Marketing

End-to-end analytics: linking ads and requests

Ad platforms know about clicks, the CRM knows about deals, and there's usually no link between them. What needs to be prepared for that report to become possible.

The ad platform reports on clicks, the website analytics on requests, the CRM on deals and money. Each system is right about its own part and knows nothing about the others. End-to-end analytics is an attempt to join three reports into one, so you can see which campaign brought a paid order and which only spent the budget.

Why join the data at all

Without a link, budget decisions are made on cost per request. That's better than nothing, but requests aren't equal. One channel brings people who buy quickly and spend a lot; another brings people who ask about the price and disappear. By cost per request these channels can look the same, while in money they differ fundamentally.

Linked data answers questions that otherwise remain guesses:

  • which campaigns bring deals, not just enquiries;
  • which keywords and placements consistently produce empty requests;
  • how much time on average passes from first visit to deal in your segment;
  • which channel brings clients who come back again.

It's important to understand the limitation: end-to-end analytics doesn't improve advertising by itself. It removes the need to guess and makes visible the part of the journey where money is usually lost.

How the data is joined technically

The whole setup rests on being able to link a CRM record unambiguously to a website visit, and a visit to an ad click. That requires three things.

Tags in links

UTM tags are the base layer. They pass the source, campaign, ad and keyword right in the page URL. The rules are simple: tag all paid links without exception, stick to one spelling (google and Google are two different sources in a report) and don't change the naming scheme retroactively.

Separately, there are click identifiers that ad systems add to links themselves. They're more precise than tags because the platform issues them, and they're usually what's used to send conversion data back to the ad platform.

The visitor identifier

Analytics tools assign the browser their own identifier. It needs to be saved in a hidden form field and sent to the CRM along with the request. Then the deal gets a key by which it can be found in the website data.

This is the most commonly skipped part. The form sends a name, phone number and comment — and that's it. At that moment the link between the ad and the future deal is broken for good, and there's nothing to restore it with later.

Fields in the CRM

The deal record should store the source, campaign, visitor identifier and date of first contact. In separate fields, not in the text of a note: a report can't be built from notes.

If records are created by hand, the fields will be filled in only some of the time. So a request from the website should reach the CRM automatically — through an integration, not by copying from email.

Where the data breaks

There are several places where the chain almost always breaks, and they're worth accounting for in advance.

Calls. A person looked at the website and dialled the number. Analytics shows a visit without a goal; the CRM shows a deal without a source. The fix is number substitution: a call tracking service shows different visitors different numbers and links the call to the visit. It's a separate service with a subscription fee, and it makes sense when calls are a noticeable share of enquiries.

Messengers. Moving to a messenger takes the person off the site, and the conversation then lives in a separate app. A partial fix is links with a parameter inserted into the first message, or taking enquiries through a bot that passes data to the CRM.

Offline deals. A client came into the office, gave their name, and a manager created the record by hand. There's nothing to link such a deal to automatically. Only organisational measures work here: a mandatory "how did you hear about us" field and discipline in filling it in.

A long cycle. If a decision takes months, devices, browsers and even contact people can change between the first visit and the deal. Part of the journey is inevitably lost.

End-to-end analytics is never 100% accurate. Its value is in comparing channels with each other, not in absolute values.

What to prepare before implementation

The tool joins what already exists. If the source data is incomplete, the report will be neatly formatted confusion. Before launch, several questions need to be settled.

  1. Define the deal stages. The CRM needs a clear pipeline with fixed stages and a clear sign of success. If half the deals have sat "in progress" for years, conversion can't be calculated.
  2. Agree on what counts as a lead. Spam, mistaken enquiries and requests from contractors should have a separate status; otherwise they end up in the denominator of every calculation.
  3. Get costs in order. You need actual spend per channel, including contractor fees and content production, not just charges from the ad platform.
  4. Check that requests reach the CRM automatically at all. Without this, there's no point going further.
  5. Assign someone responsible for filling in fields. Automation covers the website and ads, but not a manager's conversation with a client.

A practical sequence

A sensible sequence looks like this.

  • Step one: tag all ad links and make sure the tags reach the page and aren't lost on redirects.
  • Step two: add hidden fields with the visitor identifier and tag to forms, and set up their transfer to the CRM.
  • Step three: set up automatic deal creation from requests and mandatory source fields.
  • Step four: build the first report — deals and revenue broken down by source and campaign. After that you can add call tracking and sending conversions back to ad platforms.
  • Step five: check the report against reality. Pick a few closed deals and trace their path by hand. Discrepancies at this stage are normal; they need to be found before the report is used to allocate budget.

The main sign the system is built correctly: the report is opened not once a quarter for a presentation but weekly, to reallocate money. If that isn't happening, the link doesn't add up somewhere. If you'd like to discuss what applies to your project, write to us — or see how our analytics work is organised.