Technology

The cycle gets smarter: How we’re using Claude for Amazon DSP campaigns (for less €€€)

woman looking at data on computer

Amazon Demand-Side Platform (DSP) advertising used to be quite daunting for sellers.

With its heavy complexities, and associated heavy budget, many sellers never even touched it.

But e-Comas is now leveraging Claude for Amazon DSP, changing the economics of advertising on this far-reaching platform. 

With Claude able to plan, launch and analyse DSP campaigns in hours rather than days or weeks, we can now offer DSP managed services at much lower rates.

Not only that, but continued use of Claude for Amazon DSP can make your campaigns smarter and smarter over time.

Here’s how it’s worked for one of our clients.

 

The client

Our client is a consumer electronics group, selling two brands in five markets: Germany, France, Italy, the UK and Spain.

They had a €14,000 budget for Amazon DSP in April-May 2026.

They wanted full-funnel campaigns in DE and FR, and focused remarketing in IT, UK and ES.

 

Reading the catalogue with Claude for Amazon DSP

We start the process by reading our client’s catalogue – using Claude to gain full intelligence on the products much faster than we could do it before.

Our process sorts the client’s ASINs into these four types, based on traffic vs. orders:

  1. Champion (High traffic, high orders)
    These ASINs are the portfolio workhorses, with proven demand and proven conversion. The campaign needs to protect and scale.
  2. Sleeping giant (High traffic, low orders)
    These are the ASINs that shoppers visit but do not buy. These need to be routed into remarketing and conversion work.
  3. Hidden gem (Low traffic, high orders)
    These ASINs convert well but too few shoppers find them. Upper-funnel reach can put them in front of new eyes.
  4. Slow mover (Low traffic, low orders)
    These ASINs should be parked unless strategic: they’re not worth the spend. The budget saved here can fund the Champions.

The beauty here is that there’s no guesswork; we were never relying on a hunch. The quadrant decides which tactic gets budget in each market.

From this stage, we can use Claude to prepare the campaigns in a matter of hours, not days.

Product intelligence from Claude

 

Claude builds the bulk files in minutes

To structure by role, we split every campaign into two types: prospecting for new shoppers or remarketing to warm ones. We split by brand and by quadrant.

The more data you use in a campaign, the more you can steer your spend. We split orders into many line items, splitting by device and inventory type, to give us more control.

That type of granularity used to take hours of bulk-file work per deployment. But Claude builds and updates the bulk files in minutes.

The time saved means that we, the agency, can offer managed DSP at much better rates for sellers.

Claude prepared the deployment; our team reviewed and pressed go.

 

The results of using Claude for Amazon DSP

Our client’s campaigns ran over April and May 2026, with fantastic results.

Across the five markets, total ad spend was €14,310, with €277,000 of attributed DSP sales – meaning a total ROAS of 19.4.

The campaigns reached 1.76m shoppers over those two months, with around 70% new-to-brand customers.

 

Claude reporting – weekly, not monthly

Claude also allows us to feed useful insights back to our client much more frequently.

We feed Claude the raw data – 199-column campaign exports, a 1.17GB supply log, reach and frequency data, and device and technology data – and get insights quickly.

In the past this was a monthly deliverable at best. With Claude, it’s weekly, with fixes attached, so we can optimise each campaign while it’s still running.

Some examples:

Frequency. Claude found the same shoppers were seeing ads 10-11 times in two markets – but retargeting stops earning after about 7 exposures. We capped exposures at 7, meaning wasted impressions stopped within days.

Supply path. Claude can read supply logs that Excel cannot open. We found in one supply log that a single exchange was taking 17% of spend. We requested an exclusion – and moved the budget to the winners.

Device mix. Claude found that phone absorbed 77% of spend at 16x ROAS, while PC and connected TV were returning 25 to 38x. We reweighted our bid modifiers for the next cycle.

woman looking at data on computer
Image by DC Studio on magnific.com

 

Core takeaway: using Claude for Amazon DSP allows the data to compound

None of the data we’ve produced in these campaigns has been disposable. Every insight, report and decision is feeding the next activation, and the account gets smarter every cycle.

The cycle looks like this:

  1. Audience intelligence

DSP and video flights reveal who actually buys in each market: genres, interests, devices, buying patterns.

  1. AMC reporting

Clean-room deep-dives join it all at shopper level: paths to conversion, frequency curves, gateway products.

  1. Project memory

Everything lands in the client’s Claude project: reports, decks, raw exports, meeting notes, actions taken.

  1. Smarter next moves

That memory shapes what comes next: PVA flights, streaming TV (including a World Cup campaign running last week), new creative.

Each cycle starts smarter as the data compounds.

 

With thanks to Marc Villaseca, e-Comas’s Performance Marketing Director, who co-wrote this blog.

 

Want to achieve ROAS of 19 by leveraging Claude for Amazon DSP? Talk to our team today!

Share This :