Advertising platforms report a large number of metrics, and most of them do not answer the only question that matters: did this make more money than it cost?
A conversion is the action you actually want — a purchase, an enquiry, a signup, a call. Until the platform can record when one happens, every other number is decoration. Clicks, impressions, and engagement all describe activity rather than outcome.
This is also the foundation everything else in the previous lesson depends on: automated bidding aimed at conversions cannot work if conversions are not measured correctly.
| Metric | What it tells you |
|---|---|
| Click-through rate | Whether the ad is relevant and appealing to who sees it |
| Conversion rate | Whether the landing page delivers on what the ad promised |
| Cost per click | What competition and relevance are costing you |
| Cost per conversion | What one actual result costs — far more useful than cost per click |
| Return on ad spend | Revenue produced for each unit spent — the closest thing to the real answer |
ROAS is revenue divided by advertising cost. Spend 10,000 and generate 30,000 in revenue, and ROAS is 3 — three units of revenue for each one spent.
The important caveat: ROAS uses revenue, not profit. A ROAS of 3 on a product with a thin margin can still lose money. Knowing your margin is what turns ROAS from a number into a decision.
Individual metrics mislead; combinations diagnose. Some common patterns:
Cost per click is the metric people watch most and it is one of the least informative on its own. A campaign with expensive clicks that produce customers is outperforming a campaign with cheap clicks that produce nothing, every time.
Platforms report the conversions they believe they caused, and this is systematically generous. Each platform sees only its own contribution and tends to claim credit for journeys that involved several touchpoints — an ad, a search, an email, a direct visit days later.
Add up the conversions claimed by every platform and the total commonly exceeds the conversions that actually happened. Privacy restrictions have also made tracking less complete than it once was.
The practical response is to treat platform-reported numbers as directional, and to sanity-check them against something independent — your own sales records, enquiry counts, or a simple question on your enquiry form asking how someone found you. If the platforms claim thirty conversions and you had eleven enquiries, the platforms are wrong.
Small numbers are noise. Three conversions from two hundred clicks tells you almost nothing, and reacting to it produces the constant-adjustment problem from the previous lesson. Wait for enough volume to be confident a difference is real before acting on it.