Why your tCPA is lying to you

Target CPA bidding is broken for B2B. Specifically, it's broken for any account where the time between a click and a close is longer than 90 days.

Which is basically all of B2B.

Google will tell you that tCPA uses sophisticated machine learning to optimize your cost per acquisition across channels and time horizons. It's true. The machine is learning. It's just learning from the wrong data.

Target CPA optimizes for what it can see. In your Google Ads account, what it can see is short-term conversions. Form fills. Gated content. Demo requests. Anything that converts within 30-60 days. It can't see the 9-month sales cycle. It can't see the conversion that shows up in Salesforce six months after the click. So it ignores it. Instead, it optimizes for the conversions it can see, which are usually the cheap, fast, low-intent ones.

Result: your cost per acquisition looks fantastic. Your actual cost per closed deal is terrible.

How tCPA actually works

The algorithm needs conversion data to optimize against. It looks at every click, waits for a conversion signal, and learns from it. Then it bids the next click up or down based on what it learned.

In e-commerce, this works great. A user clicks an ad at 2pm. They buy at 2:15pm. Conversion recorded. Algorithm learns. Repeat a thousand times. The algorithm gets really good at predicting which clicks will convert.

In B2B manufacturing, a user clicks at 2pm in January. Nothing happens. February, they're back. Nothing. March, they talk to sales. April, they're in an opportunity. July, they close. Conversion recorded.

But by then, the algorithm has already given up on January traffic. It made decisions based on what it saw in February. Those decisions were bad. It should have bid higher on January. It didn't know to.

The algorithm optimizes against the visible conversions. The invisible ones (the ones actually generating revenue) don't factor in. So it starves the campaigns that generate real deals and oversizes the campaigns that generate quick junk.

The damage this does

Here's what I see in tCPA-heavy accounts.

Form fills go up. Click volume goes up. CPL looks fantastic. But when you trace those leads to Salesforce, the picture gets weird. Conversion rate on those leads is 3-5%. The ones that do close take forever. Sales is complaining about garbage.

Meanwhile, the high-value terms that actually drive real deals are underfunded. The algorithm sees them as expensive. They don't convert within the 30-day window, so the algorithm assumes they're bad. Budget gets pulled away. Real revenue opportunity gets starved.

The blended CPA looks great because the cheap junk is dragging down the average. But the actual cost per closed deal is climbing. The margin is disappearing. And you can't see it unless you're looking at Salesforce data.

I ran this experiment with a client once. We looked at CPL from tCPA bidding: $67 a lead. Looks fine. Then we traced those leads to closed-won in Salesforce. Average deal size: $40,000. Close rate: 3.2%. Real cost per closed deal: $2,093.

We switched to manual CPC bidding with a focus on high-intent, account-specific signals. CPL went up to $147. Close rate went to 18%. Real cost per closed deal: $817.

The CPL almost doubled. The actual cost per deal dropped 60%.

The counter-argument, and why it doesn't work

The pushback is always "Value-based bidding fixes this." Google pushes Value-Based Bidding (VBB) as the answer to tCPA's limitations. It's supposed to optimize for revenue, not conversions.

It helps. But only if you have actual revenue data flowing back into Google Ads in real time. Most B2B accounts don't. They're still uploading GCLIDs once a month from a spreadsheet. The revenue data is three weeks old. The algorithm is still optimizing against stale signals.

And even with VBB, the latency problem remains. If your deal closes six months after the click, the algorithm is still waiting. It's still making decisions based on incomplete data.

Value-Based Bidding is better than tCPA. But it's still optimizing against a 90-day window in a market with a 180+ day cycle.

What to do instead

Stop using tCPA in B2B accounts. Move back to manual CPC bidding or, if you need automation, use static target CPCs tied to explicit business metrics.

Here's what I do. I calculate the expected customer lifetime value for each product line. I back into a target CPA that lets the company hit margin. Then I set static target CPCs that ensure we hit that CPA target. No algorithm. No dynamic decisions based on invisible data. Just discipline.

If the target CPA is $300, and my average first-click CPC is $15, I can afford to bid on 20 clicks to generate one conversion. So I bid fixed CPCs at roughly $15 until the conversion rate data proves it should be higher.

This is less "sophisticated" than tCPA. It's also actually working.

Or, if you want to use algorithmic bidding, wire up your CRM properly first. Build the BigQuery pipeline. Make sure Salesforce stage data is flowing back into Google Ads via the API on a schedule. Make sure the algorithm is actually seeing real revenue data, not just form fills. Then use Value-Based Bidding.

But don't use tCPA and pretend it's working. It's just optimizing against the visible part of a much longer cycle. Which means it's optimizing wrong.

Your CPL will look great. Your actual cost per deal will be a disaster. And you'll only see it if you're looking at Salesforce.