AMC Instructional Queries Worth Running First
Start with a reach-and-overlap query before anything else — it's the fastest way to see whether DSP and sponsored ads are double-counting the same shoppers, and it sets up every query that follows. After that: a path-to-conversion query, a new-to-brand-by-channel breakdown, and only then a holdout or incrementality query once the first three have shaped a real hypothesis.
What this looks like in a real account
Why order matters here
AMC's query library is large enough that starting with the wrong one wastes real time — a sophisticated incrementality test built before you understand your own account's overlap and path structure is answering a precise question you haven't actually framed correctly yet. The sequence below is built around getting the cheap, foundational answers first, then using what they show you to design the more expensive, more rigorous tests with an actual hypothesis behind them.
First: reach and overlap
Before anything else, run a query that answers a single question: of the shoppers exposed to DSP in a given period, what share were also exposed to sponsored ads, and vice versa. This is the query behind the double-counting problem covered elsewhere on this site, and it's foundational because every other attribution question you ask is downstream of knowing whether your channels are reaching the same people or genuinely different ones. It's also the cheapest query to build — it needs only exposure data, not purchase-path sequencing.
Second: path to conversion
Once you know your overlap, build a path-to-conversion query: the ordered sequence of touches before each purchase, across every Amazon ad product. This tells you what share of purchases are single-touch versus multi-touch, and what the most common multi-step sequences look like. It's the query that turns a vague sense of "our channels probably work together" into an actual, countable pattern — and it's the input every subsequent attribution-model decision on this site depends on.
Third: new-to-brand by placement
Break new-to-brand share out by supply source or campaign type rather than reading the account blend. This is where a lot of real signal hides — in our own DSP book, new-to-brand share ranged from 15.9% on off-Amazon exchange inventory to 42.0% on Amazon's own shopping surfaces, in the same 31-day window across 27 advertisers. A blended account-level number would have hidden that entirely. Running this query early tells you which placements are actually doing acquisition work before you make any budget decision based on a headline ROAS number.
Fourth: only now, a holdout or incrementality query
Once the first three queries have given you a real map of overlap, path structure and acquisition-by-placement, you're in a position to design a holdout test that's actually testing something specific — a suspect placement, a channel with high assist counts but weak last-touch numbers, or a segment where the path data suggests a genuine but unproven contribution. Running an incrementality query first, without that groundwork, means guessing at what to test rather than targeting the query at the account's actual open question.
The common mistake, including ours
The mistake is reaching for the most sophisticated-sounding query first — usually a data-driven attribution build or an ambitious multi-channel incrementality test — because it demos well, before running the cheaper foundational queries that would have told you whether that sophistication was even pointed at the right question. We've built a holdout test early in an engagement, before checking overlap, and had to discard part of the read when the overlap query — run afterward, out of order — revealed the holdout group itself hadn't been cleanly separated from sponsored ads exposure the way the design assumed.
A fifth query worth adding once the first four are routine
Once the core sequence is a standing habit, a spend-concentration query is worth building next: which line items or campaigns account for the bulk of spend, versus which sit in a long, thin tail. In our own DSP book, 18.3% of live line items consumed 80% of spend across a single month, while the entire zero-return tail burned only 0.7% of the budget — the opposite shape from sponsored ads waste, where a losing tail of search terms can run to a third of spend. Knowing that shape changes where you actually spend review time: on DSP, the long tail isn't where the money is, so a review that starts at the top of an alphabetical line-item list rather than a spend-sorted one is spending attention in the wrong place.
What to do if a query in this sequence comes back confusing
If the reach-and-overlap query returns near-zero overlap on an account running meaningful spend in both channels, check the join logic before concluding the channels truly never intersect — a mismatched key is a more common explanation than genuinely disjoint audiences. If the path-to-conversion query returns an oddly flat, all-single-touch distribution, check the lookback window before concluding your funnel has no real multi-touch behaviour — a too-short window will silently truncate real earlier touches out of every path.
| Order | Query type | What it sets up |
|---|---|---|
| 1 | Reach and overlap | Whether DSP and sponsored ads are double-counting shoppers |
| 2 | Path to conversion | The real sequence and touch-count distribution behind purchases |
| 3 | New-to-brand by placement | Which supply sources are actually doing acquisition work |
| 4 | Holdout / incrementality | A targeted causal test on the specific channel the first three flagged |
Which one you should actually pick
Any advertiser with AMC access and someone who can write or adapt a query can run this exact sequence themselves at no additional cost beyond the time it takes. reMKTR runs this same order on every new managed DSP account before recommending any budget change, because skipping straight to a sophisticated test has cost us real time before, as part of the same discipline behind Full Circle's $500M+ in managed Amazon spend across 100+ brands.
Shortlist on the job, not the feature grid. Pull your search-term report for the last 90 days and total the spend against terms that produced no orders — 33.6% on the account above. Then ask each vendor on your list what they would do about it in week one, and see who answers with a process rather than a screenshot.
Common questions
Do I need to run all four queries before I can trust any of them?
Not strictly, but running them in this order means each query is better-informed by the one before it — particularly the holdout test, which is far more useful once you know what you're actually testing.
How long does a typical AMC query take to build?
It varies widely by complexity and whether you're using SQL, a no-code template, or the Ads Agent natural-language interface — a reach-and-overlap query is comparatively quick; a well-designed holdout query takes considerably longer to set up correctly.
Should I start with a pre-built AMC template instead of writing custom queries?
Yes, where one exists for your question — templates lower the technical bar considerably and cover the reach, overlap and path-to-conversion use cases well. Custom queries become more necessary for bespoke incrementality designs.
What's the biggest time-waster when starting with AMC?
Building a sophisticated incrementality test before running the foundational overlap and path queries — it's easy to end up testing the wrong thing, or discovering a design flaw only after the test has already run.
We show the method before the number.
Claim the free auditRead next
- Acorn Cost: What Decides the Invoice, and What to AskPricing · acorn cost
- Skai Review: A Media Buyer's Honest ReadReview · skai review
- Pacvue Pricing: No Public Number — What to AskPricing · pacvue pricing
- Skai vs Pacvue: Contracts, Not Feature GridsHead to head · skai vs pacvue