that conversion drop seasonality or a real problem?
A dip can be August, a broken tag, or a report that is not finished yet. The order you check in decides how fast you know.
Does server-side tagging slow down your site?
The Core Web Vitals objection, answered honestly: what really leaves the browser, what stays, and when a container costs you speed.
Why every attribution model tells a different story
Same property, same week, three answers. What model disagreement really tells you, and which part of it is missing data.
How does server-side tracking actually work?
One conversion traced from the browser to the ad platform, and the three hops that decide whether a migration pays off.
Why your data breaks when checkout lives on another domain
The checkout domain shows up as a traffic source, sessions double, and Google Ads disagrees with the order table. Here is why.
Why Google Ads counts the same conversion twice
More conversions than the client has orders. Here are the four things that make Google Ads count one event twice.
How do you brief a dev team for tracking work?
Vague tracking requests come back as another sprint. What a data layer brief has to pin down before a developer touches it.
Does marketing mix modelling fix your measurement problem?
MMM is pitched as the answer to signal loss. What it actually fixes, what it quietly inherits from your data, and where it leaves a gap.
How do you get ad spend and CRM revenue into one report?
Cost sits in the ad platforms, revenue sits in the CRM, and campaign name is not a join key. What it takes to put them in one row.
Can AI write your client reports?
AI drafts the commentary in seconds. What it cannot do is notice that the number it is explaining was never right.
When has your agency outgrown GA4?
Sampling, 14-month retention and (other) rows. Three signs an account has outgrown GA4, and what to move to instead.
Why does Google Ads show more conversions than GA4?
Four documented counting rules explain most of the gap. What is left after that is not a reporting quirk, it is missing data.
Your CPA is rising and nothing changed, except your data
Cost is billed and complete. Conversions are collected and incomplete. That asymmetry raises a CPA no campaign change explains.
Why Safari and iOS quietly eat your conversions
Your totals look survivable because the loss is not spread evenly. It sits on Safari and iOS, where the browser deletes the evidence.
Every tool shows a different number, which one is lying?
Ads, Meta and GA4 will never agree, and that is fine. How to tell a counting difference from a real hole in your data.
How do you tell a client their data has been wrong?
The conversation agencies dread, in order: what to say first, what to put on screen, and why the correction usually runs in your favour.
Can you charge clients for tracking work?
Why tracking hours end up unbilled, and the three invoice lines that turn the work into something a client accepts.
Your marketers set up the tracking. Do they know what they built?
An unowned tracking setup gets worse without anyone touching it. What owning the measurement layer actually involves.
Consent Mode v2 without throwing data away
Compliant and measurable is not a trade-off. The four decisions inside Consent Mode v2 that decide which one you get.
What GA4 sampling hides from your client reports
Sampling, (other) rows, thresholding and retention each bend a client report differently. How to tell which one you are looking at.
Why enhanced conversions and CAPI underdeliver on a weak setup
You followed both setup guides and the numbers barely moved. The layer underneath decides what either feature can give back.
Your ROAS looks great, so why is the business not growing?
Platform ROAS can climb while client revenue stands still. Four documented reasons why, and the number that would have said so sooner.
Which questions should you ask before hiring a tracking specialist?
Every tracking vendor says server-side, adblocker-proof and compliant. Five questions that tell you which one means it.
How do you get more visitors to accept your cookie banner?
The accept rate on your client's cookie banner scales every number you report, and it is the one metric nobody at the agency owns.
Modeled or measured: which conversions are you reporting?
The conversions in your client report are part counted and part estimated, and no column tells you where the line sits.
Cookiebot, Usercentrics or custom: does your CMP choice matter?
Two of the three options share one owner, and none of them decides your consent rate. Here is what the vendor choice really settles.
Can ChatGPT actually analyze your marketing data?
It runs real code on your file, so the arithmetic is right. What a GA4 export left out before the upload is the actual limit.
Why your tracking breaks every time the website changes
Every release quietly moves something your tracking depends on. Why nothing errors, and what actually catches it in time.
What happens to your tracking during a website migration?
Your redirect map protects rankings. It does nothing for visitor identity, ad click parameters or the before-and-after comparison.
You inherited a broken tracking setup. Now what?
Taking over an account nobody documented, without pretending you can repair the numbers the client has already seen.
What to check in the tracking before you onboard a new client
The logins are the easy part. The question that decides whether a client's measurement survives the handover is who owns it.
POAS or ROAS: which one should you report to clients?
ROAS and POAS divide by the same number, and it is the wrong one. How deep a profit figure has to go before it changes a decision.
Does server-side tracking make sense for lead generation?
Every server-side post is written for a webshop. What actually changes when the conversion is a form fill, and where the case falls apart.
Your ads optimize on form fills while the revenue sits in the CRM
Lead volume up, revenue flat. What has to be true before closed deal value can reach the ad platform, and when it cannot.
Your campaigns optimize for form fills, not revenue
More leads at a lower cost per lead, and the same revenue. The optimization target is usually the reason, not the campaign.
Why your form conversions never match the leads in the CRM
The two counts measure different objects, after different filters, in different time zones. Which one to trust, and how to shrink the rest.
Are bots inflating the conversions you optimize on?
Invalid click filtering does not cover the conversion that click triggered. What that does to your bidding, and where to filter instead.
Your tracking is first-party and Safari still deletes the cookie
You went first-party, the container sets the cookie server-side, and Safari still caps it at seven days. The reason is DNS.
Which parts of your measurement stack should you own?
Build versus buy is the wrong question. Sort your measurement stack into layers, then decide which ones the client has to keep.
How do you keep tracking consistent across thirty client accounts?
Thirty accounts, thirty dialects. Keeping tracking consistent is a specification problem, not a dashboard problem.
Why 15 to 30% of your conversions never reach GA4
GA4 and Google Ads never quite agree for a reason. Here is where the missing conversions actually go, and how to get them back.
First-party cookies and the 400-day difference
Two identical first-party cookies, one lasts a week, one lasts a year. The difference quietly shortens every client's attribution window.
In-house data hire or external specialist: what does each really cost?
You cannot justify a full-time data hire, but the missing conversions keep costing you. Here is the real maths on each option.
Is server-side tracking worth it for your agency?
Server-side sounds expensive and the payback feels vague. Here is the honest business case, and when it is actually worth doing.
Why your tracking alerts fire on noise and miss real problems
Threshold alerts fire on a quiet Sunday and stay silent when one conversion action dies. The fix is not a tighter number.
What does server-side tracking really cost?
Hosting is the cheap part. Here is the honest breakdown of what a server-side setup costs to own, and where the money really goes.
Stape vs Taggrs: does the server-side tool actually matter?
Both host the same Google server container. Here is the decision that actually changes how many conversions you recover.
How much data is your cookie banner throwing away?
Some of the loss you agreed to, the rest your banner configuration chose for you. How to tell them apart, and recover the second half.
Why AI gives wrong answers about your marketing data
The model is rarely the problem. Three things in your data decide whether an AI answer holds up on a client call.
How does white-label tracking work for agencies?
Clients want server-side, consent-safe measurement. Here is how to offer it under your own brand without hiring a specialist.
What a first-party tag gateway actually does
Everyone says go first-party, few explain what it changes. Here is what a tag gateway actually does, in plain terms.