Isn't cost-per-lead a standard metric for a reason?

It's standard because it's easy to measure, not because it predicts profit. It's a fine input. It's a terrible optimization target on its own — and we have the data to show it.

What do you optimize instead?

The number that maps to money for your business — usually cost-per-buyer or contribution margin, instrumented end to end. The right target is different for every model, which is exactly why we measure before we assume.

Do we need custom software to work with you?

No. Most engagements never need a line of custom code. Growth engineering is a way of working — measure, then build only when the measurement demands it.

Can you prove the cost-per-lead finding?

Yes. It comes from a real engagement spanning 150+ events and $11M+ in tracked revenue. We publish it anonymized to protect the client, but the analysis is real.

Jump sections

Methodology / Growth Engineering

We optimize the number that maps to money.

Most marketing optimizes the metrics the ad platform hands you — cost-per-lead, cost-per-click, return on ad spend. We instrument the full funnel, find the number that actually predicts profit, and build toward that. Sometimes that means writing software. That's growth engineering.

Let's build something

The metric everyone optimizes doesn't predict profit.

The entire industry optimizes cost-per-lead. It's the number the dashboards put in front of you, so it's the number agencies chase — cheaper leads, lower CPL, month over month. It feels like progress.

It isn't. Cost-per-lead is a proxy, and a bad one. A cheaper lead is not a more profitable lead — those are different questions, and the gap between them is where most ad budgets quietly leak. The only way to know which lever actually moves profit is to instrument the whole funnel and measure it, all the way down to the cost of an actual buyer.

So that's what we do first. Before we touch a campaign, we wire up the measurement. Optimization without instrumentation is just expensive guessing.

150+ events · $11M+ in tracked revenue — the data, not an assumption.

What most agencies optimize

r = 0.00no correlation with profit

What actually predicts profit

r > 0.7tracks directly to revenue

The Lightdrop data model

How we connect every signal — from first touch to revenue — into a single instrumented view.

Ad Layer

META_CAMPAIGN

stringPK

objective

string

status

enum

AD_SET

stringPK

meta_campaign_id

stringFK

event_id

stringFK

geo_targeting

json

status

enum

Collection

LEAD_FORM

stringPK

event_id

stringFK

platform_form_id

string

fields

json

wired_to_intake

bool

LEAD

stringPK

event_id

stringFK

form_id

stringFK

full_name

string

status

enum

received_at

datetime

Event

EVENT

stringPK

name

string

venue

string

city

string

latitude

float

longitude

float

event_date

date

status

enum

Audience

AUDIENCE

stringPK

event_id

stringFK

channel

enum

platform

string

external_id

string

AUDIENCE_MEMBER

stringPK

audience_id

stringFK

lead_id

stringFK

added_at

datetime

REACH_SEGMENT

stringPK

upcoming_event_id

stringFK

radius_mi

float

member_count

int

Outreach

CAMPAIGN

stringPK

event_id

stringFK

channel

enum

status

enum

BLAST

stringPK

campaign_id

stringFK

reach_segment_id

stringFK

blast_type

enum

scheduled_at

datetime

state

enum

MESSAGE_SEND

stringPK

blast_id

stringFK

lead_id

stringFK

channel

enum

status

enum

sent_at

datetime

Infrastructure

GUARDIAN_JOB

stringPK

name

string

cadence

string

last_run

datetime

last_status

enum

JOB_RUN

stringPK

guardian_job_id

stringFK

started_at

datetime

finished_at

datetime

result

enum

SLACK_ALERT

stringPK

type

enum

message

string

sent_at

datetime

What the data said

150+ events and $11M+ in tracked revenue, fully instrumented.

We ran this analysis for a multi-market live-events operator with a lot of spend and a lot of events to learn from. We instrumented the full funnel across 150+ events and more than $11M in tracked revenue, then looked for the relationship everyone assumes is there: cheaper leads, more profit. There was none. Cost-per-lead showed zero correlation with profitability — r = 0.00. Optimizing it harder would have changed nothing about the bottom line.

What actually drove the business was volume. As the operator ran more events per month — more than doubling its cadence — return on ad spend held steady at around $17 for every dollar spent. The system scaled without losing efficiency. The lever was throughput, not lead price — and we only know that because we measured instead of assumed.

r = 0.00

Correlation between cost-per-lead and profitability across 150+ events. Optimizing CPL harder would have changed nothing about the bottom line.

$17 ROAS

Return on ad spend, held steady as the operator more than doubled its event cadence. The lever was throughput, not lead price.

How growth engineering works in practice

Four principles. The difference between an agency that advises and one that builds.

Instrument before you optimize.

Wire up measurement down to cost-per-buyer and profit before spending against a goal. You can't optimize a number you can't see — and the visible numbers are usually the wrong ones.

Instrument before you optimize.

Wire up measurement down to cost-per-buyer and profit before spending against a goal. You can't optimize a number you can't see — and the visible numbers are usually the wrong ones.

Build the tool when the measurement needs one.

When the decision you need to make has no off-the-shelf tool, we write the software — integrations, decision tools, data pipelines. That's our custom tooling work, and it's how marketing engineering stops being a metaphor.

Build the tool when the measurement needs one.

Treat migrations as engineering.

Re-platforms are where growth dies quietly — lost catalog, dropped redirects, vanished rankings. We treat them as a build, not a copy-paste, which is the whole point of our zero-loss migration method.

Treat migrations as engineering.

Operate the system. Don't hand off a plan.

A deck doesn't compound. A running system does. We build the infrastructure and then we operate it — which is why our incentives stay aligned with the outcome, not the deliverable.

Operate the system. Don't hand off a plan.

A deck doesn't compound. A running system does. We build the infrastructure and then we operate it — which is why our incentives stay aligned with the outcome, not the deliverable.

One method, across every engagement

Growth engineering isn't a service you buy on its own — it's how we run all of them. It's underneath our e-commerce growth, B2B marketing, and crowdfunding work. The campaign is what you see. The engineering is why it compounds.

Questionsbuyersask

“

I've worked with Lightdrop on multiple projects over the last 5 years and am always amazed with the level of strategy, guidance and execution they bring to the table. Lior and his team are masters in digital and brand marketing.