Your cheapest leads might be your most expensive mistake.

A cheap lead can use up an expensive salesperson’s afternoon. Before cutting CPL again, follow a sample of submissions through qualification and opportunity creation. Find out whether the campaigns producing the most leads are also producing conversations worth pursuing.

The cheap-lead calculation that changes the meeting

Consider a hypothetical choice between two campaigns. Use leads old enough for sales to have evaluated them.

Campaign A has a €20 cost per lead; B has €50. If reporting stops there, A gets more money. Yet B produces twice the accepted demand for the same spend. My first move is to understand that quality difference, not reward A for producing more forms.

Define acceptance with sales using criteria they can apply consistently. Add a rejection reason such as unsuitable company profile, unsupported market, student research, duplicate or unreachable contact. These categories lead to different actions. “Bad lead” does not.

In the CRM, record original acquisition source, lead creation date, review date, acceptance and reason. Compare mature cohorts and distinguish unreviewed leads from rejected leads. If sales has not worked a campaign’s leads, the analysis cannot fairly condemn its quality.

Check routing and response before changing targeting. An abandoned suitable enquiry points to handling; repeated poor fit points elsewhere.

Count the conversations worth having

Qualification data can record demand quality or disguise a handling failure. A record marked “unreachable” after one late response is not equivalent to a prospect who was contacted promptly and proved unsuitable. Keep fit, contactability and review status distinct. That makes rejection reasons actionable and prevents a targeting change from being used to solve an inbox problem. The commercial definition should be simple enough that marketing and sales classify the same record the same way.

See how the arithmetic changes

Consider a calculation example, not a benchmark. Campaign A spends €5,000, produces 100 leads and creates 10 qualified opportunities. Campaign B spends €5,000, produces 50 leads and creates 15 qualified opportunities.

B looks worse if you stop at CPL. It looks more efficient at the opportunity stage. You would still need to examine deal value, sales effort and eventual wins before allocating the entire budget.

Agree what “qualified” means

Use criteria the team can apply consistently: relevant need, company fit, buying role, geography or other requirements that actually matter. Do not silently redefine the stage to improve a report.

Record reasons for rejection. “Bad lead” is not enough to guide a campaign change. Wrong company size, irrelevant problem, duplicate enquiry and no response point to different interventions.

Trace a sample through the journey

Follow enquiries from the campaign or entry page into the CRM. Review searches, audiences, landing-page messages and follow-up. A mismatch might originate in targeting, an overbroad promise or delayed sales contact.

Review mature cohorts rather than comparing fresh leads with older opportunities. If the typical qualification delay is substantial, the latest campaign can appear worse simply because its outcomes are incomplete.

Improve the signal carefully

Where appropriate, connect later CRM outcomes with acquisition reporting. Google’s enhanced conversions for leads documentation describes a way to improve matching for offline lead outcomes. Setup still needs reliable data, consent handling and validation.

Do not move bidding to a sparse, delayed event merely because it sounds closer to revenue. Establish data quality, event volume and reporting lag first. A staged pilot is easier to evaluate than a sudden account-wide switch.

The qualified-demo Google Ads case shows this strategy in an acquisition journey. If the weak point is the page itself, use the landing-page guide before assuming targeting is the only lever.

What I would review weekly

I would examine query or audience quality, qualification rate, cost per useful conversation and the reasons sales rejects demand. I would also check response time. Marketing should not be blamed for a lead that nobody contacted appropriately.

The goal is shared learning: which demand deserves more investment, which promise needs changing and which handoff is losing value.

Find the source field that quietly changes the story

Create an authorised test lead through the relevant form and record its captured source. Update or resubmit it through a later permitted test journey. Check whether the original source remains, changes or disappears.

Then follow the record into qualification and any linked opportunity. Confirm that the sales result joins to the intended lead rather than a duplicate. Use one distinct lead count and keep unreviewed records separate from rejected ones.

If the source is overwritten, fix field ownership before judging campaign quality. If the outcome attaches to another record, repair the join before importing it into reporting. Those failures can make a strong campaign look weak without any change in the people it attracts. The diagnosis begins with the record, not the blame.

Build a CRM field map before changing optimisation

Identify the fields required to connect an enquiry with its source and later outcome. Include the event definitions, relevant timestamps, ownership and identifiers supported by the chosen implementation. Review consent and data-handling requirements with the appropriate owners rather than assuming any available field can be sent anywhere.

Check how data enters the CRM, what overwrites it and how duplicates are handled. A first source may be lost when a later form replaces the value. An opportunity may be created under a different record. These issues can make a campaign appear weak or strong for reasons unrelated to its demand quality.

Use the map as an implementation brief. It should describe the information flow and acceptance checks, not prescribe a platform-specific setup without reviewing the account and current documentation.

Audit a sample of good and weak enquiries

Choose records across important campaigns or entry-page groups and trace the journey. Include accepted and rejected examples. Review the original promise, known intent, response timing and qualification outcome. Keep the review within authorised systems and avoid exporting unnecessary personal details.

Look for patterns rather than searching for a single bad lead that supports an existing opinion. A campaign can contain mixed demand. The question is which patterns are repeated enough to justify a change and which require more observation.

Bring sales feedback into the review, but ask for specifics. “These people are not serious” may describe unclear buying intent, unrealistic expectations or a poor follow-up experience. Turn the complaint into a diagnosis the team can act on.

Fix the promise as well as the targeting

A landing page can attract weak fit even when the targeting is reasonable. An offer of a free audit may generate interest from people seeking a document rather than ongoing help. A broad claim about serving every business may hide the conditions required for the service to work.

Clarify the audience, scope and next step. Explain what the conversation will cover and which situations the offer suits. Do not introduce unnecessary friction just to reduce submissions. A good-fit buyer should not have to complete your entire discovery process before earning a conversation.

Test the change against meaningful outcomes. If submissions decline but useful conversations remain stable, the journey may be filtering better. If both decline, investigate whether the page became confusing or excluded suitable visitors.

Introduce deeper feedback in a controlled sequence

First verify the CRM definitions and data flow. Then validate any supported import or reporting connection with a sample. Review timing, duplication and available event volume before allowing a deeper outcome to guide bidding or budget decisions.

A small, delayed event stream can be commercially valuable but operationally difficult for an algorithm to use. Keep an appropriate intermediate signal while evaluating the deeper data where necessary. The correct sequence depends on the account and current platform guidance, not a universal rule that every campaign should immediately optimise to revenue.

What the qualified-demo case adds

The qualified-demo case study reports a move from €420 to €285 per qualified demo while media spend stayed approximately flat. The diagnostic point is the mismatch between inexpensive form fills and sales quality. The intervention connected acquisition context with CRM outcomes before changing the bidding signal. Read it for that sequence, rather than treating the percentage as a target every account should reproduce.

Field note: the account was answering the wrong question

In the qualified-demo engagement, the company was spending approximately €28,000 a month on Google Ads. The headline lead cost looked acceptable. Sales was still spending scarce time on unsuitable conversations.

The useful evidence came from following a sample of enquiries into the CRM, not from another campaign dashboard. Some inexpensive submissions rarely became qualified demos. Other, more expensive leads were accepted more often. That changed the order of the work: agree what a qualified demo means, capture acquisition details, validate the imported outcomes, then test a change in bidding.

I would not skip the validation step just because the proposed conversion is closer to revenue. A delayed or inconsistent sales stage can train the account on noise. In this case, qualified-demo cost moved from €420 to €285 over 90 days with approximately flat spend. That is the reported engagement outcome, not a universal forecast.

The question I want answered in the meeting is precise: which rejected records would a campaign change actually prevent? If nobody can point to those records, we are discussing a feeling rather than an acquisition diagnosis.

Sales acceptance is a useful checkpoint, not the finish line

An accepted lead may still produce a weak opportunity. Sales may also change its standards when capacity or commercial priorities change. A rising acceptance rate needs context before becoming the new optimisation target.

Review mature opportunities alongside acceptance, and record material changes in qualification practice. Keep rejection reasons consistent enough to distinguish fit from handling. Unreviewed leads must remain visible.

If accepted demand grows but opportunities weaken, inspect the promise, buying readiness and sales definition before expanding spend. If the whole downstream pattern improves, use that evidence to refine the acquisition signal gradually. The feedback loop should move closer to value to the business without pretending any single CRM stage measures the complete value of a customer.

Leave room to challenge the sales definition

A rejected lead is not automatically evidence of bad marketing. Review the rejection reason, response time, contact attempts and whether the requested offer matched what the campaign promised. Use a shared definition of fit and retain reasons that can be audited. A smaller company outside today’s sales model may still be appropriate for a future offer; it should not be silently counted as accepted demand. The useful review distinguishes poor targeting, an unclear promise, a failed handoff and a deliberate commercial boundary. That makes the feedback more actionable than a single quality score.

Qualification without the blame game

Should we stop tracking CPL?

No. Keep it as a useful input. Pair it with later stages so efficiency is not judged at the easiest event to count.

Are more expensive leads always better?

No. Cost does not establish quality. Follow the outcomes and compare the full economics before drawing that conclusion.

Sources and further reading

Google Ads: enhanced conversions for leads: Lead measurement and implementation guidance.

Google Ads: offline conversion imports: Connecting later outcomes with ad measurement.

References checked 9 October 2026. Sources support the documented facts and tool behaviour; the recommendations and worked examples explain how I would apply them.