Every few months someone finds my eBay campaign in an old case study and asks me to replicate it. Seventy-five thousand dollars in paid social spend that returned twenty-five point seven million dollars, a 342 to 1 return, across eleven markets. It's a real number. It survived the strictest test we could throw at it, a randomized holdout, which is more than most legendary case study numbers can say. I'm proud of the work. I also think chasing it is the wrong lesson to take from it, and I say that as the person who ran it.
Here's what that case study doesn't show you: the specific conditions that made it possible. eBay had an enormous, warm, high-intent audience already sitting in first-party data, tens of millions of registered users with purchase histories, in an era when the platforms could still match that data at rates nobody sees today. We had a media budget large enough to test aggressively across markets without one bad week sinking the quarter. And critically, we were optimizing toward an outcome, marketplace transactions, that had a short, trackable path from ad click to completed sale. Take any one of those three conditions away and the same creative, the same targeting, the same team, would have produced a fraction of that return. Take away two and you have an ordinary campaign. The number was never a formula. It was an alignment, and alignments don't replicate on demand.
Why an outlier makes a terrible benchmark
I bring this up because I still see brands, usually mid-size, ambitious, working with a tight budget, set 342 to 1 as an internal benchmark after reading a case study like mine. That's not a target. That's an outlier being treated as a baseline, and it damages the account in a specific, watchable way. A team graded against an impossible ratio learns to hit it the only way it can be hit: by shrinking. Cut everything except the warmest retargeting audiences and the brand terms, and the ROAS climbs toward the fantasy while the business the campaign exists to grow flatlines. The benchmark selects for harvesting over growth, and the better the team executes against it, the smaller the future gets. I have watched competent marketers optimize an account into a corner trying to reach a number whose real lesson was about conditions, not effort. Meanwhile teams hitting an honest 5 to 1 on genuinely incremental spend feel like failures, when they are outperforming most of the industry on the metric that matters.
There's a second, quieter problem with the legendary number: nearly every version of it you encounter, including sometimes mine when it gets retold without the holdout context, is a platform-attributed figure, and platform attribution grades its own homework. Paid social identity resolution has degraded significantly since iOS privacy changes, and most attribution setups now overcredit the channels that are easiest to track, not the ones actually driving the sale. If your paid social dashboard still shows numbers that look too good, they usually are. The eBay figure holds up because a randomized control group of people who never saw the ads let us count only the revenue the campaign actually caused. Ask that question of most case study numbers, what was the control, and the legend usually gets quieter. I'd rather tell a client their real blended return is solid and defensible than let them keep a number on a slide that collapses the moment finance asks for an audit.
What to build instead
What I actually recommend building toward: a media mix across paid social, paid search, programmatic, and connected TV that's sized to the audience you can realistically reach at a profitable CAC, measured with a method that survives scrutiny, not just the platform's own reported ROAS. Set the benchmark from your own economics, margin, customer lifetime value, payback window, rather than from someone else's outlier, and grade the account on incremental return, verified periodically with holdouts scaled to your budget. Geo-based tests do for a mid-size advertiser what our randomized markets did for eBay, at a fraction of the complexity. That's less exciting than a headline number. It's also the version that still works in month fourteen, after the novelty of a new channel or a clever creative angle has worn off and you're managing the account on fundamentals.
The eBay campaign taught me plenty. The durable lesson wasn't the ratio; it was the measurement discipline that made the ratio believable, and that discipline transfers to any budget while the number transfers to none. Mostly it taught me how rare that specific alignment of conditions actually is, and how much of the real work happens in the unglamorous, disciplined middle where most advertising budgets actually live. That's the work we do for clients at Echelon on targeted advertising engagements, and it's a very different conversation from the one people expect when they've read the case study first.