Server-side GTM is usually sold on technical grounds. Better data quality. Reduced client-side load. Privacy compliance. All true, all important, and all completely abstract to the person signing off the budget.
Here's what they actually want to know: what does it do for my numbers?
We can answer that. When we deployed server-side GTM for Unily, they recovered 70% of previously lost session data. That's not a rounding error. That's a step change in what you can see, what you can attribute, and what you can optimise.
Let's unpack what that actually means for ad performance.
Where the data goes missing
Before we get to what you gain, it's worth understanding what you're losing right now. Browser-side tracking loses data in three ways:
Ad blockers. Roughly 30-40% of desktop users run some form of ad blocker. Many of these also block analytics scripts. Your GA4 tag simply doesn't fire for these users. They visit, they browse, they convert - and your analytics sees none of it.
Intelligent Tracking Prevention (ITP). Safari and Firefox aggressively limit cookie lifetimes. First-party cookies set by JavaScript are capped at 7 days (sometimes 24 hours). A user who visits on Monday and converts on Thursday may show up as two separate sessions - or the conversion may not be attributed to the original source at all.
Consent rejection. Under GDPR and ePrivacy, users who reject analytics cookies generate no data at all. In markets with strong consent enforcement, this can be 30-50% of traffic. These users still cost you ad spend. They still convert. You just can't see them.
The cumulative effect is that your analytics is working with a fraction of reality. Not a representative sample - a biased one. Users who accept cookies, don't use ad blockers, and visit frequently enough to stay within ITP windows are over-represented. Everyone else is invisible.
What happens when you recover that data
Server-side GTM shifts tag execution from the browser to your own server infrastructure. The practical effect:
Cookies become first-party and server-set. ITP doesn't apply to server-set cookies the same way. Cookie lifetimes extend from days to months. Returning users are recognised correctly. Attribution windows actually work.
Ad blockers can't block what they can't see. Because the data flows server-to-server (your server to Google's, or Meta's, or whoever's), browser-based blockers don't intercept it. You're not circumventing consent - users who reject cookies are still excluded. But users who accept cookies and happen to run ad blockers are no longer invisible.
Consent Mode works properly. With server-side GTM, you can implement Consent Mode v2 with full fidelity. Consented users generate complete data. Unconsented users still contribute to behavioural modelling without storing personal data. The signal improves for everyone.
Here's where it compounds. Ad platforms - Google Ads, Meta, TikTok - rely on conversion data to optimise bidding. The more conversions you feed back, the better the algorithm performs. This creates a cascade:
- More observed conversions → the algorithm has more signal to learn from
- Better attribution → conversions are attributed to the right campaigns and keywords
- Smarter bidding → automated strategies (tROAS, tCPA) make better decisions
- Lower cost per acquisition → you stop overspending on channels that look good because alternatives are under-measured
- Better audience building → remarketing lists and lookalike audiences are built from a more complete picture
Unily's 70% data recovery didn't just fix their dashboards. It restored executive confidence in reporting - which sounds soft until you realise that "we don't trust the numbers" is usually followed by "so we'll just keep doing what we've always done."
Let's make this concrete. Suppose you spend £100k/month on paid media and your current analytics captures 60% of conversions (a conservative estimate given ad blocker and ITP losses).
Your observed CPA is £50. Your actual CPA is probably closer to £30 - but you can't prove it, and neither can the algorithm.
After sGTM deployment, you recover most of that missing signal. The algorithm now sees the real conversion volume. Three things happen:
- Campaigns that were being over-credited lose budget to campaigns that were being under-credited. This reallocation alone can improve ROAS by 10-20%.
- Automated bidding strategies converge faster because they have more signal. Less wasted spend during learning periods.
- Remarketing audiences grow, because you're now building them from a more complete user base.
The exact numbers vary by vertical, consent rates, and ad blocker penetration. But the direction is always the same: more data in, better decisions out.
How we do it
We use our UPDATE framework to migrate clients to server-side GTM as part of our tracking solutions. Six steps:
- Understand - audit your current setup, quantify data loss, identify what you'd recover
- Plan - design the server-side architecture (Cloud Run, load balancing, failover)
- Deploy - spin up the server container, configure clients, tags, triggers and variables
- Audit - validate data parity between client-side and server-side
- Transition - shift live traffic, monitor, resolve discrepancies
- Evaluate - measure the uplift in data volume, attribution accuracy, and ad platform performance
The whole process typically takes 4-8 weeks depending on complexity. The infrastructure cost is modest - Cloud Run pricing scales with traffic, and for most sites it's a few hundred pounds per month.
The question isn't whether - it's when
Every month you run browser-side-only tracking, you're making decisions on incomplete data. The gap isn't getting smaller. Safari's restrictions are tightening. Chrome's Privacy Sandbox is evolving. Ad blocker adoption is growing.
Server-side GTM isn't a nice-to-have for next year's roadmap. It's a data quality issue that's costing you money right now.
If you want to know what you'd recover, talk to us about an sGTM readiness assessment. We'll quantify the gap and show you what comes back.
Clean data in is the foundation for everything else. Once your tracking is solid, the next step is unifying it in an intelligence platform - and from there, building the governed agents and data products that turn data into decisions.
