Most attribution stops at the exact moment a conversion happens somewhere other than the website. Aparagon’s doesn’t have to.
A lot of what counts as a conversion never happens online: a signed quote after a home-services estimate, a vehicle test drive, an in-person membership sign-up, a phone call that closes the deal. A website pixel was never built to see any of that, so for a huge share of businesses, ad measurement simply stops at the exact point where the conversion happens, leaving the advertiser with no way to credit the campaign that actually worked.
AMP matches an advertiser’s own CRM data, however that outcome gets logged, against Amazon’s ad-exposure data inside Amazon Marketing Cloud, tying an online ad to a real-world outcome. Google and Meta offer something comparable, but only through a developer-built integration and a complex data pipeline an in-house team has to maintain. AMP does the same job through a platform that’s already built, against a signal set neither of them has access to: Amazon’s own consumer database.
Amazon matches the casino’s own enrollees against its known customer identities: more than 4 in 5 resolve to a real Amazon account.
From that identified group, AMP checks who was also served the campaign’s ads, tying a specific ad exposure to the in-person enrollment it produced.
AMP breaks the match down by campaign and by how many impressions it took: proof an ad worked, plus an answer for exactly what to run next.
A casino loyalty program that, by law, can only be joined in person makes this concrete. There’s no signup page, no online form, nothing a pixel could ever have seen. Amazon first matched more than 4 in 5 of the casino’s own enrollees back to a known Amazon customer identity, then checked who from that identified group had also been served the campaign’s ads, tying a specific online ad exposure to an in-person enrollment this advertiser had no way to prove before.
That’s not where it stops, either. Performance shows which specific campaign drove the outcome, so an advertiser can tell a strong placement from a weak one. That’s a level of detail no single top-line number can give you. Frequency shows how many times someone needed to see the ad before it worked, so an advertiser can tell how much exposure was needed to secure the conversion.
See what a closed-loop measurement setup looks like for your business.
Get StartedMost audience targeting starts with a search someone actually typed. Aparagon’s doesn’t have to.
Many advertisers can’t generate enough web traffic from keywords alone, and others don’t have a clear keyword to bid on at all. In those cases, such as when the need is driven by lifestyle and affluence signals that never show up in a query, traditional search has nothing to grab onto and broad awareness means paying to reach people who were never going to convert. These advertisers need other signals, ones that describe who a customer already is.
AMP uses Amazon’s own consumer data, such as a customer’s actual purchase history across categories that have nothing to do with the advertiser’s product, as a stand-in for demand that has no search behavior of its own. Instead of guessing at a keyword, the advertiser targets the kind of customer who already buys adjacent to their category, then measures whether that stand-in signal actually finds real buyers, or just resembles them on paper.
Amazon consumer data that relates to the target customer, luxury and wellness, stands in for a demand signal that doesn’t otherwise exist.
These audience segments are tested head-to-head, across Amazon’s premium properties and vast publisher network, to see which one finds buyers.
Once a segment proves out, investment scales against it, and AMP tracks cost per booking, not just clicks, to confirm it’s still working.
A mobile IV-therapy company ran exactly this test, after months of underperforming search campaigns. Audience segments built from Amazon consumer signals, such as high-net-worth asset ownership, wellness purchases, and luxury retail activity, were split evenly across multiple line items and delivered to targets within a radius of Los Angeles. It started as a smaller, roughly $4,000 test in month one, then scaled up to a $10,000 relaunch across months two and three.
Across those two campaigns, the mechanism produced 37 real bookings. Cost per booking dropped by roughly a third from the first test to the relaunch, the opposite of what usually happens when spend scales up, and a sign the audience signal itself was doing real work, not just buying more impressions. The advertiser generated a more than 4X return on ad spend.
See what a closed-loop measurement setup looks like for your business.
Get StartedMost “new movers” audiences rely on a third-party matched list. Aparagon’s doesn’t have to.
Some conversions are more likely to happen in a narrow window tied to a real-world event. For example, most people don’t search for trash-collection service until they’re mid-move. By that time, a competitor is almost certainly showing up too. Most advertisers know to build a “new movers” audience for exactly this reason. The usual way: a third-party file matched to consumers after the fact, which means a real lag between the move and the match, and real audience lost along the way.
AMP gives an advertiser a second option: Amazon’s own new-mover life-event segments, built directly from verified account and shopping signals instead of an external file. Same trigger, same targeting idea, but sourced natively, with no match step and no lag. Instead of asking an advertiser to take that on faith, AMP lets them test it directly against their existing matched audience: same market, same budget. The difference shows up in the campaign results, not a sales pitch.
The same “new movers” concept, built two ways: a third-party file matched after the fact, and Amazon’s own native life-event segment.
The matched third-party audience runs directly against Amazon’s native segment: same campaign parameters except for the audience.
Cost per lead and real signups are measured for each source, not just impressions delivered, to see which signal actually performs better.
A large regional waste-services advertiser first tested a “new movers” audience in month one, built from a third-party file of recent movers matched to Amazon identities, spending its full month-one budget across their targeted zip codes over the course of the entire month. It generated leads at $19.39 apiece, a modest return given how narrow and tightly targeted that first geography was, but a real baseline this advertiser would use to judge everything that came next.
The next month, they ran the same campaign concept again, but this time used an audience sourced directly from Amazon’s own native movers segment instead of the matched third-party file. Cost per lead dropped to $3.14, roughly six times cheaper, while still producing real leads at real volume: 319 versus 129 in the first test, with every other campaign parameter held constant.
See what a closed-loop measurement setup looks like for your business.
Get StartedMost performance measurement expects a conversion within days. Aparagon’s doesn’t have to.
Insurance involves a multi-step conversion process: a customer requests a quote, the carrier underwrites it, and only then does a policy bind. Weeks can pass between an ad click and a completed sale. Measurement built for a fast, single-session online conversion has almost nothing to show in that window, so an advertiser watching only for a finished sale can go months without a single data point confirming the campaign is working at all.
AMP doesn’t wait for the final sale to say something about a campaign. It tracks the full path a customer moves through, from early interest, to deeper engagement, to an eventual acquisition, so an advertiser has real evidence of a campaign working long before any policy is bound. No single stage proves a sale on its own. However, read together, and read over time as budget scales, they show a customer moving through the exact stages an insurance sale requires.
The advertiser has run the same life-event audiences — recent mover, household change, new car — since its very first month.
Monthly budget started at $1,000 in the first month and climbed to $4,000 by the fifth month, without ever swapping audiences.
Engagement further down the funnel was silent for the first three months, then appeared and nearly doubled the month after, as budget kept climbing.
This local insurance advertiser began with a $1,000 monthly budget against its three life-event audiences. Early-stage activity was holding roughly flat: 220 in month one, 119 in month two, 204 in month three, while the budget nearly doubled over the same stretch. Nothing deeper in the funnel had shown up yet, the expected shape for a multi-step conversion process that includes weeks of underwriting before any policy could bind.
By month four, with budget at nearly $2,000, deeper engagement appeared for the first time: 263 conversion events beyond the first stage. In month five, at $4,000 in budget, that measure nearly doubled to 477, with early-stage activity growing to a total of more than 1,300. Still missing is a bound policy, needing closed-loop data that meets AMP’s 100-conversion threshold, the next milestone.
See what a closed-loop measurement setup looks like for your business.
Get Started