Performance marketing: measurement before media
Performance marketing only works when measurement comes first. What to define, track and test before you scale spend, and which metrics to trust.
By Roozbeh Nazari · CEO
Performance marketing is usually described as marketing where you pay for results: clicks, leads, sales. The description hides an assumption. It only works if you can measure those results accurately and connect them to the money you spend. Many teams start with media, launching campaigns and choosing bidding strategies, and discover later that the numbers they optimised against were incomplete or misleading. This guide argues for the opposite order: build the measurement first, then buy media against it.
Why measurement comes before media
Modern ad platforms optimise automatically towards whatever conversions you tell them to count. Google Ads describes conversion measurement as the way to learn which keywords, ads and campaigns drive valuable customer activity, and its automated bidding learns from that data. If the conversion signal is wrong, the platform will efficiently buy more of the wrong thing.
Common examples are easy to spot once you look:
- Counting every form submission as a conversion, including spam and job applications.
- Counting page views of a thank-you page that can be reached without submitting anything.
- Tracking a lead but never learning whether it became a customer.
- Double counting the same sale in several platforms, each claiming full credit.
None of these problems is solved by a better campaign structure. They are solved by deciding what a result actually is and measuring it properly.
Performance marketing metrics that matter
Start by writing down the business outcome, then work backwards to the signals you can measure.
- Business outcome: revenue, qualified pipeline or booked appointments, agreed with finance or sales.
- Primary conversion: the action closest to that outcome that you can track reliably, such as a qualified lead, a purchase or a booked call.
- Secondary signals: earlier actions such as key page views or content downloads, used for diagnosis rather than bidding.
- Efficiency metrics: cost per qualified lead, cost per acquisition and return on ad spend, calculated from the primary conversion, not from clicks.
- Quality checks: the share of leads that sales accepts, and the time it takes a lead to become an opportunity.
In Google Analytics 4, recommended events give you a consistent way to record actions such as generating a lead or completing a purchase, and they update the predefined reports automatically. Mark only the events that represent real value as key events, and import only those into the ad platforms.
Set up the measurement stack
A practical stack for most businesses has four layers.
First, clean tracking on the website: a tag manager, consistent event names and consent handled correctly for your markets. Check events in DebugView or real-time reports before you trust them.
Second, conversion tracking in each ad platform, fed by the same definitions. Google Ads can use the Google tag or GA4 key events as data sources.
Third, a way to bring offline results back. When sales happen on the phone, in a meeting or in a CRM, offline conversion imports and enhanced conversions for leads let you send those outcomes back to the ad platform, so bidding learns from customers instead of form fills. Google is moving these uploads to its Data Manager API, so check which method your setup uses.
Fourth, a reporting view that combines spend and outcomes by channel, built on the same definitions so that every team reads the same numbers. Our analytics and data work usually starts here.
Test incrementality before you scale
Platform-reported conversions tell you what happened after an ad was shown, not what the ad caused. Before you increase budgets significantly, test whether the extra spend brings extra results. Options range from simple on/off or geographic tests to marketing mix models. Google's open-source Meridian model, for instance, is designed to help teams estimate channel contribution and plan budgets using aggregated data, with experiments used to calibrate it.
You do not need a sophisticated model on day one. You need the habit of asking whether a result would have happened anyway. Our guide to multi-touch attribution explains when attribution models help and when they mislead.
Then buy media
With definitions, tracking and a testing habit in place, the media work becomes clearer: which channels reach the right people, which offers convert and how much you can afford per outcome. Our paid media team plans campaigns this way. We do not promise a fixed return on ad spend, because that depends on your market, offer and sales process, but we do make sure every result we report can be traced back to something real.