Why closed-loop reporting matters
A lead-generation business pays for clicks and is paid for customers, and the two are separated by days or months and by a hand-off between teams. Everything the ad platforms and analytics measure sits on the near side of that gap. They can report the cost of a form submission with precision and the value of a customer not at all, because the customer is created in the CRM, outside their view.
The consequences are practical. Automated bidding optimises towards whatever conversion it is given; given form fills, it finds the cheapest form fills, which are rarely the best customers. Budget is allocated on cost per lead, so a channel that produces many poor leads looks better than one that produces few good ones. And cost per acquisition (CACCost to acquire one new customer.) is reported at the form, where it is a guess, rather than at the won deal, where it is a fact.
Closing the loop means the CRM tells the platforms which leads became qualified, which became customers, and what they were worth, keyed to the click that started it. The platforms can then optimise towards qualified leads or revenue, reports can show cost per customer by campaign, and attribution can be run on outcomes. None of it is exotic: every major platform documents a way to receive outcomes from a CRM, and the pattern is the same in each case. Capture an identifier at the form; store it; send it back with the outcome.
The CRM for marketers hub places this guide beside the fundamentals and pipeline guides it depends on. The outcomes sent back are stage changes, so the stages in the pipeline guide have to be defined first; the identifiers are stored in source fields, so the fields described in the CRM fundamentals guide have to exist.
Capturing identifiers at the form
The join between a click and a customer is an identifier captured on the landing page and written into the CRM record at creation. Which identifiers depends on which platforms need to receive outcomes later.
| Identifier | Where it comes from | How it is captured | Who uses it later |
|---|---|---|---|
gclid | Appended to the landing URL by Google Ads auto-tagging | Read from the URL, stored in a first-party cookie, written to a hidden form field | Google Ads offline conversion import |
| Other Google click identifiers | Google documents further identifiers for some device and browser contexts | Same as gclid; store whichever is present | Google Ads offline conversion import |
fbclid and the browser cookies Meta documents | Appended by Meta ads; the pixel writes click and browser cookies | Read and stored with the form submission | Meta Conversions API event matching |
| Campaign parameters | Set by the marketer on every tracked link | Read from the URL, persisted for the session, written to hidden fields | CRM source fields; analytics channel reporting |
| Analytics client identifier | Set by the Google Analytics tag in a first-party cookie | Read from the cookie, written to a hidden field | Google Analytics offline event import and Measurement Protocol |
| Hashed email and phone | The form fields themselves, hashed before sending | Captured by the conversion tag at submission, as the help centres document | Enhanced conversions for leads; Conversions API matching |
| Landing page, referrer, form name, timestamp | The page and the submission | Hidden fields | Diagnosis when a parameter is missing or ambiguous |
| Consent state | The consent banner's stored choice | Written with the submission | Deciding what may be sent back at all |
Two capture problems account for most failures. The first is persistence: a person lands on one page with the identifier in the URL and converts on another without it, so the identifier has to be stored in a first-party cookie or local storage on arrival and read back at submission, with a storage lifetime at least as long as the platform's documented click window. The second is forms that do not pass hidden fieldsA form field the visitor never sees that records values such as source, campaign and landing page, so each lead arrives in the CRM with its origin. through: embedded, third-party and multi-step forms often strip unknown fields, and the fix is usually a form setting or a small script, verified by submitting a test lead and reading the record. When records arrive with no identifier and no parameters, the diagnosis for a blank or wrong CRM lead source traces each sourceless group to the form or capture step that lost them. The same capture is what makes a comparison of in-platform forms and landing pages readable, as in the native lead form test, where every lead must carry its test arm into the CRM. When the whole site is replaced and not a single form, the tutorial on moving a website without breaking the CRM puts every form, script and rule that depends on this capture in order.
Campaign parameters (UTMURL tags identifying traffic source. values) follow the same path and are the fallback when a click identifier is missing. The naming conventions guide sets out how to structure them so that the campaign on the lead matches the campaign in the ad account without a translation table.
Offline conversion imports
An offline conversion import is the oldest of the patterns and the simplest to explain: the CRM sends the platform a list of click identifiers with the name of the conversion that happened, when it happened and, optionally, what it was worth. The platform matches each identifier to the click it issued and records the conversion against that click's campaign, ad group, keyword and audience.
Google Ads documents this as an import conversion action whose source is a CRM or other offline system. The
documented flow is: create the conversion action with the import source; make sure auto-taggingA Google Ads setting that adds a click identifier (GCLID) to every ad click's landing URL. is on so that
gclid is appended to landing URLs; store the identifier on the lead; and upload the outcomes, with the
identifier, the conversion name, the conversion time and an optional value and currency, by file upload,
scheduled upload from a spreadsheet or a file location, through the API, or through one of the CRM
integrations Google lists. Google documents that the conversion time must be after the click and within the
conversion action's window, and that the time zone has to be stated correctly; the help pages in the sources
give the current formats and limits, which this guide does not restate. The
conversion tracking setup tutorial covers creating
actions and tags in the account.
Google Analytics (GA4Google's event-based analytics platform.) has two documented routes for the same job. Data Import lets you upload offline events keyed to the client identifier captured at the form, and the Measurement Protocol lets a server send events directly to the property, which Google documents as a supplement to tag-based collection rather than a replacement for it. In either case the imported event can be marked as a key event; the key events guide covers how key events are chosen and how they flow on to Google Ads.
Meta's route is the Conversions APIServer-side event sharing with Meta., covered in the next section, since Meta documents CRM outcomes as server events rather than as a separate import.
Which stage to send is a decision, not a default. Sending every form fill duplicates what the tag already reports. The useful outcomes are the ones the platform cannot see: a lead accepted by sales, a lead qualified after a conversation, a deal won, and the won amount. Many teams send two: a qualified-lead conversion used for bidding, because it arrives soon enough to be useful, and a won-deal conversion used for reporting, because it arrives too late and too rarely to steer bids on its own. The won-deal upload is one action among several that should fire together when a deal closes, and the tutorial on building a closed-won workflow puts it in order with the customer stage, prospect suppression and the customer exclusion lists. Where outcomes go up by file instead, the offline conversion upload SOP runs each upload as a routine, from the consent filter and the import results to the count check against the CRM.
Enhanced conversions and Conversions API from the CRM
Click identifiers fail in predictable ways: the parameter is stripped, the cookie expires, the person converts on a different device, or the lead arrives by phone. The platforms' newer mechanisms match on hashedConverting an identifier such as an email address into a fixed string with a one-way function, so platforms can match records without the raw value. first-party data instead, and both are documented as working from a CRM.
Google documents enhanced conversions for leads as follows: at form submission, the conversion tag captures the lead's email address or phone number, hashes it, and sends it to Google; later, when the lead reaches the outcome stage, the CRM uploads that outcome keyed by the same hashed data rather than by a click identifier, and Google matches it to the ad interaction on its side. The upload uses the same import mechanisms as the previous section. The enhanced conversions guide explains the web and leads variants, how the hashing works and how to validate that uploads are matching; this guide only notes that the CRM's part is to hold the same email or phone the tag captured, unchanged, and to upload the outcome with it.
Meta documents the Conversions API as a server-side channel for events, including events that happen in a CRM after a lead is created. The CRM, or an integration in front of it, sends a lead event with the outcome, the customer information parameters Meta documents (hashed email and phone among them) and the click and browser identifiers captured at the form, so that Meta can match the event to the ad. Where the same event was also sent by the pixel, Meta documents deduplication by a shared event identifier; CRM stage changes that the pixel never saw have nothing to deduplicate. The Conversions API guide covers the event flow, deduplication and verification, and the pixel setup tutorial covers the browser side it pairs with.
Three things are common to both mechanisms. The data has to be captured at the form, not reconstructed later from whatever the CRM holds, because the match depends on the same value being seen at both ends. It has to be normalised the way the platform documents before hashing, or the hashes will not match. And consent has to permit it, which the final section covers.
Closed-loop reporting
With outcomes flowing back, the platforms can show qualified leads and won deals by campaign, and bidding can target them. That is the operational loop. The reporting loop runs the other way: spend from the platforms joined to outcomes in the CRM, so that cost per customer and return by campaign can be read in one place, with the business's own stage definitions rather than the platform's attribution rules.
The report is a table with one row per campaign (or source, or ad group) and columns for spend, leads, qualified leads, won deals, revenue and the derived rates and costs. Spend comes from the platforms, usually through a connector or an export; everything else comes from the CRM, grouped by the campaign value in the source fields. The join keyThe shared field, such as date and channel, a campaign ID or a click ID, on which rows from two data sources are matched in a blend or a query. is the campaign name, which is why naming has to match. The report lives wherever the business already keeps its reporting; the attribution section covers the models that can be applied once the touches and outcomes are in one place.
Reconciliation is part of the routine. The platform's count of imported conversions and the CRM's count of outcomes sent will not match exactly, because some uploads fail to match a click, some fall outside the window, and some are deduplicated. A monthly comparison, with the unmatched share tracked over time, catches a broken capture before it costs a quarter of data. The platform versus analytics discrepancies guide and the Google Ads versus analytics numbers guide explain why the systems disagree and which differences are normal.
A commercial insurance broker closes the loop
A commercial insurance broker captures gclid, campaign parameters and a hashed email on its quote form and
stores them on the contact. Two conversion actions are created in Google Ads: "qualified quote", uploaded
nightly when sales marks a lead as qualified, and "policy bound", uploaded with the premium when a policy is
issued. In the first month 1,140 qualified quotes are uploaded and 1,020 match a click; of the 120 unmatched,
most arrive from a landing page the agency built without the cookie script, which is fixed in a week. Bidding
is switched to target the qualified-quote action. Three months later the account reports 310 bound policies
from paid search at an average cost of £265 each, against a form-level cost per lead of £38 that had
previously been the only number anyone looked at; two campaigns with the cheapest leads turn out to have the
most expensive policies and are cut, and the budget moves to a campaign whose leads had looked expensive at
the form.
Consent and data minimisation
Everything above moves personal data between systems, and the rules that govern that vary by jurisdiction; this section is general guidance, not legal advice. The practical principles are the same ones the privacy and consent guide sets out for on-site measurement.
Capture the consent state with the lead and respect it when sending outcomes. Where a person declined the relevant purpose at the banner, the identifiers and hashed data for that person should not be sent to the ad platforms, and the integration has to check the stored state rather than assume it. The platforms document their own consent requirements for the mechanisms described here, and the help pages in the sources are the authority on what they currently require.
Send the minimum. An outcome upload needs an identifier or a hashed match key, a conversion name, a time and optionally a value; it does not need the lead's name, job title, company or notes, and the integration should not send them because a field was available. The same applies to the analytics side: an imported event needs a client identifier and an event name, not the contents of the CRM record.
Hash where the platform documents hashing, with the normalisation the platform documents, and never send raw contact details where a hashed value is what the mechanism expects. Keep the stored identifiers only as long as they are useful: the platforms' click windows bound how long a click identifier can be matched, and a retention rule that deletes identifiers after that period is both tidier and easier to defend.
Document the flows. A one-page description of which fields leave the CRM, for which purpose, to which platform, under which consent condition, is the artefact a privacy review will ask for and the one a new team member needs. It is also the first place to look when an upload stops matching.
To decide which CRM stages, identifiers and imports a programme needs before connecting anything, see the measurement plan for lead generation, which writes them into one specification.
Common mistakes
- Sending form fills back to the platform. The tag already reports them; the loop adds nothing until a stage the platform cannot see is sent.
- Identifiers lost between landing and submission. Persist them on arrival in first-party storage and read them back at the form; test with a real click.
- Forms that drop hidden fields. Embedded and multi-step forms often strip them; submit a test lead and read the record before trusting the integration.
- Campaign names that do not match. The join between spend and outcomes is the campaign name; mismatched naming forces a translation table that nobody maintains.
- Uploading outside the window or with the wrong time zone. Google documents both as reasons an import is rejected; check the help page for the current rules rather than guessing.
- Reconstructing match keys from the CRM. The hashed email has to be the one captured at the form, normalised the documented way, or it will not match.
- Ignoring consent state on the way out. A declined purpose on the banner has to stop the outcome upload for that person.
- No reconciliation routine. Unmatched uploads climb quietly; compare counts monthly and track the unmatched share.
The CRM and marketing automation hub places this guide in the wider stack, and the conversion tracking expertise page describes how the author approaches the whole measurement layer.
