How to Optimize Amazon DSP Campaigns: A Practitioner's Guide

Running DSP is not like running Sponsored Products. You cannot just set bids and check back weekly. DSP requires active management across audiences, creatives, placements, and measurement — and the optimization levers are completely different from what most Amazon advertisers are used to.

Here is how we approach DSP optimization after managing tens of millions in cumulative DSP spend at Skale.

Start With Audiences, Not Bids

The biggest mistake we see in DSP accounts is brands obsessing over CPMs and bids while ignoring audience construction. A perfectly optimized bid on the wrong audience is still a waste of money.

We build audiences in tiers. Tier 1 is retargeting: people who viewed your product or added to cart but did not purchase. This is your highest-converting audience and should get the most budget. Tier 2 is competitor conquesting: people who purchased or browsed competing ASINs. Tier 3 is in-market and lifestyle segments: broader audiences for upper-funnel awareness.

Each tier gets its own campaign with its own budget, bid strategy, and success metrics. Mixing tiers in the same campaign makes optimization impossible because the conversion rates are so different.

Creative Testing Is Non-Negotiable

We rotate at least 3-4 creative variations per campaign at any given time. Every month, we cut the bottom performer and introduce a new challenger. Over a quarter, this compounds into significantly better click-through and conversion rates.

What to test: lifestyle imagery vs. product-on-white, benefit-led headlines vs. promotional headlines, video vs. static display, and different calls to action. We have seen headline changes alone swing CTR by 40-60%.

One thing that consistently wins: custom creatives that include social proof. Star ratings, review counts, "Best Seller" badges — these elements boost CTR because they build instant credibility with shoppers who don't know your brand.

Bid Strategy by Objective

We do not use the same bid strategy across campaigns. Retargeting campaigns optimize for ROAS with a target cost-per-purchase. Upper-funnel campaigns optimize for reach and detail page view rate (DPVR). If you try to optimize an awareness campaign for ROAS, you will kill the reach and defeat the purpose.

For retargeting, we typically start bids 10-15% above the suggested range and let the algorithm compete aggressively for high-intent impressions. For upper-funnel, we start at or slightly below suggested and scale based on DPVR performance.

Frequency Capping Matters More Than You Think

Without frequency caps, DSP will show your ad to the same person 30+ times in a week. That is a waste of budget and it annoys potential customers. We cap retargeting at 3-5 impressions per day and upper-funnel at 2-3 per day. If someone has seen your ad 15 times and has not clicked, showing it a 16th time is not the answer.

Measuring What Matters

Amazon's default DSP reporting uses a 14-day lookback window with last-touch attribution. That overstates DSP's contribution. We supplement Amazon's reporting with total ASIN-level sales data, organic rank tracking, and branded search volume to measure the true halo effect.

The most important metric for DSP is not ROAS — it is incremental sales. Are you selling more total units than you would without DSP? At Skale, we use geo-based holdout tests and time-series analysis to answer that question definitively, not hypothetically.