Meta, Google and Shopify all report different sales numbers
- What we'd do
- Server-side tagging, UTM rules, BigQuery model, MER reporting and incrementality tests
- What moves
- Reporting disputes down, budget moved on agreed numbers
Close to nothing in software at mid-size volumes: GA4 standard, Google Tag Manager and Looker Studio are free, BigQuery and server-side hosting add a modest cloud bill, and setup and upkeep are the main cost.

Red Balloon is a marketing analytics agency for Indian brands whose ad platforms, GA4 and CRM disagree about what actually sold. We build one measurement system you own: a tracking plan, GA4 and Tag Manager, server-side events, a warehouse and reports that tie spend to orders or pipeline. It suits marketing heads who have to defend a budget to a founder or CFO, and founders tired of three dashboards telling three stories. What sets the work apart is the starting point: the revenue number your finance team already trusts, worked backwards to every channel, with every property, container and dashboard sitting in your company's accounts from day one.
Some fixes are a few hours of in-house work: a missing UTM rule, a duplicate tag, a key event counted twice. Bring in a GA4 consultant when the gaps are structural instead. The usual signs are purchases or leads in GA4 that never match the backend, paid social claiming sales your store never shipped, offline outcomes such as test drives or admissions that never reach the ad platforms, and a site rebuild planned with nobody owning the dataLayer. Each of these quietly misdirects budget for every week it is left alone.
The first deliverable is not a dashboard. It is a written tracking plan that names every event, parameter, key event and owner, followed by a conversion tracking setup that is tested against real orders or CRM records before anyone reports on it. Only then do we build attribution views and, where spend justifies it, incrementality tests. Reporting runs weekly on a short list of agreed numbers, and the tag health checks and anomaly alerts keep running after handover, so a broken tag is caught the same day rather than at month end.
Also covered on this page
What happens, how long it takes, what you get and what we need from you.
Tag and data audit, stakeholder interviews
You get: Audit
We need: Access to GTM, GA4, ad platforms and CMS
Measurement plan, events, consent
You get: Measurement plan
We need: KPI decisions
GTM, server-side, GA4, Conversions APIs, CRM links
You get: Working tracking
We need: Developer time, cloud account
BigQuery, data models, dashboards
You get: Dashboards
We need: Data access
QA, release checks; geo test re-run when a channel's spend changes by more than 30% (rule of thumb)
You get: Monthly health report
We need: Advance notice of site changes
Gurugram eyewear retailer with Shopify plus 20 stores, running Google, Meta and marketplace ads
Every metric has a formula, a target, a source and a reporting rhythm.
| Metric | How it's calculated | Target | Source | Cadence |
|---|---|---|---|---|
| Order match rate | GA4 purchases ÷ backend orders, same dates and timezone | Rule of thumb: within 10% of backend orders; alert when a day falls outside | GA4 + Shopify or backend | Daily |
| MER | Total revenue ÷ total marketing spend | Floor set from contribution margin | BigQuery or Looker Studio | Weekly |
| New-customer CAC | Total marketing spend ÷ new customers | Below the allowed CAC from margin and repeat rate | Backend + spend data | Weekly |
| Channel ROAS (data-driven) | Channel-attributed revenue ÷ spend; GA4 data-driven model, default 90-day lookback for key events | Read against incrementality results, not alone | GA4 advertising reports | Weekly |
| Consent rate | Users granting analytics or ads consent ÷ users shown the banner | Known and tracked; modelled data covers part of the gap | Consent management platform dashboard | Monthly |
| Unassigned traffic share | (not set) and Unassigned sessions ÷ total sessions | Set from audit; drive it down | GA4 traffic acquisition report | Weekly |
| Incremental lift | (Test − control outcome) ÷ control outcome in a geo or holdout test | Lift that holds up statistically | Experiment analysis in BigQuery or a notebook | Per test |
The manual work we take out of this service, and the tools we use.
Tracking personal data needs notice, consent where required and retention limits. Main duties apply from 13 May 2027.
Never send names, emails or phone numbers to GA4; hash only where ad tools allow.
Applies to EEA and UK visitors; Indian sites with European traffic need it.
A useful benchmark for access controls on analytics and warehouse data.
We track: Google Analytics Help: What's new, Google Tag Manager release notes, Meta for Developers changelog, MeitY notifications.
Check which key events are imported into Google Ads and how.
Any history you still need must already be exported.
Chrome cookies continue but Safari still blocks them; first-party and server-side data still matter.
Plan consent and retention controls now.
Tags without a plan create mess
Events, naming, owners, consent approach
Attribution models disagree
Geo or holdout test designs
Undocumented setups decay
Tag map, data dictionary, handover
Releases break tracking
Test checklist, automated checks
Dependence is costly
Training, dashboards, documentation
Every engagement is tied to one business metric, tracked before and after over a set period. Named case studies with figures follow as clients approve publication.
All workTracked as patient enquiries, before and after, over a fixed period
Performance Marketing →Tracked as site visits booked, before and after, over a fixed period
Product Engineering →Tracked as hours saved per month, before and after
AI & Automation →Tracked as price premium or repeat rate, before and after
Brand & Positioning →Tracked as organic demo requests, before and after
Search & Content →Tracked as qualified leads in launch month, against plan
Launch & Go-to-Market →Straight answers from the practice lead. Highlighted text marks Red Balloon figures we confirm before publishing.
Three to six weeks for a tracking plan plus GA4, GTM and ad-platform implementation; attribution reporting then needs one to two months of clean data before it is reliable.
GA4 with web and server-side Google Tag Manager, feeding BigQuery as the warehouse, plus conversion APIs that send outcomes to Google, Meta and LinkedIn. Smaller teams can stop at GA4 and Looker Studio.
Give notice and take consent before tracking that identifies people, with a banner that actually blocks tags until consent. Keep personal data out of URLs and events, and set retention limits.
No single model: GA4 data-driven attribution for day-to-day channel decisions, platform numbers only within each platform, and incrementality tests or marketing mix modelling for budget splits across channels.
Your company should own the GA4 property, GTM containers, BigQuery project and Google Cloud billing, with the agency added as a user. Agency-owned accounts can take history and configuration with them.
A paid two-week Growth Audit covers your funnel, ads, site, tracking and automation gaps, and ends with a 90-day plan you can run with or without us. The fee is credited against your next engagement.
Prefer to talk? Call +91 97114 48083, or email info@redballoon.in.
A person from the team replies to every enquiry.