The Ecomm Analyst

ROAS drops as you increase spend because you buy your cheapest conversions first. The people easiest to convert, the ones already searching for you or already on your list, get reached at low cost. Every additional dollar has to find someone slightly less interested than the last person, and that costs more. This is not a tracking failure and it is not your media buyer losing their touch. It is the normal shape of a demand curve, and the sooner you plan around it the fewer arguments you have.

The reason it feels like a failure is that ROAS is an average, and averages hide the thing you actually need to see.

Average ROAS versus marginal ROAS

Say you spend $10,000 in a month and attribute $40,000 in revenue. Your ROAS is 4.0. Next month you spend $15,000 and attribute $52,500. Your ROAS is 3.5, and someone in the Monday meeting says performance is down.

Performance is not down. Look at what the extra money bought. You spent an additional $5,000 and got an additional $12,500, which is a ROAS of 2.5 on the incremental spend. That 2.5 is the number that should drive the decision, because it is the only one that describes what happens if you spend more. The 3.5 is a blended average of a great first $10,000 and a mediocre next $5,000, and it tells you nothing about what the next dollar will do.

I calculate marginal ROAS whenever spend moves meaningfully between periods. Take the change in attributed revenue, divide by the change in spend. It is crude, it ignores lag and seasonality, and it is still more decision-useful than the headline figure. If marginal ROAS is above your break-even threshold, the spend increase was correct even though average ROAS fell. If it is below, you overshot, regardless of how healthy the average still looks.

Where the curve usually bends

The decline is rarely smooth. In practice I see a plateau followed by a fairly sharp drop, and the location of that bend is specific to the account rather than to the industry. A brand with a large warm audience and a narrow product range hits the wall earlier than one with broad appeal and a deep catalog.

Three things move the bend. Audience saturation is the obvious one, and you can see it in rising frequency alongside falling click-through rate. Creative fatigue is often mistaken for saturation, and the tell is that a genuinely new concept restores performance while a new audience does not. And auction competition, which is entirely outside your control, shifts the whole curve down during peak retail weeks regardless of anything you do.

Distinguishing these matters because the responses are different. Saturation means expand targeting or accept a lower ceiling. Fatigue means produce. Competition means wait, or budget for it in advance.

The brand search trap

There is a specific version of this that catches almost every growing brand. As you scale prospecting, more people search your brand name, and if you are bidding on your own brand terms those searches convert at an absurd ROAS. Your average ROAS looks resilient because a growing share of your spend sits in the cheapest possible bucket.

What has actually happened is that your prospecting is subsidizing a line item that takes credit for it. Strip brand search out and recalculate. If the remaining ROAS has fallen much faster than the blended figure suggested, you now know where you really are.

Watching it without doing arithmetic every week

The practical version of this is a chart with spend on one axis and ROAS on the other, plotted weekly over a long enough window to see the shape. You want at least six months, and you want it per channel, because the curves bend at different points.

Any tool that can group ROAS by time and by channel will produce this. ThoughtMetric, which sponsors this blog, exposes spend and ROAS grouped by channel and campaign, along with a separate new customer ROAS metric that is genuinely useful here, since the marginal dollar should be buying new customers rather than orders you would have received anyway. It runs $99 per month for 50,000 monthly pageviews with all features included at every tier, and offers a two-week free trial. Whatever you use, the requirement is that it can plot the relationship over time rather than just reporting this month’s average.

What to tell the person asking why ROAS is down

Bring three numbers. Average ROAS for the period, marginal ROAS on the spend increase, and your break-even threshold. If marginal sits comfortably above break-even, the correct action is usually to keep going and to stop treating average ROAS as a performance grade. If marginal has fallen below break-even, you have found your ceiling for now, and the conversation moves to creative, offer, or margin rather than to budget.

Falling ROAS during a scale-up is a sign the plan is working, right up until the marginal number crosses the line. Knowing where that line sits is the whole job. I wrote about setting the threshold itself in what is a good ROAS for e-commerce.