Advertising platforms are good at reporting impressions, clicks, and conversions recorded within their own systems. Web analytics shows visits and actions on the website. A CRM stores leads, deals, and payments.

The problem begins when these datasets remain disconnected.

The marketer sees inexpensive leads. The sales team reports that their quality is poor. Management sees total revenue but cannot identify which campaigns actually produced paying customers.

End-to-end analytics connects the journey from the first advertising interaction to the completed deal. It helps the business answer not only:

“How many leads did advertising generate?”

but also:

• how many paying customers came from each channel;
• how much it cost to acquire one customer;
• how much revenue and profit each campaign produced;
• where leads are lost;
• which sources produce many enquiries but few sales;
• where the advertising budget should be reallocated.

What is end-to-end analytics?

End-to-end analytics is a system that connects data from several stages of the customer journey:

1. advertising source;
2. website or application visit;
3. target action;
4. enquiry or call;
5. CRM record;
6. qualification;
7. deal;
8. payment;
9. refund, repeat purchase, or subsequent revenue.

The business can therefore analyse the commercial result rather than only the top of the funnel.

For example, one channel may generate leads for $5 and another for $12. If one person in fifty buys from the first channel but one in five buys from the second, the more expensive lead may be considerably more profitable.

This is why optimising only for CPL often produces poor decisions.

What does the minimum data chain look like?

A business does not need to build an expensive data warehouse immediately.

A minimum working structure may look like this:

Advertising platform → website → form or call → CRM → deal status → payment amount → report.

The relationship between the source and the customer must be preserved throughout this chain.

Common identifiers include:

• UTM parameters;
• advertising click identifiers;
• first-touch and last-touch source;
• visitor or session identifier;
• phone number;
• email address;
• lead ID;
• customer and deal IDs in the CRM.

If the relationship is lost between the form and the CRM, or between the CRM and the payment, the report is no longer genuinely end-to-end.

Which data should be collected?

Advertising platform data

Advertising platforms usually provide:

• campaign;
• ad group;
• advertisement;
• keyword or audience;
• impressions;
• clicks;
• spend;
• platform-recorded conversions.

This data explains the cost of traffic but does not independently confirm a sale.

Website data

The website should record:

• source and campaign;
• views of critical pages;
• form submissions;
• clicks on phone numbers or messengers;
• checkout start and completion;
• registrations;
• bookings;
• other commercially meaningful actions.

Not every button click should become a primary conversion. Events should represent the actual customer journey.

CRM data

The CRM should contain:

• enquiry source;
• lead creation date;
• responsible sales representative;
• status;
• reason for rejection;
• deal value;
• payment date;
• product or service;
• repeat sale;
• refund, when applicable.

When sales representatives do not update statuses and values, even a technically sophisticated analytics system will produce reports from unreliable data.

Calls and offline enquiries

If a meaningful share of customers calls the company, form tracking alone is insufficient.

Call tracking can connect a phone call to its advertising source. Enquiries from messengers, offline locations, and manually entered sources also need a clear identification method.

Otherwise, part of the revenue will be classified as direct, unknown, or manual, distorting channel performance.

Which metrics matter?

CPL — cost per lead

CPL = advertising spend / number of leads

It is useful for analysing the top of the funnel but does not measure lead quality.

A low CPL does not necessarily produce a low customer-acquisition cost.

CPA or CAC — customer-acquisition cost

A simple formula is:

CAC = acquisition costs / number of new paying customers

Depending on the reporting objective, acquisition costs may include only advertising or also marketing salaries, agency fees, software, and other related expenses.

The important requirement is to use the same definition consistently.

ROAS — return on advertising spend

ROAS = advertising-attributed revenue / advertising spend

If advertising costs $10,000 and the associated revenue is $30,000, ROAS is 3, or 300%.

A high ROAS does not guarantee profitability. Cost of goods, discounts, refunds, and operating expenses must also be considered.

ROMI

Different companies use different ROMI formulas. The business should therefore define its own calculation explicitly.

One practical version is:

ROMI = (gross profit attributed to marketing − marketing costs) / marketing costs

Using gross profit rather than revenue avoids treating products with very different margins as equally valuable.

Conversion between funnel stages

End-to-end analytics should also show transitions such as:

• click → lead;
• lead → qualified lead;
• qualified lead → proposal;
• proposal → sale;
• sale → repeat purchase.

This helps distinguish an advertising problem from a website or sales-process problem.

LTV

For subscriptions and repeat purchases, the first payment alone is insufficient.

A channel that attracts more expensive but loyal customers may outperform a source that produces a cheaper initial purchase.

However, LTV should not be based on a very short history or optimistic assumptions. For young products, it is safer to report actual revenue and forecasts separately.

Why do different systems report different numbers?

An advertising platform, Google Analytics, and the CRM may show different results even when the integration is operating correctly.

Common causes include:

• different attribution windows;
• different credit-allocation models;
• repeat visits;
• cross-device journeys;
• cookie restrictions;
• missing user consent;
• data-processing delays;
• duplicated events;
• calls and offline sales;
• different time zones;
• refunds and cancellations;
• different conversion definitions.

This does not always indicate an error.

The objective is not to force every interface to display identical numbers. It is to establish a consistent method for making business decisions.

What is attribution?

Attribution determines how conversion value is assigned across marketing touchpoints.

A customer may:

1. first see a social advertisement;
2. later search for the company;
3. return directly;
4. submit a form;
5. pay after speaking to a sales representative.

Which channel produced the sale?

The answer depends on the attribution model.

Last click

All credit is assigned to the final known source before the conversion.

It is simple but often undervalues channels that created the initial demand.

First click

All credit is assigned to the first known interaction.

It can help analyse audience acquisition but ignores later touchpoints that supported the decision.

Data-driven attribution

A data-driven model distributes credit using observed customer paths and the estimated contribution of each interaction. In GA4, the model may consider factors such as interaction order, device type, and time to conversion, while attribution-path reports help show which channels initiated, assisted, and completed key events.

Any attribution model is still a model rather than proof of causation.

If a report assigns 40% of conversion value to a channel, switching the channel off will not necessarily reduce sales by exactly 40%.

For material budget decisions, attribution should be supplemented with controlled experiments where practical.

How can actual sales data be returned to advertising platforms?

In lead-generation businesses, the meaningful conversion often occurs after a sales conversation rather than on the website.

It is therefore useful to return later funnel events to advertising platforms, including:

• qualified lead;
• scheduled meeting;
• signed agreement;
• confirmed payment;
• revenue value.

Google Ads supports enhanced conversions for leads, allowing offline conversions to be matched to advertising interactions using click IDs and hashed first-party data such as email addresses or phone numbers. This can support measurement and optimisation for later funnel outcomes rather than form submissions alone.

Meta Conversions API supports sending events from websites, applications, CRM systems, and offline sources through a server connection. It can complement browser events, but correct event deduplication and responsible data processing remain necessary.

Is server-side tracking necessary?

In a standard client-side implementation, events are sent from the user’s browser to analytics and advertising platforms.

Server-side tracking introduces an intermediate server layer where data can be validated, enriched, filtered, and routed to the required systems.

Google states that server-side tagging can improve control, data quality, and page performance by reducing client-side tag activity. It also introduces separate infrastructure, configuration, and maintenance requirements.

Server-side tracking is not the mandatory first step for every business.

It becomes more reasonable when:

• advertising spend is material;
• several systems consume the same events;
• offline conversion reporting is important;
• browser-side tracking is unstable;
• technical maintenance resources are available;
• the company can manage consent and personal-data requirements.

If CRM deals are not updated, UTM parameters are regularly lost, and forms create duplicates, server infrastructure will not fix the foundational problem.

Implementation levels

Level 1: basic visibility

Suitable for a business that is beginning to connect marketing and sales.

Required work includes:

• standardising UTM parameters;
• defining critical events;
• passing source data into the CRM;
• introducing mandatory statuses;
• recording payment amounts;
• building a basic channel report.

This level alone can reveal the difference between the number of leads and the number of customers.

Level 2: complete lead analytics

Additional components may include:

• call tracking;
• messenger integrations;
• advertising spend imported through APIs;
• offline conversions;
• lead qualification;
• rejection reasons;
• sales-representative reports;
• automatically updated dashboards.

This level is appropriate for companies operating several channels with a meaningful volume of leads.

Level 3: centralised data infrastructure

It may include:

• a dedicated data warehouse;
• server-side tracking;
• unified customer identity;
• product and billing data;
• cohort and LTV analysis;
• data-quality monitoring;
• custom attribution models;
• forecasting.

This becomes justified when business scale and the cost of poor decisions exceed the cost of the infrastructure.

Building it too early is unnecessary overengineering.

Common mistakes

Optimising for every lead

Spam, duplicates, recruitment enquiries, partnership requests, and irrelevant messages should not be valued in the same way as qualified leads.

No shared definition of a sale

Marketing may count a form submission, the CRM may count a created deal, and finance may count only received payment.

A management report needs one agreed definition.

Losing the source before the CRM

If UTM data remains only in web analytics and is not attached to the lead and deal, advertising cannot be connected to payment.

Duplicating conversions

The same event may be sent through the browser, server, CRM, and a third-party integration.

Without event IDs and deduplication rules, one sale may be counted several times.

Ignoring refunds

Revenue reporting should account for cancelled orders, refunds, and unpaid deals.

Mixing different dates

An advertisement may be clicked in January, the enquiry created in February, and payment received in March.

Reports based on click date, lead date, and payment date answer different questions.

Trusting the dashboard without verification

A polished report does not guarantee correct data.

The business should regularly verify:

• test leads;
• actual deals;
• values;
• currencies;
• duplicates;
• missing sources;
• update delays.

When is end-to-end analytics useful?

It is particularly valuable when:

• the business uses several advertising channels;
• time passes between enquiry and payment;
• a sales representative closes the deal;
• some enquiries arrive by phone;
• customer cost differs materially from lead cost;
• repeat purchases are important;
• the advertising budget cannot be allocated confidently;
• the CRM already contains sufficiently reliable data.

When should you avoid starting with a complex system?

A complete implementation may be premature when:

• only one small advertising channel is used;
• lead volume is low;
• the CRM is not maintained;
• sales representatives do not update statuses;
• funnel stages have not been defined;
• payment values are not recorded;
• the product and offer change continuously.

The business should first establish basic discipline:

1. consistent UTM parameters;
2. one CRM;
3. mandatory statuses;
4. recorded payment values;
5. a weekly report.

Automating disorganised data does not produce reliable insight.

How should implementation begin?

Define the business questions

Do not start by selecting a platform.

First determine which decisions the report must support:

• which channel should be scaled;
• which campaign should be stopped;
• where customers are lost;
• how much a new sale costs;
• which product produces more profit;
• how effectively sales representatives work.

Document the funnel

For every stage, define:

• its name;
• the transition condition;
• the responsible person;
• mandatory data;
• event date;
• final status.

Choose the source of truth

A common model is:

• advertising spend from advertising platforms;
• website behaviour from analytics;
• deal status from the CRM;
• confirmed revenue from CRM, billing, or finance.

The business should define which source takes priority when data conflicts.

Audit data quality

Check:

• how many leads have no source;
• how many deals have no value;
• how many duplicates are created;
• whether sales representatives use statuses consistently;
• whether test events match across systems;
• whether the source disappears after calls or repeat visits.

Launch a minimum report

Begin with a limited set of metrics:

• spend;
• leads;
• qualified leads;
• customers;
• revenue;
• CPL;
• CAC;
• ROAS or ROMI;
• conversion between stages.

Do not begin by creating dozens of charts.

Conclusion

End-to-end analytics is not primarily about combining as many systems as possible in an attractive dashboard.

Its purpose is to connect marketing spend with real commercial outcomes and help the business make better-supported decisions.

A reliable system begins with a simple foundation:

• consistent source tagging;
• correct events;
• CRM discipline;
• recorded sales and values;
• agreed metric definitions;
• regular data-quality checks.

Only after this foundation works should the business add automated cost imports, call tracking, offline conversions, server-side tracking, and a dedicated data warehouse.

Connect advertising data to confirmed revenue

Describe your advertising channels, current CRM, and the customer journey from first contact to sale.

The Prodexa team will help you:

• define the minimum necessary architecture;
• identify where data is being lost;
• connect advertising, the website, and the CRM;
• build reporting based on customers and revenue;
• implement the system in stages without unnecessary infrastructure.