When you increase advertising spend, efficiency goes down. This phenomenon is called the law of diminishing marginal returns, because on the margin, each additional dollar invested brings lower returns. It occurs because you’ve used up all the ‘low hanging fruit’, and have to expand out to less relevant inventory. To understand why this happens, imagine showing up at a car auction, hoping to buy a car. If you only need to buy one car, you can keep entering different auctions that look to be good value, and bowing out gracefully when the price gets too steep. Eventually you’ll find a good deal. If instead you needed to buy 100 cars that day, the economics look very different. You may need to enter every single auction – regardless of the quality of the car – and outbid everyone else no matter how high they bid, or else you risk not hitting your quota. The price you pay per car will go way up, and the average quality way down: it’s unavoidable.
Modern ad platforms also work on an auction model, and therefore the same rules apply. If you increase your advertising spend, you need to buy more impressions, and that means tolerating lower quality, and outbidding everyone else. When an ad platform is optimizing to a cost per acquisition target, they’re essentially using their knowledge of the auctions and relative quality of the impressions to buy on your behalf. If the ad platform is targeting correctly, the first set of people targeted will be the lowest cost, highest intent to convert people available. Expanding to a wider audience means accepting people who have less intent and are more expensive to reach, so each incremental dollar yields a lower return. This is why you can’t just double your ad budget and expect the same return on ad spend – you’ll almost always see higher costs as an audience reaches saturation. Over time this problem gets worse not better, as ad fatigue kicks in and decreased tolerance for your ads lowers the ceiling.
This pattern will hold unless you find new creative that beats what you’re currently running. Only by improving the click-through rate or conversion rate of your ad, can you jump to a higher curve. That’s why it’s important to A/B test: you might find one ad works better than another at low ad spend, but when you scale it up it declines rapidly. Ads need to be at the same level of spend at the same time, to judge their relative effectiveness. Efficiency gains can translate to a higher return for the same spend, or more conversions at the same return on ad spend (ROI). Ad accounts looking for scale will typically leverage winning combinations to grow top-line sales, whereas more mature accounts will bank the extra profit. Either way all performance spikes are short lived, as ads start to decay as the audience gets tired of seeing them, another form of diminishing returns. There is a power law distribution to ad performance: 80% of your results will come from 20% of the variations you test. You’re unlikely to be able to predict what will perform ahead of time. The solution: maintain a rapid pace of testing, killing losing variants early, to make space for fresh ideas. Once you identify a winner, plot a new curve and choose the spend level that best accomplishes your goals.