Connect Google Ads to the Leads and Sales in Your CRM

Google Ads can count a website enquiry without knowing whether it became a suitable lead or a paying customer. Offline conversion tracking fills in that missing part of the journey. Your CRM records a meaningful outcome, such as a qualified opportunity, and a supported connection sends the outcome and eligible matching information to Google.

Start with one clearly defined outcome and one reliable transfer route. For a new implementation, consider enhanced conversions for leads through Google Ads Data Manager or a supported CRM integration. Use a custom API when the business genuinely needs engineering control. Creating a conversion action alone does not connect the CRM, validate the data or prove that an outcome can be attributed to an ad.

Here is the simplest way to picture the job.

Follow one lead from the ad to the sale

This is a hypothetical example, with no real customer or advertising data. Cedar Commercial Maintenance receives an enquiry from a facilities manager after an ad click. The form submission tells Cedar that somebody wants a conversation. The salesperson later confirms that the property is in Cedar’s service area, the requested work is offered, and there is an active purchasing requirement.

That confirmation is Cedar’s qualified opportunity. Two weeks later, an accepted maintenance agreement becomes a separate contract accepted outcome.

Google does not need Cedar’s entire sales file. It needs the relevant event, the destination conversion action, the true outcome time, and supported information that can connect the outcome with an advertising interaction. An approved value can be included where useful. Internal notes explaining the customer’s building, negotiations or personal circumstances stay in the CRM.

The enquiry, qualification and contract are three milestones in one commercial journey. Adding them together would not reveal three customers. That distinction matters both when reading reports and when choosing what bidding should pursue.

For Cedar, the first implementation could measure qualification while the team establishes dependable contract reporting. Another business might already have clean sales records and begin with completed purchases. Choose an outcome your team can identify consistently and report promptly. A sophisticated connector cannot repair a stage that means something different to every salesperson.

Choose a route your team can maintain

There are several ways to transfer offline outcomes. Begin with the software and operating capacity you already have, then verify that the route supports the particular outcome you need.

Google maintains a direct-source catalogue. Check the conversion use case, rather than assuming that a product listed for audience uploads supports the same conversion workflow. Also distinguish a direct Google connection from a CRM vendor’s own advertising integration. They can have different triggers, backfill behaviour and supported fields.

A spreadsheet is reasonable when its contents are deliberate and someone owns the refresh. It becomes fragile when several people paste exports into it, rename columns or replace the file each week. A direct connector can reduce that manual work, but still needs a person who understands what its events mean.

The walkthrough below uses Google Sheets as the source and Data Manager as the connection. It is a documented configuration path checked against Google’s instructions on 9 October 2026. It is not a live-account demonstration: no CRM connection, tag deployment or conversion submission was tested for this guide. The fictional examples are for planning and must not be uploaded.

Existing integrations need a migration plan rather than a second producer switched on beside the first. List every system already sending the chosen outcome. Otherwise, a new connection can silently duplicate or change measurement that was already running.

Decide exactly when the outcome has happened

Write the qualification rule in language the sales team can use during an ordinary conversation. For Cedar, “service fit confirmed and active purchasing requirement recorded” is more useful than “good lead.” It identifies evidence a salesperson must establish.

Now choose the CRM object. A contact represents a person; an opportunity represents a potential piece of business. One facilities manager could request work at several properties. If Cedar qualifies opportunities, exporting the contact every time somebody edits its lifecycle stage may misrepresent the business.

For the first outcome, answer these questions:

• What observable event makes the record qualify?
• Which CRM record contains the evidence?
• Where is the first qualifying date and time stored?
• Can that same event happen again, and under what circumstances?
• Which conversion action will receive it?

Keep the answers short enough to inspect together. Cedar’s answer might be: “The first confirmed qualification of each opportunity creates one qualified-opportunity event. Reopening that opportunity does not create another. A genuinely separate request creates a new opportunity.”

Try the rule on a difficult case before configuring the connection. A prospect may be a perfect service fit but have no current requirement. Another may have an urgent need outside the service area. If the rule includes both without distinction, the eventual Google Ads signal will describe a different outcome from the one management expects.

Give each milestone its own time

The outcome time is when the milestone actually happened. It is not the time of the ad click, the export or the next day’s data import.

Suppose Cedar confirms qualification on Monday at 2:30 p.m. and the administrator exports the record on Tuesday morning. The event still belongs to Monday. Retain the timezone or an explicit offset, especially when sales staff work across Canadian provinces.

Use stage history or a dedicated first-qualified timestamp when available. A column labelled “last modified” can change because somebody corrected a postal code. Treating that edit as the conversion time would move an old outcome into a new reporting period. If the CRM has no trustworthy history, establish the field prospectively instead of inventing historical dates.

Decide what the value means

For an accepted contract, Cedar could use an approved contract-value definition. Finance should settle whether that means signed revenue, collected revenue or another consistent measure, including how taxes, cancellations and changes are handled.

For a qualified opportunity, the final sale value is still unknown. An estimated value needs a defensible business basis and maintenance. Do not enter a large round number just to make qualified leads look more important in a dashboard. It is acceptable to begin without a speculative per-lead monetary value and resolve the bidding strategy separately.

If the team cannot agree on its sales stages, fix that first. The CRM implementation guide covers the wider process. This article focuses on getting one dependable outcome from that process into advertising measurement.

Prepare the website and the source data

Enhanced conversions for leads combines eligible first-party customer information with offline outcomes. Google describes a website collection step and an import step using corresponding customer data. Retain genuine GCLIDs where available; Google says they remain required when a tag is not collecting user-provided data. A CRM email column alone does not establish a complete implementation. See Google’s explanation of enhanced conversions for leads.

Have the website owner confirm what happens when an enquiry is successfully submitted. Does the relevant tag receive the intended information? Does the CRM receive that same person’s record? Does the consent configuration behave as approved? A contact created manually later may have a different collection history from a website enquiry.

For a GCLID-based path, auto-tagging and reliable capture into the lead system matter. GCLIDs are case sensitive. Check that redirects, embedded forms and CRM transformations preserve the actual identifier; never substitute a campaign name or a UTM value. Google’s GCLID setup instructions describe the collection requirement.

If the site uses Google Tag Manager, the implementer should follow the current enhanced-conversions configuration guide. Preview mode can show whether the intended tag fires and receives the expected input before publishing. Agree on a permitted test procedure first, because form tests can transmit data. Inspect the successful submission path as well as errors, abandoned forms and repeat submissions.

An online enquiry event remains a separate measurement question. If that event is unreliable, use the GA4 lead-tracking guide alongside the website work. Do not relabel every form event as a qualified lead to make an offline setup appear complete.

Keep the transfer source small and deliberate

For the Sheets route, prepare a dedicated source containing only approved outcome records. Keep general CRM exports, staff comments and unrelated contact lists elsewhere. Restrict access to the people who maintain or review the connection.

The following is Cedar’s planning map, not Google upload headers and not a ready-to-import file. The actual fields must be mapped in the selected connection.

Google’s data-preparation reference lists supported fields and formats. For enhanced conversions for leads, map at least an eligible email or phone field. Map the actual conversion date and time. Use the supported order identifier where appropriate for event identity. Data Manager can normalize and hash relevant customer fields automatically, so do not add a second hashing step casually.

Check permission before connecting the sheet

Google’s customer data policies require appropriate disclosure of third-party advertising measurement and consent where legally required. They also restrict sensitive-category measurement and the information that can be uploaded. Hashing does not remove those obligations.

Have the responsible privacy or legal adviser determine the policy that applies to your business and customers. A newsletter subscription, a sales representative’s permission to call, and permission for advertising measurement are not interchangeable records. Keep the applicable disclosure, choice and time accessible internally, including later withdrawals.

For Cedar’s planning exercise, records with unresolved permission are held out until someone resolves them. That is a conservative operating choice for the example, not a universal legal determination. Do not change an empty consent field to “granted” to increase the number of eligible rows.

Connecting a source grants access and enables data sharing. Review the actual permission and policy screens with the authorized owner. The walkthrough describes where those steps occur; it does not mean a reader should accept unfamiliar terms or grant access without understanding the scope.

Walk through Google Sheets and Data Manager

This route assumes the business has settled the outcome, prepared the website collection, and approved the data use. Google’s current help pages show more than one conversion-creation layout. Follow the equivalent offline-source option in the account you actually manage; do not force an old screenshot’s labels onto a different interface.

1. Prepare one maintained source file

Create a dedicated Google Sheets file for the approved outcome export. Use a stable file name and stable column names. Put headers in the first row and the data in the first sheet: Google’s connector reads that first sheet. The person connecting it needs Editor access to the file and administrator access in Google Ads. See the Google Sheets connector requirements.

For Cedar, the first source would contain qualification events only. That is easier to review than a mixed export of enquiries, qualifications, appointments and contracts. If multiple outcomes share a source later, each connection needs an explicit filter so it selects the intended records.

Ask the administrator to show where tomorrow’s newly qualified opportunity will appear. If the answer is “someone will remember to paste it,” assign that person and a backup before proceeding. Automation is optional; dependable ownership is not.

2. Create or choose the intended conversion action

In Google Ads, open Goals → Conversions → Summary and begin creating the offline conversion action. Google’s Data Manager guidance describes the Import route through CRMs, files, or other data sources, while another current setup page uses Conversions offline. Choose the offline source and the category that reflects the milestone, such as Qualified lead or Converted lead.

Name Cedar’s action clearly, for example “CRM qualified opportunity.” Avoid a generic “Lead” name that could be confused with the online enquiry. Review the value, count and window settings against the outcome definition. Existing measurement needs a deliberate reuse or migration decision. Google’s Data Manager setup guide covers creating or associating the conversion action.

At this stage, confirm which account owns the action. A manager account used to administer campaigns is not automatically the conversion owner. If cross-account measurement is already in use, involve the person who maintains it before creating another action in a client account.

3. Connect the source for conversions

Select Connect a new data source, choose Google Sheets, and choose the direct connection. Select Conversions as the use case. Review the customer-data policy, use the authorized Google Drive account, and review the connector’s access request. Select the intended file.

Cedar is connecting a qualification source. An audience-list setup would serve a different purpose. Check the use case before proceeding, particularly if someone previously connected the same product for Customer Match.

If the file is already connected, inspect that connection rather than creating an unexplained duplicate. The important association is the source data plus the intended conversion action. A product appearing under Connected products does not by itself prove that the outcome has been assigned correctly.

4. Map the fields and inspect the meaning

Use the source-to-destination dropdowns to map the event information. Review every automatically suggested mapping. Similar column names can hide different meanings: “created_at” may mean contact creation, whereas Cedar needs the first-qualified time.

Work across a single approved source record with the sales administrator. Point to the qualification evidence, its timestamp and the matching information. Then inspect the corresponding mapped fields. This catches errors that a format check cannot: a date can be perfectly valid while representing the wrong business event.

Apply only transformations that have a clear purpose. For dates, inspect the timezone handling. A fallback timezone is appropriate only if it truthfully describes the source records. It should not conceal a sheet containing mixed, unknown local times. The general connection instructions cover mapping, optional filters and transformations.

5. Review the connection and its schedule

Name the connection so the source and outcome are identifiable. Review the chosen file, selected fields, filters, conversion action and schedule before finishing. Save the completed association, then return to the conversion action to confirm that a source is attached.

Google’s connection-management instructions provide manual, daily and weekly scheduling options, with daily as the maximum import frequency described there. Refresh the source before the scheduled import. For example, Cedar could prepare its approved export before the morning run; that timing is an operating example, not a prescribed Google schedule.

Avoid scheduling an import for the same instant the source is being replaced. Agree on how the source refresh signals completion. A run that succeeds against yesterday’s unchanged sheet can still leave today’s qualification events missing.

6. Inspect the first run and the conversion diagnostics

After an authorized run, open the connection’s run details and read any errors. Confirm that the selected source was read and that the expected eligible records were considered. Keep the source snapshot or an internal reference to it, so the team knows what the run attempted.

Then inspect the conversion action’s offline data quality from the Conversion summary page and open its diagnostics. Google’s offline data diagnostics show data-quality alerts and history. A successful connection and healthy formatting are useful evidence, but they do not establish that every CRM record belongs in Google-attributed reporting.

Finally, compare a defined, sufficiently mature group of outcomes with the appropriate report. The next section explains that comparison. Only after these checks can the team describe what its own implementation actually demonstrated. This guide’s documented walkthrough does not claim that those account-level steps have already succeeded.

Read the result without forcing the totals to match

There are three different questions after an import:

1. Did the source contain the right business outcomes?
2. Did the connection receive and process the intended records?
3. Which outcomes appear in Google Ads under the selected attribution and reporting settings?

A comparison becomes confusing when each person answers a different question. Sales may count every qualified opportunity from all channels. The integration owner may count rows processed today. The advertising report may assign conversions to earlier ad interactions. Those totals can differ without any arithmetic error.

Consider another hypothetical Cedar example. Its export contains 30 candidate rows. Two are repeated copies of existing qualification events, three are held because their measurement eligibility is unresolved, and one lacks a trustworthy qualification timestamp. That leaves 24 eligible unique outcomes: 30 = 2 + 3 + 1 + 24.

The first check is whether those six exceptions are explained and the remaining 24 are the intended events. It is too early to conclude that Google should report 24 conversions. Some may have no eligible Google advertising connection, and reporting settings can affect the result.

Compare the same outcome and date basis

Select the exact conversion action in the advertising report. Do not compare Cedar’s qualified opportunities against a total that also includes form submissions and telephone enquiries.

Google explains that standard conversion reporting is organized around the ad interaction date. The relevant by conv. time columns support a conversion-date comparison. Upload date is a different date basis. Timezone, processing delay and the One counting setting can also explain discrepancies. Google’s offline reporting discrepancy guide covers these distinctions.

In practice, choose a finished group of qualification events, record when you extracted it, and compare it with a report using the intended action and date basis. Refresh that same group after the applicable delay. This is more informative than comparing today’s changing CRM total with today’s changing advertising total.

Keep all-channel CRM outcomes visible in business reporting. Advertising attribution answers a narrower question. It does not replace the CRM’s record of a customer, and a missing attributed conversion does not mean that the sale never happened.

Allow time without ignoring a broken process

Check the documented timing for your route and identifiers. Google’s discrepancy guidance says processing can take longer for GBRAID- or WBRAID-keyed conversions, potentially up to 72 hours. That is a reporting expectation, not evidence that every unresolved discrepancy will fix itself after three days.

If the first scheduled run never happened, waiting for attribution will not help. If the source is empty, investigate the CRM export. If the source and run are correct but the reporting period is immature, allow the relevant delay before escalating a missing-result claim.

Track delays where they arise. A salesperson waiting a week to mark a lead qualified creates a business-process delay before Google receives anything. Improving that handoff can be more useful than repeatedly changing the connector.

Keep new measurement separate from bidding

It is possible to install a technically valid connection that teaches bidding to pursue the wrong outcome. Suppose Cedar imports every appointment booked, including reschedules and appointments outside its service area. The integration may operate exactly as configured while the business signal remains unsuitable.

Review a sample of the actual source outcomes with sales before making the new action a bidding objective. Ask whether the events represent the quality you intended and whether important categories are systematically missing. A signal that covers only one salesperson’s records may distort what the account appears to value.

Google distinguishes primary and secondary conversion actions. A primary action can be used for bidding when the relevant standard goal is selected; secondary actions are generally observational. There is an important exception: actions in a custom goal are used for bidding even when marked secondary. Review campaign goal selection as well as the action’s label. See Google’s primary and secondary action guidance.

A staged observation period can make sense, but its length should reflect the sales cycle, data volume and observed reliability. There is no universal number of days that makes an unreliable qualification definition safe. Keep the account’s existing measurement understandable during the transition and avoid changing the outcome definition, bidding objective and transfer route simultaneously without a reason.

If the new data confirms that many enquiries are unsuitable, use the lead-quality diagnosis guide to investigate demand, offers and follow-up. The import makes those outcomes available; it does not automatically solve the commercial cause.

Fix the first broken part of the journey

A useful support request identifies where the expected event stopped. “Google Ads is missing leads” is too broad. “These approved qualifications are present in the source, but this connection run rejected their date field” gives the implementer somewhere to begin.

For “Unknown click” errors, the population matters. Enhanced conversions can include eligible outcomes from other channels, so not every event will be attributed to Google. A widespread failure among outcomes expected to have an advertising connection deserves investigation. Start with website collection and corresponding source data, using the discrepancy guide rather than relabelling those leads.

Separate retrieval periods from attribution limits

A connector’s extraction period determines what it reads. An attribution or upload-age limit determines what can still be used. The conversion action’s window adds another condition. Passing one does not satisfy the others.

Google’s GCLID setup page describes a 90-day retrieval range for Sheets and several other file sources; some database sources use 14 days, while its Salesforce and HubSpot connections have their own initial and subsequent retrieval behaviour. Google’s offline import FAQ separately describes limits of 90 days for GCLID-based uploads and 63 days for enhanced-conversion uploads using personal information. Confirm the rules for the exact route and identifiers you use.

For Cedar, a salesperson holding an approved outcome in a personal spreadsheet may consume part of the available time before the export runs. Set a prompt, repeatable transfer process and alert on ageing records. If a sale falls outside the supported path, retain it in the CRM and business reporting. Never backdate the sale to make it eligible.

Handle changes to the underlying business event

A later loss does not always mean an earlier qualification was wrong. Cedar may correctly qualify an opportunity and then lose the contract to another supplier. If the measured event is qualification, that event still happened.

A duplicate record, an incorrectly qualified lead and a cancelled contract are different cases. Decide the appropriate treatment from the event definition and the transfer route’s actual correction support. Do not automatically upload a negative value, delete a row or create a new event and assume Google will interpret it as the intended reversal.

This is especially important before importing monetary values. Ask how an accepted contract’s value can be corrected and whether the integration can retract a mistaken conversion count. The answer belongs in the route-selection decision, not only in a support ticket after launch.

Developer appendix: build and migrate carefully

The configuration path above can be managed without writing an uploader. Custom implementations need additional controls. These notes explain the important design choices; they are not a complete provider schema or deployment script.

Check current API access before reusing old code

For a new custom build, begin with Data Manager API’s offline conversion documentation. Google’s deprecation register describes an access restriction effective 15 June 2026 for legacy offline uploads: developer tokens with no offline-upload requests between 17 December 2025 and 15 June 2026 are restricted from UploadClickConversions.

The same register now says developer tokens have been sunset in favour of the updated access model, while that historical restriction remains. This does not establish that every existing connector stopped working. It does mean an old Google Ads API script is not a safe starting assumption. Verify the application’s current authorized route and the vendor’s migration position. See the dated deprecation register.

Keep the application account, the account owning the conversion action and any manager used for access explicit. A migration can authenticate successfully while pointing to the wrong conversion owner.

Preserve event identity through retries

Create a stable internal key for the business milestone. Cedar could identify the first qualification of an opportunity with a reference such as DEMO-OPP-17-QUALIFIED-1. That is an illustrative internal reference, not a usable upload example.

The key should survive a repeated export, a retry and a contact merge. Do not generate a new key from the upload time. Otherwise, the same business event becomes a different event whenever delivery is uncertain.

Keep a small internal ledger with the destination, event key, source reference, payload revision and processing state. A timeout means the result is uncertain; it does not establish that the event was never received. Resolve that uncertainty through available diagnostics and the documented replay behaviour before creating another identity.

For Data Manager API, the event representation includes eventTimestamp, an eventSource and supported identifiers; a stable transactionId supports deduplication and corrections. The operating account must own the conversion action. Follow the offline event requirements, including encoding when user data is sent. The source worksheet in this guide is deliberately not that schema.

Assign preparation to one component

For a direct API implementation, apply the API’s current normalization, hashing and encoding rules before sending customer fields. Email and phone preparation differ, and Gmail-specific changes must not be applied to every domain. Phone numbers use E.164 before hashing. Preserve click identifiers exactly rather than passing them through email-cleaning code. Google’s formatting reference is the implementation authority.

State whether each input is raw, normalized or already hashed. A double-hashed email can look structurally plausible while failing to match. Tests should prove that the correct preparation occurs once and that an unexpected state is rejected.

Consent needs similarly explicit mapping. The API has distinct adUserData and adPersonalization signals, with granted, denied and unspecified states. These consent fields represent signals; they are not the evidence supporting the organization’s decision. Preserve that evidence internally and ensure a batch default cannot silently replace an individual restriction.

Understand adjustments before promising reversals

Data Manager API matches a transaction ID within the same conversion action to apply supported adjustments. Value and currency can be restated; matching information has specific supplementation limits, and other fields are not freely overwritten. An unmatched transaction ID can create a new conversion. Required event fields and identifiers still need to accompany an adjustment. See Google’s adjustment reference.

The API migration guide explicitly says Data Manager API does not support a retraction that removes both value and count. Restating value to zero leaves the count. A business that needs count reversals must evaluate that requirement before selecting or migrating its workflow.

Do not assume the Sheets connection exposes every API capability, or that a CRM vendor implements corrections the same way. Verify the selected route. Cedar’s operating rule for a lost opportunity, an invalid qualification and a cancelled agreement should be documented separately because each describes a different change.

Test what each layer can actually prove

Start locally with fictional records and no transport connection. Test a first qualification, an identical replay, a missing timestamp, an unknown destination, a denied eligibility decision, and a changed value. Keep synthetic markers blocked from all external transfer, including validation requests.

An authorized API implementation can later use validateOnly for basic request checks without ingestion. That still sends a request to Google and does not prove attribution. For real ingestion, retain the request ID and inspect destination-level results. Google’s diagnostics procedure starts status checks after 30 minutes, continues with backoff to a terminal result, and allows processing to take up to 24 hours. Reporting has its own timing.

The offline conversion planning kit contains design worksheets and a deliberately non-uploadable fictional exercise. Its local teaching checks illustrate business-rule handling only. They do not validate a live schema, consent system, actual matching or an end-to-end integration. Keep the kit separate from production exports and build an authorized implementation from current route documentation.

Know what “ready” should mean

A useful handover can explain one actual outcome all the way through: why it qualified, where its time came from, why the data was eligible, which connection handled it and what the available diagnostics showed. It also identifies who responds when the next run fails.

Ask sales to own the outcome definition, the CRM administrator to own the source, and the advertising or integration owner to own the destination and monitoring. One person can hold more than one role, but none should be assumed away. Review the handoff after changes to forms, CRM stages, source columns or agency access.

Begin with a meaningful event you can trust, then extend the measurement when the first route is dependable. Canada Create’s Google Ads management services can help connect that measurement to account decisions. Better feedback gives those decisions a firmer basis; it does not guarantee lower costs or more sales.

Frequently asked questions

What is offline conversion tracking in Google Ads?

It sends outcomes recorded in your CRM, such as a qualified opportunity or a closed sale, back to Google Ads. Google Ads can count a website enquiry without knowing whether it became a suitable lead or a paying customer, and this fills that gap. Start with one clearly defined outcome and one reliable transfer route.

What are enhanced conversions for leads?

Enhanced conversions for leads let Google match outcomes from your CRM to ad interactions using eligible, supported customer information. For a new setup, consider it through Google Ads Data Manager or a supported CRM integration. Send only the relevant event and matching information, and keep internal sales notes in the CRM.

Can my CRM send leads and sales back to Google Ads?

Often yes, through the CRM's documented Google Ads integration, a direct Data Manager connection, Google Sheets or a supported partner. Before committing, confirm that the route sends the specific contact, opportunity or transaction event you need and that your subscription includes it. Choose a route your team can actually maintain.

Is there a template for uploading offline conversions to Google Ads?

Google Sheets connected through Data Manager works when a modest, controlled export can be kept current. Prepare one maintained source file, create or choose the intended conversion action, connect the source, map the fields and review the schedule. Inspect the first run and the conversion diagnostics before trusting the numbers.

Why does Google Ads say my conversion action is not set up for uploads?

The upload is pointing at a conversion action that was not created to receive imported conversions. Create or choose an action intended for that source, then check that the connection maps outcomes to it. Creating an action alone does not connect the CRM or validate the data.

Which CRM milestones should I send to Google Ads as conversions?

Send each meaningful milestone as its own event, such as a qualified opportunity and an accepted agreement, with its true outcome time. An enquiry, a qualification and a sale are three steps in one journey, not three customers. Administrative edits like a changed phone number should never create a new conversion.

Why don't Google Ads conversions match the totals in my CRM?

Compare the same outcome on the same date basis first, since Google Ads and a CRM may record different times. Not every outcome can be matched to an ad interaction, so the totals will rarely line up exactly. Allow for delays, but investigate a gap that suggests a broken step.

Should bidding use offline conversions as soon as they start arriving?

Keep new measurement separate from bidding until the data has shown itself to be reliable. Watch several runs, check diagnostics and confirm the events mean what you defined. Only then decide whether automated bidding should pursue that outcome.