Value-based bidding for lead generation: stop buying leads, start buying customers

There is a moment, in almost every lead generation account we audit, when the Google Ads numbers and the sales team’s numbers stop telling the same story. Google says cost per lead is down 30%. Sales says the leads have got worse. Both are right, and the reason is always the same: the campaign is optimising precisely toward the objective it was given.

Why optimising for lead count attracts junk

Maximise conversions and target CPA share an implicit assumption: every conversion is worth the same as every other. To the algorithm, the form filled in by a student researching a dissertation and the one from a procurement manager with an approved budget are the same event. If the first costs £20 and the second £120, the rational choice — for a system asked to minimise cost per conversion — is to buy the first.

That is not a malfunction. It is successful optimisation toward the wrong objective. And it gets worse over time, because the better the system becomes at finding cheap leads, the more the mix drifts toward generic queries, low-cost placements and people who will fill in any form put in front of them.

A worked example

Take two campaign groups with the same monthly spend, £4,000, and an average deal value of £6,000:

Under target CPA, campaign A looks two and a half times more efficient, and the system will shift budget and auctions toward it. With a value signal, the same campaign turns out four times worse. Same data, opposite decision. The whole case for value-based bidding sits in that table.

How to put a value on a lead

A lead’s value is not a customer’s value. It is a customer’s value multiplied by the probability that this lead becomes one. As a formula: expected value = close rate from the current stage × average deal value.

The starting point is the funnel stages your CRM already records. Suppose these, with historical stage-to-stage rates measured over the last twelve months:

The last column is the one that matters. If you send Google £300 for the lead, then £750 for qualification, then £1,500 for the proposal and then £6,000 for the contract, that one customer is worth £8,550 in your account: you are inflating the value of everyone who progresses, and the reported ROAS no longer means anything. The increments, by contrast, add up exactly to the deal value.

Two possible architectures

• One conversion action per stage, each carrying its incremental value. The most readable option in reporting, and the most robust when stages are cleanly defined in the CRM.
• A single conversion, updated over time with conversion adjustments: a restatement changes the value of the original conversion as the lead progresses, a retraction removes it when the lead turns out to be spam or out of scope.

Retractions deserve a special mention because they are the most underused tool in the box: taking obviously fake leads out of the count teaches the system what not to buy, and costs almost nothing. Google recommends uploading adjustments promptly, and no sooner than 24 hours after the original conversion; late adjustments update reporting but carry less weight in bidding.

The technical side of getting data in — enhanced conversions for leads, offline import, the move to the Data Manager API — is covered in our guide to offline conversions and the Data Manager API. Here we assume the data arrives: the question is what to do with it.

The latency problem

If your sales cycle runs three months, optimising solely on signed contracts means training the system on clicks a quarter old. The signal arrives late, it is sparse, and some of it falls outside the conversion window. The result is an algorithm that learns slowly and reacts badly to change.

The practical answer is to make your main signal the furthest stage that reliably arrives within a few weeks of the click and has enough volume. In most accounts that is the qualified lead. Later stages still get sent, but that one does the heavy lifting. Conversion windows and their effect on the numbers are covered in our piece on data-driven attribution across Google Ads and GA4.

Conversion value rules

Conversion value rules let you adjust the value of a conversion at auction time based on location, device and audience. For example: leads from a region where you close twice as often are worth 100% more; leads from an existing customer list are worth less, because you would have had them anyway.

They earn their place in two situations: when a difference in value is known but does not flow through the CRM, and in the early phase before the value loop is complete. They are not a substitute for the CRM. A rule is an average applied to every lead in a segment; sales data is the truth about the individual lead. If you have the latter, rules should only correct what it cannot see.

From target CPA to target ROAS: the transition plan

Switching strategy is where the damage gets done. Correct values sent to a campaign still on target CPA achieve nothing: the strategy ignores them. Wrong values sent to a campaign on target ROAS do harm quickly. Order matters.

1. Values first, strategy second. Start sending values while the campaign is still on target CPA and leave them running for at least four to six weeks. That serves two purposes: confirming they arrive correctly, and building the history the new strategy will start from.
2. Check the volume. Our rule of thumb: a campaign that cannot generate thirty or so value-bearing conversions a month is a candidate for a shared portfolio strategy, not its own target ROAS. Below that, average value swings too much from week to week.
3. Start from observed ROAS. Set the initial target close to what you actually achieved over the preceding weeks, not what you would like. An ambitious target on day one throttles volume, and since the August changes targets bear on delivery more than before — see Smart Bidding Exploration and promotion mode.
4. Use an experiment. If volume allows, a 50/50 campaign experiment gives you a clean comparison instead of a before-and-after muddied by seasonality.
5. Move the target in small steps. 10-15% at a time, at least two weeks apart. Every change reopens a settling period.
6. Judge on the CRM, not the platform. ROAS in Google Ads is built on the expected values you gave it. The real test is the pipeline: qualified leads, proposals, contracts.

Maximise conversion value with or without a target?

Target ROAS is an option within Maximise conversion value, not a separate logic. Without a target, the system spends the full budget chasing the most value; with one, it chases the most value while holding a minimum return. With a tight budget, starting without a target for a few weeks is reasonable and shows you where return settles. With a budget that is generous relative to demand, running without a target lets the system spend into unprofitable auctions: better to set a cautious target from the start.

When it is not yet time

Value-based bidding is not always the right answer. It is not when:

• The CRM does not record stages reliably: wrong values are worse than none, because the system takes them seriously.
• Volume is too low even with campaigns pooled: target CPA on a well-defined qualified lead is the better choice.
• Customer value really is uniform: if every contract is worth roughly the same, the gain comes from qualification, and optimising on that is enough.

In every other case, continuing to optimise for lead count means paying an ever-smarter algorithm to find ever-less-useful contacts. It also bears on how much to invest: in our article on how much budget Google Ads actually needs, the quality of conversion value is one of the four checks to run before raising spend.

If you want to know what each lead in your account is really worth and whether it is time to change strategy, the free audit starts exactly there: CRM stages, values, volumes. You can see which services cover this ground and answers to common questions in the FAQ.