When you reward a metric, people optimise for the metric — not the outcome it was meant to measure. In PPC, this turns ad accounts into rat farms: impressive dashboards, terrible business results. The fix is measuring what actually makes you money.
Start a conversationIn 1902, French colonial authorities in Hanoi were losing a battle against rats. Their solution: pay a bounty for every rat tail handed in. Logical enough — until rat catchers realised they could sever the tail and release the rat alive to breed more. Some enterprising locals started rat farms to maximise their earnings.
The bounty was achieved. The rat population grew.
In 1937, a paleontologist in Java paid locals per bone fragment discovered. Rather than preserve complete specimens, people smashed intact skulls into pieces to multiply payments.
In 2016, Wells Fargo set aggressive cross-selling targets. Employees opened over two million unauthorised customer accounts to hit their numbers. Billions in fines followed.
Different industries. Different centuries. Same mechanism. And it’s happening in your PPC account right now.
“When a measure becomes a target, it ceases to be a good measure.” — Charles Goodhart, economist. Once you attach rewards to a specific metric, people stop optimising for the underlying goal and start optimising for the number itself. Google’s algorithm is no different to a human employee in this regard.
Google’s automated bidding is extraordinarily good at finding more of whatever signal you give it. That is also what makes it dangerous.
If you tell it to optimise for form fills, it will find people who fill in forms. If many of those form fills are spam or unqualified leads that never convert, it learns to find more spam and unqualified leads. It has no way to know a lead was worthless unless you tell it.
The algorithm is not strategic. It is a very fast, very obedient rat farmer. It will optimise exactly what you measure, at scale, relentlessly. If your measurement is wrong, you are paying Google to be more efficient at the wrong thing.
Metric gaming is not always deliberate. Often it is the predictable result of a culture where hitting targets is rewarded regardless of what is underneath them.
If your agency is measured on impression share, they will find ways to grow impression share. If your in-house team’s bonus depends on the number of conversions reported in Google Ads, they will find ways to report more conversions. This is not malice — it is rational behaviour in response to the incentives you have set.
The fix is creating an environment where missing a volume target while improving genuine business outcomes is not just acceptable, but celebrated. That is harder than tweaking your conversion settings. It is also where the real leverage is.
The rat farm is not a Google problem or an agency problem. It is a measurement culture problem. And it starts with what you choose to put on the dashboard.
Revenue or profit (not just spend and conversions). Cost per qualified lead or cost per sale — verified against your CRM, not just reported by the platform. Incremental contribution where you have tested it. Margin by campaign where product data allows. Everything else is context, not a target.
The Rat Farm Effect describes what happens when you optimise for a metric instead of the outcome it represents. In PPC, this means campaigns that look successful on paper — high click volumes, low CPAs, strong ROAS — but are actively undermining business profitability because the metrics being chased are disconnected from real revenue.
Goodhart’s Law states that when a measure becomes a target, it ceases to be a good measure. In Google Ads, this plays out when Smart Bidding optimises for the conversion signal you have set up — if that signal is a soft conversion like a form fill or PDF download rather than actual revenue, the algorithm gets better and better at finding people who trigger that signal, not people who buy.
POAS stands for Profit on Ad Spend. Unlike ROAS (Return on Ad Spend), which measures gross revenue relative to ad spend, POAS accounts for product margins, shipping, returns, and other costs. A campaign with a 500% ROAS can still be loss-making if your margins are thin — POAS makes that visible so you can bid and budget on what actually makes money.
View-through conversions (where someone saw an ad but did not click, then converted later) can provide context about brand exposure, but they should never be used as a primary Smart Bidding signal. They routinely over-claim credit for sales that would have happened regardless of the ad, inflating reported performance without reflecting genuine incremental impact.
The most reliable test is comparing platform-reported conversions against your CRM or order management system. If Google Ads reports 50 leads but your CRM only shows 30, the gap is soft or duplicate conversions. Incrementality testing — pausing campaigns and measuring the revenue impact — is the gold standard for understanding what your ads are genuinely driving.
High click volume achieved through broad match or low-quality placements fills your campaign with unqualified traffic. Smart Bidding learns from these signals and bids more aggressively for similar users. Over time you end up paying more to reach people less likely to convert, while your cost per actual sale climbs and your budget gets wasted on traffic that was never going to buy.
Most accounts I audit are optimising for the wrong things — and the dashboards look great. If you want to know what your Google Ads are actually doing for your business (versus what they are reporting), get in touch.
Call 07410 907 104 Start a conversation