A GA4 property takes ten minutes to install and is then never configured. The data flows, the reports exist, and nobody can answer a straightforward question about which channel produced revenue.
GA4 setup is configuring Google Analytics 4 beyond the basic tag: defining the events that represent real business actions, marking the right ones as key events, configuring e-commerce tracking with accurate values, filtering internal and bot traffic, linking Google Ads and Search Console, and building reports around the questions the business actually asks. Installing the tag takes ten minutes. Configuration is the part that makes the data usable.
GA4 arrived with an event-based model that is genuinely more flexible than what came before, and a default configuration that tells you almost nothing specific to your business. The gap between those two is where most properties sit.
The practical test is simple. Can you answer, from your own GA4, which channel produced the most revenue last month, what a lead is worth by source, and where people abandon your checkout? If not, the property is collecting data rather than producing information.
Not a feature list. These are the areas where the work actually changes the outcome.
Deciding what to measure before implementing anything. Which user actions represent real business outcomes, what parameters they need to carry, and consistent naming so reports stay coherent as the site grows. Retrofitting a naming convention after six months of data is painful.
Marking the events that matter and, crucially, not marking the ones that do not. A property with fourteen conversions has none. Everything looks like success and the Ads algorithms receive a diluted signal.
The full purchase funnel with accurate item data and values: view item, add to cart, begin checkout, purchase. This is where most implementations are incomplete, and an incomplete funnel means you cannot see where you are losing money.
Internal traffic filtered, bot traffic excluded, referral exclusions for payment gateways so a Razorpay redirect does not break session attribution, and cross-domain measurement where you span domains. Payment gateway attribution loss is extremely common and quietly misattributes your best revenue.
Google Ads, Search Console and Merchant Center linked, so paid and organic performance can be seen together and conversions can be imported where that improves bidding.
Consent Mode where you serve markets that require it, and a configuration that is defensible under DPDP Act expectations. Worth setting up deliberately rather than retrofitting under pressure.
Custom reports and funnel explorations built around the questions your team asks. The default reports are generic by design; the useful ones have to be built.
Where you are moving from an older setup, preserving what can be preserved and being clear about what cannot. Historical data does not transfer between models, and pretending otherwise causes trouble later.
In this order, because each step depends on what the previous one found.
What decisions this data should support. Measurement designed backwards from questions is usable; measurement designed forwards from available events is not.
Tag Manager container, event tracking, e-commerce data layer, key events and platform links.
Real transactions and real form submissions tested end to end, reconciled against backend records. Reported numbers that do not match reality are worse than no numbers.
Custom reports and explorations, plus a walkthrough so your team can use them without asking us.
Events, key events, e-commerce and data quality settings complete.
Documented, with a clear structure someone else can maintain.
Reported figures reconciled against backend orders or CRM records.
Built around your actual questions, with a handover session.
A fixed-fee implementation project, scoped on site complexity, a lead-generation site and a multi-currency store are different jobs. Ongoing analytics support folds into a marketing retainer where one exists. We usually recommend doing this before any significant advertising spend, because optimising campaigns against unreliable data costs more than the setup does.
See indicative pricingProbably, and it is quick to check. Look at whether your key events reflect real business outcomes, whether e-commerce values match your backend, and whether payment gateway referrals are excluded. Most properties fail at least one.
Different attribution models and different windows. They are counting differently by design, not malfunctioning. Understanding which to use for which decision matters more than making them match, and they will not match.
Historical data does not migrate between the two models. If you exported before shutdown you can retain it for reference, but it is not comparable like for like. Anyone claiming a clean migration of historical data is overstating it.
Not necessarily. Standard GA4 is adequate for most businesses. Server-side becomes worth its cost when data loss from ad blockers and browser restrictions is materially affecting decisions, which is more common in high-spend e-commerce.
Server-side GTM and CAPI for accounts where measurement loss is costing real money.
ExploreA full account review with a prioritised fix list, yours to keep.
ExploreCrawl, index, speed and schema diagnosis with a prioritised fix list.
ExploreSend a few details and we will reply on WhatsApp with a first read on what we would do, before any talk of fees.
