Build the benchmark from your own history rather than a published average. Your last twelve months, segmented by channel and season, describes your actual market, audience and margins. An industry figure blends businesses with different economics, so beating it or missing it tells you almost nothing actionable.
Why are published benchmarks so unhelpful?
Because they average across businesses that have nothing in common with yours. A cost per acquisition benchmark spanning enterprise software and impulse retail describes neither.
They are also usually self-reported by the party publishing them, drawn from their own customer base, which is a selection rather than a market.
What should replace them?
Your own history, segmented enough to be comparable. The same channel, the same season, the same offer type.
- A rolling twelve months as the baseline.
- The same month last year, for anything seasonal.
- Your best comparable campaign, as the ceiling rather than the average.
- Break-even, as the floor below which nothing is acceptable.
How do you benchmark something with no history?
Accept that the first period is measurement rather than performance, and say so before it starts. A new channel's first month sets the baseline; judging it against an invented target creates an argument about an arbitrary number.
Where an external reference is genuinely needed, a competitor's publicly visible activity is more useful than an industry average, because at least it shares your market.
When is a benchmark actively harmful?
When it becomes the target. A team optimising to hit a benchmark will optimise the metric rather than the outcome, and cost per acquisition in particular is easy to improve by simply spending less on everything that scales.
Benchmarks are for interpretation, not for goals. The goal should be the commercial outcome; the benchmark only tells you whether a number is unusual.
Frequently asked questions
- How long before my own data is a reliable benchmark?
- Enough cycles to see seasonality, which usually means a year. Before that, compare against the previous period and label the baseline as provisional.
- Should benchmarks be shared with the client?
- Yes, along with how they were derived. A benchmark whose origin is unexplained invites the suspicion that it was chosen to flatter the result.
- What if performance is always below benchmark?
- Check the benchmark before the campaign. A baseline drawn from an unusually strong period will make ordinary performance look like failure indefinitely.