Tracking calls and getting them into Google Ads are two different jobs. The first tells you what happened; the second changes what happens next, because imported conversions are what your reporting ranks campaigns by and what Smart Bidding learns from. This guide covers the second job: why importing beats counting everything, the three transport methods and how to choose, value imports, and the troubleshooting table for when uploads bounce.
Prerequisites: call tracking capturing GCLIDs per session (the setup guide gets you there) and a working grasp of the click-ID mechanic (two-minute version in Offline Conversion Tracking).
Why import calls instead of counting them all
The naive configuration — every call becomes a conversion — actively damages accounts. Spam calls, solicitors, wrong numbers, and existing customers all become “conversions” the bidding algorithm dutifully learns to buy more of. Importing means choosing what crosses the line into Google Ads:
- A filtered subset — calls over a duration threshold, first-time callers only, spam-screened.
- A qualified subset — calls a scoring system or human marked as real opportunities.
- A valued subset — qualified calls carrying dollar values for tROAS bidding.
Each step up the ladder makes the imported signal scarcer and truer, and the bidding results follow the signal quality. Import architecture is where that ladder gets built.
Method 1: native platform integrations
What it is: your call tracking platform’s built-in Google Ads connection — authorize the account, map calls to a conversion action, set the sending rules, done.
Strengths: automated, near-real-time-ish, no maintenance, and rule changes (say, moving from “60+ seconds” to “scored qualified”) happen in the platform UI without touching pipes. This is the right default for the large majority of accounts.
Limits: you send what the integration supports. Most handle filtered and scored calls plus values cleanly; unusual conversion logic (multi-touch dedup rules, exotic value formulas, CRM-side qualification) can exceed what the checkbox settings express.
Choose it when: your qualification logic lives in the call platform. Which is most accounts, most of the time.
Method 2: scheduled Google Sheets imports
What it is: Google Ads can ingest conversions from a Google Sheet on a schedule. Your side of the bargain: keep a sheet populated with rows of GCLID, conversion name, conversion time, value, currency — from a call platform export, a CRM report, or a script.
Strengths: total control of the logic with zero API programming. Anything you can compute into a row, you can import — including CRM-qualified calls, deduplicated cross-system conversions, or human-reviewed weekly batches. Wonderfully debuggable: the sheet is the log.
Limits: batch cadence (your schedule, not real time), format discipline (timestamp and timezone errors are this method’s signature failure), and the operational fragility of anything that depends on a spreadsheet staying healthy.
Choose it when: qualification happens outside the call platform (CRM review, sales feedback) or you need custom logic without engineering investment.
Method 3: the Google Ads API
What it is: direct programmatic conversion uploads from your systems.
Strengths: everything — real-time-capable, arbitrary logic, enhanced matching options, fully integrated into your data infrastructure.
Limits: it’s software. You’re signing up for development, credentials management, error handling, monitoring, and ownership when it breaks at 2am before a board meeting.
Choose it when: you have engineering resources, high volume, and requirements the first two methods genuinely can’t express — commonly as the mature endpoint of a closed-loop CRM architecture. If you’re unsure whether you need the API, you don’t yet.
Choosing: the decision in one paragraph
Start with Method 1 and stay there while it expresses your logic. Add Method 2 the day qualification moves outside the call platform — typically when CRM outcomes enter the picture. Graduate to Method 3 when volume and complexity make the sheet feel like duct tape. Methods coexist fine — many accounts run native integration for fast qualified-call signal and a scheduled import for slower CRM-confirmed revenue — just point them at separate, clearly named conversion actions so nothing double-counts.
Importing values for value-based bidding
All three methods carry values; the question is which values.
- Flat average (every qualified call = your average job value): one afternoon of setup, immediate improvement over valueless counting.
- Tiered by call type (emergency service ≠ estimate request ≠ parts question): your scoring taxonomy assigns the tier.
- Actual revenue (CRM closed-won amounts, imported late but true): the tROAS endgame.
Two rules keep value imports honest. Match the conversion action’s counting settings to the value logic (a revenue action should count “every,” not “one”). And when real revenue arrives weeks after the click, lean on the qualified-call action for timely signal while the revenue action carries the truth — the timing interplay is worked through in How to Calculate True ROAS.
Troubleshooting rejected uploads
The greatest hits, with fixes:
| Symptom | Likely cause | Fix |
|---|---|---|
| “Click not found” rejections | GCLID malformed, truncated, or from another account | Verify auto-tagging; check redirects aren’t stripping URL params; confirm account match |
| Conversions rejected as too old | Upload after the click-lookback window | Import a faster milestone (qualified call) alongside slow revenue; verify current window limits |
| “Conversion time before click time” | Timezone bug | Standardize timestamps with explicit timezone at the source |
| Imports land but in the wrong bucket | Conversion action renamed/duplicated | Freeze conversion action names; change deliberately and rarely |
| Volume looks low for recent days | Normal — imports post to click date | Compare complete weeks only; resist “fixing” it |
| Everything silently stopped | Expired authorization on the integration | Re-auth; add a weekly volume glance so silence gets noticed in days, not months |
That last row deserves the emphasis: import pipelines fail quietly. One recurring five-minute check — “did last week’s conversions arrive in expected volume?” — is the entire monitoring program most accounts need.
Frequently asked questions
Can I import only qualified calls?
Yes — that’s the recommended architecture. Send qualified calls (platform-scored or CRM-confirmed) as your primary conversion action for bidding, and keep raw call counts, if you track them at all, as a secondary observation-only action.
How often should imports run?
Native integrations: leave them on their automatic cadence. Sheets: daily is the standard; weekly is acceptable for low volume. What matters more than frequency is beating the lookback window and keeping the cadence consistent so week-over-week comparisons mean something.
Why were my conversions rejected?
Overwhelmingly one of four: bad/missing GCLIDs, uploads outside the click window, timestamp/timezone errors, or conversion-action name mismatches. The table above maps symptom to fix; the GCLID chain test in the offline conversion guide isolates capture problems from upload problems.
Imports flowing? The next lever is what you feed them: call scoring to raise signal quality, then Smart Bidding With Call Conversions to cash it in.