Marketing Analytics: Every Tool in the Stack, Explained

The five layers, bottom-up. Everything in layers 1-4 is free or has a free tier; layer 5 is mostly method, not software.

Most Indian businesses we audit have Google Analytics installed and cannot answer a simple question with it: which channel brought the customers who actually paid? The tool isn't the problem. The problem is that "analytics" got treated as one plugin to install, when it is really a small system of tools that each answer a different question — and half of them were never connected.

This guide explains every entity in that system in the same shape: what it is, the question it answers, the question it can't, whether it's free, when you genuinely need it, and the mistake we see most often. If you only read one section, read the setup order at the end — it is the difference between web analytics that decorate a report and marketing analytics that change a budget.

The stack at a glance

Eighteen tools and methods, one table. "Free" means the vendor's own plan is free at the scale a small or mid-sized business uses it; paid tiers exist for most of them and prices change, so we don't quote them.

THE FIVE LAYERS

What each layer is for

Layer 1 — Collect: the tags on your site

Everything else in this guide reads from this layer. A missing tag, a duplicated tag or a form that fires the "thank you" event twice will quietly corrupt every report built on top of it, which is why the setup order at the end starts here and stays here for a week.

Google Analytics 4 (GA4)

What it is: Google's web and app analytics platform, and the default "analytics" on most Indian websites. GA4 records events — a page view is an event, a click is an event, a purchase is an event — and stitches them into sessions and users.

What it answers: how many people came, from which channel, which pages they saw, how long they stayed, which of them did the thing you care about (a "key event", GA4's name for a conversion), and roughly what that was worth.

What it can't: tell you why anyone left (that is the behaviour layer), show you the Google queries they searched (Search Console), or give you exact numbers once a report gets large — GA4 samples and applies privacy thresholds, so two reports can disagree by a few percent. The standard, free property also keeps event-level data for a limited window; if you want history you can query years later, export it to BigQuery.

When you need it: from day one. Even a five-page site should have GA4 collecting before anyone spends a rupee on ads, because you cannot backfill data you never collected.

The mistake we see most: a GA4 property with 40,000 sessions and zero key events. Traffic without conversions is trivia. Marking the form submit, the WhatsApp click and the call click as key events is a one-hour job that makes every other tool in this guide useful.

Go deeper: GA4 explained: events, key events, reports and Explorations — and, if you need the property and data-stream steps, GA4 setup for marketers.

Google Tag Manager (GTM)

What it is: a container you install once, which then loads and controls every other tag — GA4, Meta Pixel, LinkedIn Insight Tag, a chat widget — from one interface, without a developer editing the site for each change. It measures nothing itself.

What it answers: not a data question. GTM answers an operations question: "how do I add, change or fix tracking without touching the theme?" Its triggers (fire on a button click, a form submit, a scroll depth) are how most events reach GA4 and the ad platforms in the first place.

What it can't: replace GA4 or any pixel — it is the delivery van, not the shop. It also can't fix a site that never exposes the data you need; a "purchase" trigger needs the order value to exist somewhere on the page (the data layer), which is a developer job.

When you need it: the moment you have more than one tag, or one tag that needs to fire on something other than a page load. For a single GA4 tag on a brochure site, the direct GA4 snippet is fine.

The mistake we see most: two GA4 tags — one in GTM, one hard-coded in the theme by a previous agency — doubling every page view. Check with the browser's Tag Assistant before trusting any number.

Go deeper: Google Tag Manager: setup guide for marketers — the container, the first five tags and the Preview loop; the conversion tracking guide covers the triggers that matter for leads and sales.

UTM parameters

What it is: five optional bits added to the end of a link — utm_source, utm_medium, utm_campaign, utm_content, utm_term — that tell GA4 exactly where a click came from. Without them, a click from your Instagram bio, a WhatsApp broadcast and an email newsletter can all land in GA4 as "direct".

What it answers: which specific campaign, post, creative or email produced a visit — the difference between "social brought 400 sessions" and "the Tuesday Reel brought 310 of them".

What it can't: survive inconsistency. facebook, Facebook and fb become three sources. A naming convention written down once beats any tool.

When you need it: the first time you share a link anywhere you control — social bios, ads that don't auto-tag, emails, QR codes, WhatsApp.

The mistake we see most: UTMs on internal links. Tagging a link from your homepage to your pricing page restarts the session as a new "campaign" and destroys the original source.

Go deeper: UTM parameters: best practices and a naming convention, then build them with the UTM campaign URL builder.

Meta Pixel and the Conversions API

What it is: Meta's tracking tag (the Pixel, in the browser) plus its server-side twin (the Conversions API), which together report what people did on your site after clicking a Facebook or Instagram ad.

What it answers: which ads produced leads and purchases, so Meta can optimise delivery toward people likely to convert, and so you can build retargeting and lookalike audiences.

What it can't: see everything from the browser alone — ad blockers and iOS privacy changes hide a slice of conversions from the Pixel, which is why Meta pushes the Conversions API alongside it.

When you need it: before your first Meta campaign. Running ads without it is spending money without a receipt.

Go deeper: Meta Pixel and Conversions API. B2B advertisers, the equivalent for LinkedIn is the LinkedIn Insight Tag, which also shows which companies visited.

Server-side tracking

What it is: the same events, sent from a server you control instead of from the visitor's browser. Google Tag Manager has a server-side mode; Meta's Conversions API is the same idea for one platform.

What it answers: "how much of my conversion data is being lost to ad blockers, browser privacy features and slow pages?" — and then it recovers a good part of it.

What it can't: be free. The tagging software is, but it needs a server, and someone has to maintain it. For a lead-gen site with a few hundred conversions a month the loss it recovers rarely justifies the cost.

When you need it: ecommerce at scale, or any account where ad spend is large enough that a 15–25% gap in reported conversions changes bidding decisions.

Go deeper: Server-side tracking: what it is and when you need it.

Consent, the DPDP Act and first-party data

What it is: the rules layer under everything above. India's Digital Personal Data Protection Act, 2023 now has its Rules — notified on 14 November 2025 with an eighteen-month phased compliance window — which require clear, purpose-specific consent notices for personal data, and consent managers based in India. Analytics and ad pixels collect personal data. This is not legal advice; it is the reason your tracking plan needs a consent plan next to it.

What it answers: "what am I allowed to collect, and what happens to my data when a visitor says no?" Google's Consent Mode is the mechanism that lets GA4 and Google Ads respect a visitor's choice and model what it can't see; a consent banner is the interface in front of it.

What it can't: be bolted on the night before a deadline. Consent changes which events fire, which changes every number in every report — plan for a visible step-change in the data and annotate it.

When you need it: now, for the planning; the compliance window is the deadline, not the start date. The upside is first-party data — the email, phone and purchase history people give you directly — which no privacy change can take away.

Go deeper: DPDP Act for marketers: consent, cookies and analytics, and First-party data: why it matters.

Layer 2 — See search: what Google and Bing see

GA4 starts at the click. This layer starts before it — which queries you appeared for, how often, where — and it is the only place the search engines tell you themselves.

Google Search Console (GSC)

What it is: Google's free report card for your site: the queries you appeared for, impressions, clicks, average position, which pages are indexed and which aren't, Core Web Vitals, and any manual action against you.

What it answers: "is Google finding, indexing and ranking my pages — and for what?" It is where an SEO strategy is measured, and the first place to look when traffic drops.

What it can't: see anything after the click — sessions, bounces and conversions belong to GA4. It also reports with roughly a two-day lag, samples query data and keeps sixteen months of history, so export what you want to keep.

When you need it: before you publish page two. Verifying the property is a ten-minute job and every day without it is a day of search data you never get back.

The mistake we see most: reading average position as a single number. A page at "position 8" is usually position 3 for some queries and position 40 for others; filter by query before drawing a conclusion.

Go deeper: Google Search Console: what it is and how to use it.

Bing Webmaster Tools

What it is: Microsoft's equivalent of Search Console, for Bing. It can import your verified sites straight from GSC, so setup takes minutes.

What it answers: the same questions as GSC, for a smaller share of Indian search — plus a share that is growing in importance, because Microsoft Copilot and a number of AI assistants draw on Bing's index. If Bing can't crawl you, some AI answers can't cite you.

What it can't: tell you about AI-answer visibility directly. That is a separate, still-immature measurement problem — see our AI search guide.

When you need it: the same afternoon you set up GSC. There is no reason to skip it.

Go deeper: Bing Webmaster Tools: why it matters more than you think.

Rank tracking (Semrush, Ahrefs and the like)

What it is: a paid tool that checks a fixed list of keywords every day and records where you rank, usually alongside competitor domains.

What it answers: "did the position for the keywords I care about move this week, and did a competitor's?" GSC gives averages across every query; a rank tracker gives a clean daily line for the fifty that matter.

What it can't: replace GSC. It only knows the keywords you told it about, and its India-wide "position" ignores the personalisation and location that shape real results.

When you need it: once SEO is a line item in the budget and someone has to report on it monthly.

Go deeper: Rank tracking: what to track and what it can't tell you, and SEO reporting: what to actually track.

Layer 3 — See behaviour: what people do on the page

GA4 tells you 70% of visitors left the pricing page. This layer shows you the visitor scrolling past the price, hovering on "Contact", and leaving.

Heatmaps and session recordings

What it is: a script that aggregates where visitors click (click maps), how far they scroll (scroll maps) and where the cursor moves, and that records anonymised replays of individual visits.

What it answers: "why is this page not converting?" — the button nobody sees below the fold, the image people click expecting a link, the form field where everyone stalls, the "rage clicks" on something that doesn't respond.

What it can't: tell you how many. Ten recordings of people stalling on a field is a hypothesis to test, not a statistic; the numbers still come from GA4 and, ideally, an A/B test.

When you need it: the moment you are trying to improve a specific page rather than bring more people to it.

Go deeper: Heatmap tools: what they show and how to read them.

Microsoft Clarity (and Hotjar)

What it is: the two tools most Indian teams actually use for the layer above. Microsoft Clarity gives heatmaps, recordings and a set of frustration signals (rage clicks, dead clicks, quick-backs) free, without a traffic cap, and integrates with GA4. Hotjar is the older category leader, with a limited free plan and paid tiers that add surveys, feedback widgets and user interviews.

What it answers: the behaviour questions above. Clarity's frustration filters are the fastest way we know to find a broken element on a page; Hotjar's surveys answer the one thing a recording can't — what the visitor was thinking.

What it can't: record everything on every site type. Both need consent handling under the rules in layer 1, and both mask form inputs by default — check that your own checkout doesn't need extra masking.

When you need it: Clarity on day one, because it costs nothing; Hotjar when you need surveys or your team already lives in it.

Go deeper: Microsoft Clarity: what it shows, and Clarity vs Hotjar; the CRO tools comparison covers both alongside the testing tools below.

A/B testing

What it is: showing two versions of a page to two halves of your traffic and measuring which one converts better, with enough visitors that the difference isn't luck.

What it answers: "did the change cause the lift?" — the only tool in this guide that answers a causal question about your site.

What it can't: work on low traffic. A page with 300 visits a month cannot detect a 10% improvement in any reasonable time; fix the obvious things heatmaps show you and test later. Google retired its free testing tool in 2023, so this layer is now paid (VWO, Optimizely, Convert) for most teams.

When you need it: when a page already gets thousands of sessions a month and a small conversion-rate change is worth real money.

Go deeper: A/B testing: the complete guide.

Layer 4 — Report: one page someone will read

Every tool above has its own interface, and nobody who signs off a budget will open five of them. This layer pulls the numbers that matter into one place — and decides which numbers those are.

Looker Studio (now Data Studio)

What it is: Google's free dashboard builder. It was Google Data Studio until 2022, Looker Studio until April 2026, and Google has now renamed it Data Studio again — most marketers still search for "Looker Studio", so we use both names. It connects natively to GA4, Search Console, Google Ads, YouTube and Google Sheets, and to almost everything else — Meta Ads, LinkedIn, Shopify, a CRM — through paid third-party connectors.

What it answers: "what happened this month, across every channel, in a link I can send the client?" A well-built report refreshes itself; the monthly reporting job becomes writing the commentary, not assembling the numbers.

What it can't: fix bad inputs (a report on double-counted GA4 data is a beautifully formatted wrong answer), or avoid the GA4 connector's API quotas on busy dashboards — heavy reports sometimes need the data extracted to Sheets or BigQuery first. Data Studio Pro (formerly Looker Studio Pro), the paid tier, adds team management, AI features and Google Cloud integration, not better charts.

When you need it: the first time someone asks you for a report twice.

The mistake we see most: forty charts. A report that shows everything shows nothing; the KPI list below decides what goes on page one.

Go deeper: Looker Studio (now Data Studio): marketing dashboards guide; SEO reporting shows a report layout built for decisions rather than decoration.

Marketing dashboards and KPIs

What it is: the short list of numbers that tells you whether marketing is working, in the order they matter — usually revenue or qualified leads first, cost per lead or acquisition second, then the channel metrics that explain them, and only then traffic.

What it answers: "is it working?" in under a minute, for someone who doesn't know what a session is.

What it can't: be copied from another business. A D2C brand's page one is ROAS, blended CAC and repeat rate; a B2B manufacturer's is qualified enquiries, cost per enquiry and pipeline. Vanity metrics — followers, impressions, "engagement" — belong on page three, if anywhere.

Go deeper: Marketing dashboard: what goes on it (and what stays off), and Marketing KPIs that actually matter.

Layer 5 — Decide: turning numbers into a budget

These are methods more than tools. They sit on top of everything above and answer the only question that ultimately matters: where should the next rupee go?

Conversion tracking

What it is: telling each ad platform — Google Ads, Meta, LinkedIn — which of its clicks became a lead or a sale, either with the platform's own tag or by importing GA4 key events.

What it answers: cost per lead and cost per sale by campaign, ad group and keyword; and it feeds the platforms' automated bidding, which cannot optimise toward a result it can't see.

What it can't: agree across platforms. Google Ads, Meta and GA4 will each claim the same conversion, and the sum will exceed the truth — that is normal, and it is why attribution exists.

When you need it: before the first campaign goes live. This is the one item in the guide we refuse to launch without.

The mistake we see most: counting a page view of the thank-you page as a lead, then discovering the sales team refreshes that page. Fire the conversion on the form's actual submit event, and set the offline sale as a second, separate conversion when the CRM allows it.

Go deeper: Conversion tracking. For phone-heavy businesses, call tracking — from free tap tracking to a dynamic phone number per source — is the missing piece most Indian lead-gen sites never install.

Attribution, marketing mix modelling and incrementality

What it is: three answers to "which channel deserves the credit?" Attribution follows individual users across touchpoints (GA4's default is data-driven; last click is the simple version). Marketing mix modelling uses statistics on aggregate spend and outcomes, no user tracking required. Incrementality tests run a channel in some cities and not others and measure the lift.

What it answers: whether the channel that gets the last click — usually brand search or a retargeting ad — actually created the demand, or merely collected it.

What it can't: be precise for a small budget. Attribution weakens as tracking gets more restricted; MMM needs a year or more of varied spend to say anything; incrementality tests need enough volume to see a difference. Below roughly ten lakh a month in spend, a sensible rule of thumb and a quarterly holdout beat a model.

Go deeper: Attribution modelling, Marketing mix modelling and Incrementality testing.

Customer lifetime value, cohorts and BigQuery

What it is: the analysis that follows a customer past the first purchase. Cohort analysis groups customers by the month (or channel) they arrived and tracks what each group does afterwards; lifetime value is the revenue a cohort produces over time; BigQuery is Google's data warehouse, which GA4 can export raw, unsampled event data into for free, so you can run exactly these queries without the interface's limits.

What it answers: "what can I afford to pay to acquire a customer?" — because acquisition cost only means something next to what the customer is worth. A channel with a high cost per sale and a high repeat rate can be the best channel you have.

What it can't: be done in GA4's standard reports alone, and BigQuery's querying moves beyond the free tier at volume. For most businesses a monthly cohort table in a spreadsheet, built from the order export, is the honest starting version.

When you need it: ecommerce, subscriptions and apps — anywhere the first purchase isn't the last.

Go deeper: Customer lifetime value: formula, examples and uses, Cohort analysis: read the table, not the average and BigQuery for marketers; Customer acquisition cost and the CAC calculator put a number next to it.

How the layers work together

Follow one visitor through the stack. She searches "cold-pressed oil Mumbai"; Search Console records the impression and, when she clicks, the click. Tag Manager loads GA4 and the Meta Pixel; GA4 records the session, sourced to organic search. She reads, hesitates and leaves — Clarity's recording shows her scrolling past the price to look for a delivery area, which the page never mentions. You add delivery areas above the price; three weeks later she returns from a retargeting ad and orders. Conversion tracking tells Meta the ad worked; GA4's attribution gives organic search the assist; the Looker Studio report shows both; the cohort table, six months on, shows that customers from organic search reorder more often than customers from ads — so the next budget moves a little toward SEO.

No single tool sees that whole story. That is the point of building all five layers, and of building them in the order that follows.

The setup order that works

Eight steps, in sequence. Most businesses can finish the first five in a week, with no spend beyond time.

1. Install Google Tag Manager, and move any existing tags into it.Remove hard-coded GA4 and pixel snippets from the theme so every tag has one home. Check with Tag Assistant that each fires once.
2. Set up GA4 through GTM and define your key events.Form submit, WhatsApp click, call click, purchase — whatever a customer actually does. Do this before you look at a single traffic report.
3. Verify Google Search Console and Bing Webmaster Tools.Submit the sitemap. This costs nothing and starts the sixteen-month clock on search history.
4. Add the ad pixels you use — Meta, LinkedIn — and connect conversion tracking.Each platform should receive the same key events GA4 does. Reconcile the counts once; expect them to differ, and know why.
5. Install Microsoft Clarity and write down your UTM convention.Both are free and both start paying back in the first week — Clarity by showing you the broken element you didn't know about, UTMs by ending the "direct traffic" mystery.
6. Add consent handling and annotate the change.A consent banner wired to Google Consent Mode, with a note in every report about the date the data changed. Plan this against the DPDP compliance window, not after it.
7. Build one Looker Studio (Data Studio) report with ten numbers on page one.Decide the KPI list first; connect the sources second. If a chart doesn't change a decision, it goes on page two or nowhere.
8. Only then: attribution, cohorts, server-side, A/B testing.These are the expensive layers, and they only pay when the cheap ones are clean. Add them when spend, traffic or repeat purchases make the questions they answer worth real money.