Cheap leads cost more: why a falling cost per lead is a warning

A cost per lead far below your account's normal range usually means Meta found cheaper people to fill your form, not a better audience — because lead generation campaigns optimise for form submissions, and precise targeting raises cost per lead rather than lowering it.

A commercial property client in Delhi NCR ran a lead generation campaign for a week. It returned eighteen leads at roughly a seventh of what a lead had cost that account for the previous year. On every dashboard involved, including ours, that read as the best week the account had ever had.

Then their team called all eighteen. The people who picked up were in villages well outside the catchment for a commercial property in Dwarka. Several had never heard of the project. Two were worth a follow-up call.

The campaign had not found a cheaper route to the same buyer. It had found a different, cheaper person entirely — and every automated system watching it, ours included, called that a success and suggested spending more.

Why this happens

Meta's lead generation objective optimises for people likely to submit a lead form. That is a behavioural profile, not a buying profile. The cheapest people to reach inside it are, systematically, the people with the most time and the least discrimination about what they tap. Meta is not malfunctioning when it finds them. It is doing precisely what it was asked.

Instant forms make this worse. Name and phone number are pre-filled from the user's profile, so submitting costs two taps and no typing. A form with no friction cannot distinguish curiosity from intent, which means the signal Meta optimises against is weaker than it looks.

Cost is an output, not a dial

Here is the part that inverts the usual reading. When interests, geography and age are chosen carefully, the resulting audience is smaller and more contested. CPMs rise. Cost per lead rises with them. A precisely targeted campaign is supposed to be more expensive than a loose one.

So a cost per lead that falls sharply on a campaign you targeted carefully is not evidence that the targeting is working. It is evidence that delivery has stopped obeying it — because the cheapest way for Meta to buy another form submission is always to find someone easier to reach than the person you asked for.

Cheapness is the symptom. The disease is that delivery escaped the targeting.

What it looks like in the data

In that campaign, two things were visible in Meta's own breakdowns and neither was on any report anyone read:

• Delivery had reached three states when the brief named two neighbourhoods. The largest regional share of spend went to a state nobody had targeted — a 25-mile city radius travels a great deal further than it sounds.
• Roughly nine rupees in every ten went to a single placement. The ad set allowed four. Nobody chose that concentration; Meta did, because it was the cheapest surface to fill.

Both are ordinary Meta breakdowns — region, and publisher platform with platform position. Both were available from the first day of the campaign. The problem was never that the data was missing. It was that nobody had a reason to open it while the headline number looked excellent.

What to do instead

The fix is not a cost floor. That was our first instinct and it is wrong, because the whole point of running ads well is getting a two-hundred-rupee lead for eighty. A rule that flags anything cheap would flag your best work.

Cost cannot tell a cheap good lead from a cheap bad one. Only a phone call can. So treat an unusually low cost per lead as a trigger to verify rather than a verdict either way:

• Check the region breakdown against what you actually targeted. Spend outside the brief is the clearest single signal that delivery has drifted.
• Check the placement breakdown. One surface taking almost everything is a decision, and if nobody made it deliberately then Meta made it for you.
• Call a sample before you scale. Twenty minutes on the phone settles a question no dashboard can answer.
• Record what the calls found, per campaign. Cost per qualified lead is the only number that reflects what you actually bought.

The uncomfortable part of this story is not that Meta delivered poor leads. It is that every measurement system pointed at the account agreed it was doing well, right up until a human picked up a phone. Eighteen calls found what a week of automated monitoring could not.