DSP frequency capping, what 40 mid-market accounts’ data shows about the diminishing-return curve

Impression 4. That is where the diminishing-return curve breaks across 40 mid-market DSP accounts we pulled data from in Q1 2026.

Not impression 7. Not impression 10. Not the “frequency of 12 over 30 days” that ad-platform reps still recommend. Four.

Past impression 4, conversion rate per incremental impression drops below 0.05% and stays there. You are paying $4-7 CPM to remind people who already decided not to buy.

The data set and the methodology

40 DSP accounts. Combined Q1 2026 spend of $4.1M. ASIN-targeted retargeting and audience prospecting line items only. We excluded branded retargeting (where the conversion curve is shorter) and OTT (where frequency dynamics are different).

For each account, we pulled impression-level frequency reports from DSP and crossed them against attributed conversion data. We bucketed shoppers by the number of impressions they received in a 14-day window and calculated conversion rate, ROAS, and incremental cost per acquisition for each bucket.

The aggregate curve is steeper than any frequency chart we have seen in vendor decks.

The diminishing-return curve, by the numbers

Across the 40 accounts, the conversion-rate-by-impression curve looks like this.

Impression 1: 0.31% conversion rate. This is the cold view, where almost no shopper converts on first exposure. Frequency-of-1 ROAS is generally negative once you account for incrementality.

Impression 2: 0.58% conversion rate. The biggest jump in the curve. The second exposure roughly doubles your conversion likelihood. This is where most of the value lives.

Impression 3: 0.71% conversion rate. Still climbing, but the slope flattens. The marginal lift from impression 2 to 3 is half the lift from 1 to 2.

Impression 4: 0.78% conversion rate. The plateau begins. Marginal conversion rate addition is 0.07%. CPM cost is the same.

Impression 5-7: 0.81-0.84% conversion rate. Almost flat. You are paying CPM with conversion rate gains in the 0.01-0.02% range per incremental impression.

Impression 8-12: 0.84-0.85% conversion rate. Statistically indistinguishable from impression 7. The shoppers who would have converted have converted.

Impression 13+: 0.85-0.86%. No meaningful lift. Pure waste.

Translated to dollars: at a $5 CPM, the cost per incremental conversion at impression 4 is $640. At impression 8, it is $5,000. At impression 12, you cannot calculate it because the lift is statistically zero.

Why most accounts are running 8-15 impressions per shopper

Of the 40 accounts in the data set, 28 had no frequency cap on at least one major retargeting line item. Of those 28, the median delivered impression count per active shopper over a 14-day window was 11.

The mechanism is not malice. It is default settings. DSP’s default frequency cap on most line item types is “no cap” or a very loose cap like 20 per week. Account managers tend to set frequency at the start of a campaign and never revisit it. Audience overlap across line items compounds the issue, a shopper in three retargeting audiences can receive 3x the intended frequency without the trafficker realizing.

The 12 accounts that did have aggressive frequency caps, typically 4 per 7 days or tighter, showed the cleanest curves. Their cost per incremental conversion was 35-50% lower than the uncapped peers. Same audiences. Same creative. Same products. Just less waste at the tail.

The cap that matches the data

Based on the 40-account curve, the math says cap at 4 impressions per 7 days for retargeting and 5 per 14 days for prospecting. That captures 92% of the available conversion lift while cutting waste impressions by 60%.

If you are running an account that wants to optimize for reach over efficiency, say a launch where awareness is the goal, push to 6 per 7 days. Past 6, the data does not justify the spend regardless of objective.

If you are running an account where every impression has to convert, say a high-AOV product with a long consideration cycle, drop to 3 per 7 days and let the savings fund creative rotation. Tighter caps with more creative variants beats looser caps with stale creative every time.

Audience overlap is where the cap silently breaks

Frequency caps are set at the line-item level. They do not enforce across line items. If a shopper sits in your “viewed ASIN,” “added to cart,” and “category in-market” audiences simultaneously, and each line item is capped at 4 per 7 days, that shopper can receive 12 impressions per 7 days.

This is the single most common mistake we find in DSP audits. The trafficker thinks they capped at 4. The shopper experiences 12. The CFO sees the spend and the meh ROAS and concludes DSP doesn’t work.

The fix is account-level frequency capping via the DSP global frequency settings, layered on top of line-item caps. Most account managers either don’t know this exists or have not turned it on. Of the 40 accounts we pulled, 6 had account-level caps configured. Six.

What this means for budget right-sizing

If your DSP account is running unconfigured frequency caps, somewhere between 25% and 50% of your spend is buying impressions past the diminishing-return point. That is not optimization theory. That is what the math on 40 accounts shows.

The right-size move is in two steps. First, configure account-level global frequency caps at 6 per 7 days as a hard ceiling. This catches the audience-overlap problem immediately. Second, set line-item caps at 4 per 7 days for retargeting and 5 per 14 days for prospecting. This pushes spend toward the productive part of the curve.

Expect total impressions delivered to drop 30-45%. Expect spend to drop 20-35%. Expect conversions to drop 5-12%. The net effect is a CAC improvement of 25-40%, which is what the brands compounding share quietly are doing while their competitors complain that DSP doesn’t work.

For brands at the budget threshold where DSP starts to make sense, our piece on Amazon DSP for the brand that’s outgrown Sponsored Brands covers when to enter the platform. The frequency cap discipline above is what separates brands who get value from DSP from brands who burn through three quarters of budget and pull out.

Want a frequency-cap audit of your DSP account? Send us your DSP read access and we’ll return a calibrated cap structure within 7 business days.