Marketing Analytics: GA4, Server-Side Tracking And The Honest Baseline

Marketing analytics at ScaleGrowth Digital is built around three working rules. Tracking has to be repaired before reporting is worth reading. The baseline a brand reports against has to be the trough it inherited, not the peak that came before it. And every monthly number has to be reproducible from a script the brand owns. The team installs (or rebuilds) GA4, GA4 measurement protocol, server-side GTM, call-tracking and product analytics, then ships monthly reports that read against the same database the leadership team already trusts.

Why analytics breaks silently

Tracking failure is rarely loud. A daily-pageview metric drops from 527 to 24, the engagement-rate metric falls from 72 percent to 52 percent, and nobody notices because the dashboard still renders. The cause, on one inherited property, was a single GTM container edit on December 11 that fired across the site for months before the audit caught it.

The same pattern shows up across other surfaces. A Laravel sync worker dies for three and a half days because Railway only restarts the foreground process. A robots.txt file gets served as Angular HTML because of a Router intercept bug. A GTranslate WAF misconfiguration returns 403 on 71 percent of crawled pages without anyone noticing in the brand’s GA dashboards because the WAF was filtering crawlers, not users. The brand does not need more dashboards. The brand needs a tracking-health gate that catches breakage in hours, not in months.

The second category of analytics failure is reporting frame. A team inherits a property where October was the peak and December was the trough. Reporting in February against the October peak makes the work look like a decline. Reporting against the December trough, honestly framed as the inherited state, makes the same data show a recovery. The framing is the work.

How the stack gets installed

Four phases.

Phase 1: tracking audit. Two weeks. The team runs a Playwright-based audit against the site to confirm which events fire, when, with what payload. GA4 BigQuery export is checked for completeness. Server-side GTM (where it exists) is checked for double-firing and silent drops. Call tracking is checked end of funnel. The output is a tracking-health report listing every issue by priority.

Phase 2: remediation. Two to four weeks depending on depth. Container edits, event-payload corrections, server-side GTM hookup, lead-source tagging, redirect chain cleanup, GA4 measurement protocol where the marketing-automation platform needs to push data to GA4 directly, and CRM webhook hookup on the lead capture surface.

Phase 3: reporting layer. Two weeks. The team writes a Python (or, where the brand prefers, dbt) pipeline that pulls from GA4 BigQuery, the CRM, the call-tracking export and any product analytics warehouse. The pipeline outputs an HTML report (12 sections, with charts and commentary) plus an Excel workbook (seven sheets with MoM primary, Daily-Averages secondary, category deep-dive and rebuilt narrative). The report rebuilds on a single command on the brand’s side.

Phase 4: monthly cadence. Ongoing. First Tuesday of every month the report goes out. Every number is reproducible. Anomalies caught between months get flagged in a Slack or email digest the same day.

This service ties into the CRO service for hypothesis testing, the marketing automation service for lifecycle attribution and the team that runs the build.

Cases we have re-baselined

For an industrial-materials manufacturer in Australia, the team inherited a property where pageview tracking had silently broken on December 11. Daily pageviews collapsed from 527 to 24. Engagement-rate dropped from 72 percent to 52 percent. The recovery story would have read as a decline if anchored to the October peak. Re-baselining against the December trough, then walking forward, the report showed leads doubling from a 120-per-month baseline to 139 in 25 days. The same data set also surfaced the Insulated 21.1 percent conversion paired with a 17.8 percent traffic decline, the Corodek 74.7 percent conversion rate (the SKU to protect), and the Wall Cladding 278 sessions and zero conversions gap. NSW and QLD together accounted for over 60 percent of every category. The monthly report has been live on the analytics route since March, rebuildable on a monthly cadence.

For a major BFSI lender, the analytics work surfaced the AI brand-mention baseline that anchors the AI visibility program: 8 percent on ChatGPT, 15.6 percent on Google AI Overview and 19 percent on Google AI Mode. The locked prompt set runs across 300 prompts and reports out in the same monthly cadence as the GA4 work. The lift on this metric, post a content-and-schema sprint, is the primary success measure.

For a multi-location F&B brand with 86 active stores, the analytics layer sat on the Laravel command centre rather than on GA4. Per-store revenue, per-SKU AOV, repeat rate and pilot performance pulled from the Rista POS sync. The per-store baseline (1.58 lakh rupees per store per month on the four pilot stores, against a founder-stated 4 lakh) was the number that corrected the Q2 spend envelope from 51 to 78 lakh down to 5.93 to 8.23 lakh. The analytics did the strategy work; without the pull from the database, the Q2 budget would have been eight times over.

Cadence

The team is one analytics lead, one data engineer, one front-end engineer (for the on-site events) and one editorial lead (for the report commentary). The brand provides GA4, GTM, CRM, BigQuery and product-analytics access on day one. Reporting is monthly. The monthly cadence includes a written narrative, the HTML report, the Excel workbook and a one-page board-ready summary.

Cost

The installation phase (six to eight weeks) is priced at twelve to twenty-five thousand US dollars internationally and eight to seventeen lakh rupees in India, depending on the depth of remediation needed. The monthly retainer is priced at two thousand five hundred to five thousand US dollars per month internationally and one and three-quarter to three and a half lakh rupees per month in India. BigQuery, GA4 360 (if needed at scale), call-tracking and product-analytics licenses are separate.

Frequently asked questions

Do you only work on GA4? No. The team works on GA4, GA4 BigQuery export, Mixpanel, Amplitude, Heap, Adobe Analytics, server-side GTM, call-tracking platforms (CallRail, Marchex), Looker, Metabase, and custom warehouses on Postgres, Snowflake or BigQuery.

How long until the first honest monthly report? Eight weeks if remediation is light. Twelve weeks if remediation is deep (broken tracking, server-side GTM hookup, redirect chain cleanup, identity collisions). The first report carries the new baseline and the framing rule (report against the inherited trough, not the prior peak).

What is the minimum engagement? An eight-week installation. Monthly retainers run on a six-month minimum because the second and third months are when anomaly-flagging starts to pay back.

Do you build the reporting pipeline in our stack or yours? The brand’s stack. The Python or dbt pipeline lives in the brand’s repository. Credentials sit in the brand’s secret manager. The brand can re-run the report without the team.

Do you provide AI visibility reporting as part of this? Yes, when the brand is also running the AI visibility service. The same monthly cadence carries an AI brand-mention rate, an AI citation-source rate and an answer-frame share across ChatGPT, Google AI Overview, Google AI Mode and Perplexity.

By ScaleGrowth Digital Editorial

{
“@context”: “https://schema.org”,
“@graph”: [
{
“@type”: “Service”,
“@id”: “https://scalegrowth.digital/services/analytics/#service”,
“name”: “Marketing Analytics”,
“serviceType”: “GA4 install and remediation, server-side GTM, call tracking and monthly reproducible reporting”,
“provider”: {“@type”: “Organization”, “name”: “ScaleGrowth Digital”, “url”: “https://scalegrowth.digital/”},
“areaServed”: “Worldwide”,
“offers”: [
{“@type”: “Offer”, “name”: “Installation phase (six to eight weeks)”, “priceSpecification”: {“@type”: “UnitPriceSpecification”, “minPrice”: “12000”, “maxPrice”: “25000”, “priceCurrency”: “USD”}},
{“@type”: “Offer”, “name”: “Monthly retainer”, “priceSpecification”: {“@type”: “UnitPriceSpecification”, “minPrice”: “2500”, “maxPrice”: “5000”, “priceCurrency”: “USD”, “unitText”: “MON”}}
]
},
{
“@type”: “FAQPage”,
“@id”: “https://scalegrowth.digital/services/analytics/#faq”,
“mainEntity”: [
{“@type”: “Question”, “name”: “Do you only work on GA4?”, “acceptedAnswer”: {“@type”: “Answer”, “text”: “No. The team works on GA4, GA4 BigQuery export, Mixpanel, Amplitude, Heap, Adobe Analytics, server-side GTM, call-tracking platforms, Looker, Metabase and custom warehouses on Postgres, Snowflake or BigQuery.”}},
{“@type”: “Question”, “name”: “How long until the first honest monthly report?”, “acceptedAnswer”: {“@type”: “Answer”, “text”: “Eight weeks if remediation is light, twelve weeks if it is deep. The first report carries the new baseline and the framing rule that reports against the inherited trough rather than the prior peak.”}},
{“@type”: “Question”, “name”: “What is the minimum engagement?”, “acceptedAnswer”: {“@type”: “Answer”, “text”: “An eight-week installation. Monthly retainers run on a six-month minimum because the second and third months are when anomaly-flagging starts to pay back.”}},
{“@type”: “Question”, “name”: “Do you build the reporting pipeline in our stack or yours?”, “acceptedAnswer”: {“@type”: “Answer”, “text”: “The brand’s stack. The Python or dbt pipeline lives in the brand’s repository, credentials sit in the brand’s secret manager, and the brand can re-run the report without the team.”}},
{“@type”: “Question”, “name”: “Do you provide AI visibility reporting as part of this?”, “acceptedAnswer”: {“@type”: “Answer”, “text”: “Yes, when the brand is also running the AI visibility service. The same monthly cadence carries AI brand-mention rate, citation-source rate and answer-frame share across ChatGPT, AI Overview, AI Mode and Perplexity.”}}
]
}
]
}