Services consulting
Analytics setup where the numbers can be trusted
Marketing analytics consulting means building a measurement layer you can make decisions from — correct tracking, honest attribution, consent handled properly — so every other marketing investment can be judged rather than argued about.
Why does analytics quality determine everything else?
Every marketing investment decision depends on a measurement you cannot verify. If conversions are double-counted, if phone calls are invisible, if consent mode has silently thinned your European data until month-to-month comparison is meaningless, then no amount of marketing expertise produces a reliable answer.
This is why we insist on fixing measurement before tactics, in every engagement, even when it was not what the client came for. It is unglamorous, it produces no launch announcement, and it is the highest-leverage work available. The most expensive marketing decisions are usually made on data that was wrong, and the cost of finding out afterwards is frequently larger than the cost of getting it right.
What are GA4's real limitations?
GA4 is a good free analytics tool that is frequently misused and occasionally mis-sold. Its limitations are structural rather than bugs, and knowing them prevents a lot of bad decisions.
| Limitation | What it means in practice | What to do about it |
|---|---|---|
| No pageviews in the admin | Page views appear in reports but not in realtime or explorations as before | Use explorations and the events report; it is a change of habit, not a defect |
| Event data retention | Standard is 14 months; beyond that, events are lost unless configured otherwise | Export to BigQuery or a warehouse if you need long-term event-level history |
| Attribution windows and modeling | Multi-touch attribution is modelled, not measured, and does not reflect reality cleanly | Treat platform attribution as one directional input, not ground truth |
| Ad blockers and ITP | Significant shares of traffic never execute your tag at all | Server-side tagging, or accept the loss and account for it |
| Consent mode thins data | Cookieless pings are modelled and behavior differs from cookieless consent | Document the modelled share; never compare raw figures across the consent change |
| Ad blockers affect server-side too | Server-side helps only if requests are not blocked upstream | Use a first-party subdomain and your own DNS |
| Cross-domain complexity | Tracking across related domains needs careful configuration | Set up link decoration and cross-domain measurement deliberately, or self-referrals will corrupt the data |
What is the measurement layer?
Analytics has five layers, and problems are usually diagnosed at one and attributed to another. Knowing which layer has failed saves an enormous amount of time.
Collection
How data travels — client-side to Google, or through your own server. Vulnerable to blockers, latency and sampling.Processing
Deduplication, consent application, identity resolution, modeling. Where most 'the numbers look wrong' problems actually live.Reporting
Dashboards, explorations and alerts. A correct model can still produce useless output if nobody reads it.Decision
Whether anyone changes what they do based on it. The layer most often missing, and the one that determines whether analytics is worth its cost.What should you actually track?
Track the minimum set that supports decisions you actually make, plus the technically necessary events. Teams routinely instrument far more than they use, which adds collection cost and complexity for nothing.
Below are the recommended event sets by business model. The distinguishing question at the start is: what decision would this number change?
| Business model | Essential events | The metric that matters |
|---|---|---|
| Content / media | Scroll depth, engaged session, newsletter sign-up, outbound click | Subscriber growth and assisted conversions |
| Lead generation (B2B) | Form submit, demo booking, content download, scroll to CTA | Qualified pipeline, not raw form fills |
| SaaS / product | Signup, activation event, trial start, invite sent, upgrade | Trial-to-paid and activation rate |
| Ecommerce | View item, add to cart, checkout start, purchase, refund | Contribution margin per session and repeat rate |
| Local service | Call click, direction request, GBP action, form submit | Booked jobs, tracked to source |
| Professional services | Contact, consultation booking, document download | Consultations booked and their value |
When does server-side tagging justify its cost?
Server-side tagging routes analytics through your own server, which makes it far harder for blockers to intercept and gives you control over data before it reaches third parties. That control is genuinely valuable for privacy-sensitive businesses.
It is not free, and it is not always worth it. Beyond the platform cost you take on operational responsibility: a first-party subdomain, DNS and TLS management, monitors for uptime, and a data flow you are now responsible for under GDPR.
| Situation | Server-side worth it? | Why |
|---|---|---|
| Significant share of traffic blocked by blockers | Yes | It is the only reliable fix for client-side collection loss |
| EU-focused business with strict data handling requirements | Likely | Gives genuine control over what is sent and where |
| Sensitive category — health, financial, children's data | Yes | Reduces the data footprint considerably |
| High-value conversions where a 5% tracking loss matters financially | Possibly | Depends on whether the lost conversions are worth the operational overhead |
| Small site, low traffic, limited technical resource | No | Operational cost exceeds the benefit; first-party domain configuration matters more |
| Already losing data for another reason | No | Fix the actual cause first — it is usually not blockers |
How do consent mode and GDPR affect this?
Google Consent Mode v2 requires specific parameters to be declared before tags fire, and it changes what you can measure rather than simply whether you measure it. With consent denied, you receive modelled data — cookieless pings that Google estimates rather than observes.
This produces a specific and manageable problem: your consented and non-consented traffic are different populations, and blending them distorts every rate you calculate. Conversion rates, engagement rates and averages all change if modelled sessions are mixed with observed ones without separation.
Separately: this website sets no cookies and runs no analytics, which is why the only data we hold is what you type into the contact form. That is a deliberate position rather than an oversight.
Related services
Frequently asked questions
Do we need Google Analytics 4 or should we use something else?
GA4 is sufficient for most businesses and is free, so the burden of proof sits with the alternative. Plausible, Fathom and similar tools are genuinely better for simple traffic reporting and are worth considering if you need privacy and simplicity more than marketing attribution. What is not worth doing is paying for a premium tool while leaving the underlying tracking misconfigured — better data collection beats better reporting every time.
Why is my conversion rate different in GA4 than in my ad platform?
Three usual causes. Your ad platform is counting a wider definition of conversion than you expect. Cross-domain or cross-device journeys are not being stitched together. Or consent mode has thinned your data unevenly. Note that a discrepancy between platforms is normal and not necessarily an error — but a large one is a reason to reconcile rather than average.
Is server-side tagging worth it for a small business?
Usually not, on cost-to-benefit grounds. It requires a first-party subdomain, DNS and TLS management, and uptime monitoring — real operational work. For a small site the benefit usually does not justify it. If blockers are genuinely removing a large share of your traffic, that is a strong argument; otherwise fix attribution and consent configuration first, which is cheaper and more often the actual problem.
What is consent mode v2 and do we need it?
Consent Mode v2 is Google's framework for handling consent signals in analytics and advertising, with additional required parameters beyond v1. If you run Google tags in the European Economic Area, you need it to be compliant. It changes what you can measure rather than whether you can measure — denied consent produces modelled pings rather than nothing, so your data becomes a mixture of observed and estimated sessions that must be handled carefully.
How many events should we track?
Fewer than you think. Track the minimum that supports decisions you actually make, plus the technically necessary events. Excessive instrumentation adds collection cost and cognitive load without improving decisions. Start from the question "what would this number change?" — if the answer is nothing, do not track it.
Can analytics tell us the return on our advertising spend?
It can estimate it, and the estimate is limited by several factors: ad blockers removing a share of sessions, consent mode modeling, privacy-preserving browser changes, and cross-device journeys that cannot be observed. If phone calls or in-person sales are your main revenue path, analytics cannot see them at all unless you implement offline conversion import. Report ROM as a modelled estimate, clearly labeled, and never as a measured fact.
How do we track phone calls as conversions?
With call tracking — dynamic number insertion, which shows a different tracking number in ads and on the website, and attributes calls to source. It works well but needs care: exclude staff numbers, handle number reuse, and make sure the reported attribution is directional rather than precise. For local businesses this is often the most important conversion you are currently missing entirely.
Our numbers changed and we do not know why — where do we look?
Start with what changed, not with what is wrong. A consent banner, a tag deployment, a cookie banner, a website migration, a consent-mode rollout and a server-side tagging change can each do this. Search Console alongside GA4 tells you whether it was collection or genuine traffic. And check whether a gap in your tag config stopped an event firing — that produces exactly this symptom and is often the simplest explanation.
Need analytics & tracking for your business?
Get a free, no-obligation review of where you stand today on analytics & tracking — and the three changes that would move the number that matters. No pitch deck.
Book a free 30-minute audit