Leads Are Not Customers: Why Lead Volume Is Quietly Costing Lead-Gen Businesses Money

Updated: 6 hours ago
The short answer: lead quantity counts form fills, not customers. When Google Ads bids to raw leads, it buys more of whatever is cheapest to submit. The fix is to separate four numbers (leads, qualified leads, accepted leads and customers) and make the earliest commercially meaningful event, usually a qualified or accepted lead that arrives within about seven days, your single primary conversion.
There is a particular kind of meeting I have sat in many times. The paid media report is green. Cost per lead is down, conversion volume is up, and the platform says return on ad spend has never been better. Then somebody from sales, usually the person who has to phone the leads, says quietly that it does not feel like a good month.
Both sides are telling the truth. The account is doing exactly what it was built to do. The problem is what it was built to do.
At Google's Think Leads EMEA conference in Dublin on November 3–4, 2026, Google's own lead generation product team spent two days making this point to a room of senior lead-gen marketers. There was a clear message, understand your KPIs first, set the right tracking and signals in place second, and lastly build your account around these KPIs to get the most out of Googles capabilities.
How a well-run account learns the wrong lesson
Smart Bidding, Performance Max, AI Max for Search and Demand Gen share one trait: they get very good at producing whatever event you mark as the goal. If the goal is a form submission, the system learns which people, queries, placements and times of day produce form submissions, and it buys more of them.
That sounds fine until you remember what a form submission actually measures. It measures that someone typed something into a box. Not that the phone number connects. Not that they live in the area you serve. Not that they can afford the product, that they make the decision, or that they ever intended to buy.
The opening session in Dublin made this clear. The host asked the audience, a room full of marketers, how many had given a throwaway or secondary email address on a form. A large share of hands went up. If the people who build lead funnels for a living do it, your prospects do too.
If the only signal you pass back is the form submit, you get plenty of leads, but no notion of downstream quality.
That is a close paraphrase of Google's LeadGen product lead in his keynote. It is an unusual thing for a platform to say about its own product, and it is correct. The algorithm is not failing. It is succeeding at the wrong task.

Four numbers that are not the same number
Most lead-gen reporting collapses four different measures into one word. Pulling them apart is the first step, because each has a different owner, a different delay and a different effect on bidding if you optimize to it.
Leads: A form submit, call or message. Owner: Marketing. Typical delay: instant. If bidding optimizes to it: volume rises; quality drifts toward whatever is cheapest.
Qualified leads: Meets your criteria: reachable, in your area, right need, right timeframe. Owner: Marketing and sales operations. Typical delay: minutes to days. If bidding optimizes to it: volume and quality balance; the usual sweet spot.
Accepted leads: Sales, a partner or a client has taken it on. Owner: Sales. Typical delay: hours to a week. If bidding optimizes to it: strong commercial signal, often fast enough to bid on.
Customers: Revenue has been booked. Owner: Finance. Typical delay: weeks to months. If bidding optimizes to it: sparse, slow learning; volume often collapses.
The gap between the first and the last of these is where lead-gen businesses quietly lose money. A campaign can double leads while customers stay flat, and every platform metric will say the campaign improved.
Why ROAS does not rescue you
The usual answer to "cost per lead is misleading" is "move to ROAS." Return on ad spend feels commercial because it has revenue in it. But the revenue the platform sees is whatever value you send it.
If every form submit is sent as worth $1, your ROAS is a conversion rate in disguise. If every lead is sent at the same estimated average value, the bidder still cannot tell a good lead from a bad one. ROAS also knows nothing about your margin, your capacity to serve the customer, or whether the customer stayed.
Value-based bidding is the right destination, but it only works when the values describe reality. Facile.it, the Italian price comparison business, told the Dublin audience it moved from Target CPA to Target ROAS using a predictive value model, and that it chose "actionable and understandable" over statistically perfect. That is the right instinct. The value does not need to be perfect; it needs to be true enough to rank one lead above another.
The one-primary-goal rule
The most useful practical guidance at the conference came from Google's search product team, in three steps:
Track the key online and offline touchpoints in Google Ads.
Categorize every conversion action correctly in the standard goal categories.
Optimize to one primary lower-funnel goal aligned to the real business KPI, ideally with a conversion delay of seven days or less.
Two details from the Q&A matter more than they look. First, if you run several primary conversions at different funnel stages, say a form, a call and an appointment, Smart Bidding builds one blended predicted conversion rate across all of them. It ends up bidding to an average of events that are not worth the same. Second, re-categorizing conversion actions does not trigger a new learning period, so cleaning this up is cheap.
Many accounts carry several primaries because someone believed more signals meant more learning. In practice it means a blurrier target. Keep one primary, set the rest as secondary, and park duplicates in "Other."
The common fix is wrong: do not bid to the sale
Once people accept that leads are not customers, the instinct is to jump to the far end of the funnel and bid to revenue. For most lead-gen businesses, that is a mistake too.
Sales arrive weeks or months after the click, and in small numbers. A bidder with twelve sales a month to learn from learns slowly and erratically. Volume drops, cost per sale looks unstable, somebody panics, and the account is switched back to form submits. I would rather a business never tried value bidding than tried it that way, because the failure teaches the organization the wrong lesson.
There is almost always an event in between: a verified, qualified inquiry; a call that lasted long enough to matter; a lead accepted by sales; an appointment booked. It is commercially real, it happens within days, and there is enough of it for the system to learn from.
Bid to the earliest event that is commercially meaningful, not the latest event you can measure. Record the sale as a secondary conversion and use it as a guardrail. If the sale rate falls while the primary goal improves, the signal needs fixing, not the target.

A worked example
Take a residential solar installer. A homeowner submits a quote request. Within a day, the sales team confirms the number works, the property is owned, the roof is suitable and the ZIP code is served. Within a week, a home survey is booked. The installation, and the revenue, arrives six to ten weeks later.
Bidding to the quote request buys renters and people who were curious about the price. Bidding to the installation gives the system a handful of signals a month, two months late. Bidding to the qualified inquiry, or the booked survey, gives it a commercially meaningful event inside the seven-day window. You can then add value on top: a survey for a larger system can carry a higher value than one for a small system. You do not need a lifetime value model to start. One honest attribute sent back quickly beats a sophisticated model sent back next quarter.
Engineer quality at the point of capture
The other half of the fix happens before the lead reaches your CRM. Google's lead format team described the quality controls now available on Google-hosted lead forms: one-time passcode (OTP) phone verification, more than 130,000 pre-approved qualifying questions with new ones approved in two to three days, conditional answers, and a built-in conversion for responses you mark as qualified. As of November 2026, Google's help pages confirm that qualifying responses and phone verification are available on Search campaigns only.
That last feature changes the economics. You can bid to the qualified response rather than the raw submit. Ask the two or three questions your sales team would ask on the first call, and let a verified phone number filter out the rest. Two cautions: these forms are billed per click, not per lead, so unqualified clicks still cost money; and frictionless forms can attract lower-intent users, so judge them on qualified volume, not total leads.
Google also reported that invalid-lead exposure has fallen 30% this year and that it has removed more than five trillion invalid interactions. Useful, but filtering obvious fraud is not the same as filtering low intent. That job is still yours.
What to tell finance
Finance teams have learned to distrust platform metrics, and they are right to. The fastest way to earn credibility is to report the four numbers side by side, every month, from your own systems rather than the ad platform's.
Lead to qualified rate, by campaign and by channel.
Cost per qualified or accepted lead, not cost per lead.
Qualified to customer rate, as the guardrail.
Revenue or margin per dollar of media, using your CRM's revenue, not the platform's.
When a platform quotes an uplift figure, ask for the metric, the baseline, the time period and whether it was a controlled test before it goes anywhere near a budget decision. Most figures presented at industry events are averages from the platform's own data, measured in the platform's own conversions. If you want a broader view of how to test claims like these, I cover the options in the four measurement frameworks every modern marketer needs.
A 30-day plan
Week 1. Audit your conversion actions. List every primary action in every campaign. Note what it measures and its typical delay. Most accounts find at least one campaign with several primaries.
Week 1. Define your four numbers. Agree with sales what counts as qualified and accepted. Write it down.
Week 2. Choose the primary. Pick the earliest commercially meaningful event inside about seven days. Everything else becomes secondary. Re-categorizing does not reset learning.
Weeks 2–3. Send the outcome back. Import qualified or accepted leads as offline conversions, matched by click ID and hashed contact details. Google now routes this through Data Manager, using enhanced conversions for leads. If you need help with the setup, that is what my Google Ads conversion tracking work covers.
Week 3. Add a value. Start with one or two flags that genuinely separate good leads from poor ones.
Week 4. Set the guardrail and test. Record the sale as secondary. Run the new primary against the old one as a campaign experiment, and judge it on cost per customer.
The real question
None of this needs new technology, and most of it costs nothing but time and an honest conversation between marketing and sales. The question underneath is simple: which event are you actually paid for, and does your bidding ever see it?
If the answer is no, the platform is not underperforming. It is doing exactly what you asked. Change the question. If you run a lead-gen business and want a second pair of eyes on it, here is how I approach Google Ads for lead generation.
Frequently asked questions
What is the difference between lead quality and lead quantity?
Lead quantity is the number of form submits, calls or messages a campaign produces. Lead quality is how many of those leads are reachable, relevant and likely to buy. A campaign can raise quantity while quality falls, because bidding to raw leads buys whatever is cheapest to submit.
Should I optimize Google Ads to sales instead of leads?
Usually not as the primary goal. Sales arrive weeks after the click and in small numbers, so bidding learns slowly. Bid to the earliest commercially meaningful event, such as a qualified lead or booked appointment within about seven days, and track the sale as a secondary conversion.
How many primary conversion actions should a Google Ads campaign have?
One, aligned to the real business goal. Several primaries at different funnel stages make Smart Bidding blend them into one predicted conversion rate. Set the others as secondary. Re-categorizing conversion actions does not start a new learning period.
How do I send qualified leads back to Google Ads?
Capture the click ID and contact details when the lead is created, mark qualified or accepted leads in your CRM, then import them as offline conversions using enhanced conversions for leads through Google Ads Data Manager.
Do Google Ads lead forms produce low-quality leads?
They can, because quick forms attract lower-intent users. On Search campaigns you can add phone verification and qualifying questions, then bid to qualified responses. Judge lead forms on qualified volume, not total leads.
Not sure which event your bidding is learning from?
Book a free 30-minute conversion signal review. We will look at your primary goals, your conversion delays and what your CRM already knows, and I will tell you what your bidding should be learning from instead.




Comments