When someone taps the phone number on your Google search ad, they are making a decision before they have seen your website. They like your ad headline, maybe your reviews, and they want to talk now. When someone calls after spending 4 minutes on your site reading your about page, your service list, and your pricing - that is a completely different decision-making process. Both produced a phone call. Both ended up in your Google Ads conversion data as identical "Call" conversions. That is the problem.
Smart Bidding cannot tell those two calls apart if you have fed them into the same conversion action. It trains on a blend of two different intent levels, finds the average of both, and optimizes toward that average. The result is a campaign that never fully optimizes toward either.
How Each Call Type Is Triggered
The mechanics of each call type are different in ways that directly affect buyer intent.
Call Asset Calls (Formerly Call Extensions)
The phone number appears directly in your search ad in the SERP. The caller taps that number before ever visiting your website. By the time they dial, they have been exposed to your ad headline, your ad description, your review star rating, and your business name - nothing else. They have not seen your landing page, your proof elements, your service details, or your pricing. They are calling on the basis of an ad impression, which is a relatively thin information exposure.
Website Call Tracking Calls
The caller clicked your ad, landed on your website, stayed for some period - reading your content, evaluating your offer, reviewing your social proof - and then picked up the phone. Your landing page acted as a qualification filter. By the time they call, they know significantly more about you than a call-asset caller does. They have passed through an additional decision gate.
"The website call has been through a qualification filter. The call asset call has not. Treating them as equivalent in your conversion data is like treating a walk-in off the street and a referred client the same way in your CRM."
Net Profit Positive
These are not subtle differences in buyer behavior. They produce measurably different qualified rates, close rates, and average deal values in every service-business account we have audited. The intent gap is real, it is consistent, and it does not average out when you mix the two call types together.
The Intent Gap in Data
| Signal | Call Asset Call | Website Call |
|---|---|---|
| Pre-call exposure | Ad headline only | Full landing page |
| Avg call duration | 1 min 45 sec | 4 min 20 sec |
| Qualified rate (service businesses) | 38% | 64% |
| Close rate (from qualified) | 31% | 44% |
| Recommended Smart Bidding priority | Secondary (volume signal) | Primary (quality signal) |
| Best campaign types | Branded search, emergency keywords | Informational search, service comparison |
The close rate comparison is where the revenue impact becomes concrete. A 31% close rate on call-asset calls means you need roughly 3.2 qualified calls to close one job. A 44% close rate on website calls means you need 2.3 qualified calls to close one job. If you are paying $90 per qualified lead and your average job value is $900, the difference between those two close rates is $288 vs. $207 in cost per closed job - a 28% cost difference on the same budget, driven entirely by which call type you are optimizing toward.
Why One Conversion Action for Both Is Wrong
When both call types feed the same conversion action and that action is marked as primary in Smart Bidding, the algorithm trains on a blended average of two fundamentally different intent levels. It cannot learn which campaign types, keywords, or audiences drive high-quality website calls versus lower-quality ad-asset calls. The blended model produces a kind of systematic mediocrity - it optimizes toward the mean of both, which means it is not fully optimizing toward either one.
The most common symptom of a blended call conversion action is a campaign that consistently hits its CPA target but fails to generate proportional revenue growth. The CPA looks on target because there are enough call-asset calls counting as conversions to keep the numbers where they should be. The revenue does not follow because the model has learned to optimize toward the lower-quality half of the call mix. The dashboard shows a healthy campaign. The bank account does not agree.
This failure mode is particularly dangerous because it does not produce any obvious error or alert in Google Ads. You have to look outside the platform - at your CRM, your booking rate, your revenue data - to see the disconnect. Most accounts running blended call conversion actions have no process for making that comparison, which means the problem persists indefinitely.
The Numbers
How to Split Them
The fix is straightforward: create separate conversion actions for each call type and set their bidding priority correctly. This is a one-time setup that takes about 20 minutes in Google Ads.
Splitting Call Conversion Actions - Step by Step
In Google Ads: Goals > Conversions > + New conversion action --- CONVERSION ACTION 1 --- Type: Phone calls > Calls from ads using call assets Name: "Call - Ad Click (call asset)" Duration: Your business threshold (e.g., 2 min for most service businesses) Counting: One per click Bidding: START as Secondary (observe only for 30 days) --- CONVERSION ACTION 2 --- Type: Phone calls > Calls to a phone number on your website Name: "Call - Website Visit" Duration: Same threshold as above Counting: One per visit Bidding: Primary (used in Smart Bidding immediately) --- REVIEW STEP (Day 30) --- Pull the split performance report: - What is the qualified rate of each type? - What is the average call duration of each type? - What is the close rate (from CRM) for each type? If call-asset close rate is within 15% of website call close rate: Move call assets to Primary (the gap is small enough to blend) If call-asset close rate is more than 15% below website call close rate: Keep call assets as Secondary Use them for audience insights only, not bidding signal
The 15% rule is a practical duration threshold guideline, not a hard law. Some businesses - particularly those in emergency service categories like locksmith or urgent plumbing, where call-asset callers are highly motivated - will find the gap is small and both types can run as primary. Most service businesses with a research phase in their buying process will find the gap is larger and should keep website calls as their primary signal.
Bidding Strategy Per Call Type
Call-asset calls are a high-volume, lower-qualification signal. The right use is not to discard them - it is to use them for what they are actually good at. They tell you which keywords and ad copy variations generate any phone engagement at all. You can use call-asset volume to identify keywords worth investing in, then use website call quality data to determine which of those keywords actually generate profitable calls. The two signals are complementary when used correctly. The problem is only when they are blended into one bidding signal that obscures the quality difference.
Practically, this means setting up audience lists from your call-asset converters and using those lists for observation in your campaigns. Over time, you will see whether call-asset callers who do not convert immediately come back and convert later via website call - which would indicate a longer consideration cycle, not a low-intent caller. That kind of cross-signal analysis is only possible when the two call types are tracked separately.
Call Quality Comparison
Call Asset vs. Website Call - Quality Metrics Comparison
Service businesses - qualified rate, close rate, and cost per revenue-generating call
Law Firm - Single Blended Call Conversion Action
A personal injury law firm running $22,000 per month on Google Ads, with a single "Phone Call" conversion action capturing both call-asset calls and website calls. Target CPA was set at $120. Smart Bidding was consistently hitting $115 CPA. Revenue from Google leads had plateaued for 8 months despite increasing the budget.
Audit revealed: call-asset calls had a 31% qualified rate and 28% close rate from qualified. Website calls had a 71% qualified rate and 47% close rate from qualified. Smart Bidding had been optimizing toward the blended average of both - finding plenty of call-asset calls to hit the CPA target while missing the higher-quality website call audience that would have driven revenue.
Fix: call assets moved to secondary, website calls to primary. Conversion volume dropped 40% immediately. Learning phase ran 45 days.
Free Calculator
Calculate the cost of your blended call signal
Enter your current call volume split, close rates by call type, and average job value. See exactly how much the blended model is costing you versus splitting the actions correctly.