The most dangerous sentence in lead generation is: "We lowered CPL, the campaign is a success." I've watched the same film in higher education, healthcare and consulting: form costs drop, the team celebrates, and two months later sales asks — "with this many applications, why are there no enrolments/appointments?" Because what was optimized wasn't the business outcome; it was the form itself. Whatever the KPI is, the system buys it. Order cheap forms, receive cheap forms.
Why quality falls as forms get cheaper
The mechanism is transparent: platform algorithms find the users who reach your stated goal most easily. When the goal is "form submission," the system finds the audience most inclined to fill forms — which is not the same set as the audience most inclined to buy. Every friction-reducing move pushes the same direction: cutting fields to three, using auto-filled native lead forms, "apply now, no obligation" copy… All of it lowers CPL, and all of it raises the share of low-intent users. That's not a platform defect; it's your instruction being executed to the letter.
The invisible half of the funnel
The form is where measurement begins, not where it ends. A healthy lead operation tracks at least four more stages after the form: reached (did they answer the phone), qualified (real need/budget), opportunity (appointment/meeting), closed (enrolment/sale). Without this chain, campaign optimization is blind flight — you cannot tell which campaign delivers "cheap but empty" and which "expensive but full" leads. In the Bay Atlantic University work, the real turning point wasn't on the ads side but exactly here: once post-application stages were reported by channel and campaign, the data finally dictated which campaigns to defund and which to double down on.
The feedback loop: teaching platforms what a good lead is
The good news: platforms can be taught. The structure I build has three steps. First, clear lead-stage definitions in the CRM (new → reached → qualified → opportunity → won), with source (campaign/channel) attached to every lead. Second, sending stage transitions back to the ad platforms — enhanced/offline conversion imports on Google, CAPI on Meta. Third, migrating the bidding goal over time from the form to a deeper stage: the system stops hunting form-fillers and starts hunting users who turn out qualified. The transition takes patience — deep-stage volume is low, learning slows — but over a three-month window the pattern is almost always the same: CPL rises, cost per customer falls.
The friction paradox: sometimes you should make the form harder
When the KPI is forms, friction is the enemy; when the KPI is sales, friction is a filter. For one education client, adding a single question — "which term are you planning to start?" — reduced form volume and visibly increased the rate at which sales found a real candidate on the other end of the line. Intent questions, budget ranges, preferred callback times — they worsen the CPL report and improve the business report. You have to decide upfront which report you want to look good.
In short: in lead gen, a KPI is not a measurement preference; it's a behaviour order. Optimize for forms and you get forms; build the infrastructure to optimize for sales and you get sales. For the higher-education deep dive on this topic, see "A lead is not an application".