What is Amazon Marketing Stream?

Amazon Marketing Stream delivers Sponsored Products advertising data at hour grain rather than the daily grain of the standard reports. Marketplace Ad Pros ingests three AMS datasets — ams_sp_traffic, ams_sp_conversion, and ams_budget_usage — so you can ask about pacing, intraday performance, and budget exhaustion by hour of day.

How long does it take for Amazon Marketing Stream data to appear?

The first hourly file typically lands about 10-15 minutes after you enable Amazon Marketing Stream on a profile. After that, each closed hour appears within minutes of the hour ending.

Can I use hourly AMS data for dayparting decisions?

Yes. The playbook includes a dayparting section that reads hourly performance and turns the finding into Amazon-native bid rules, so the analysis closes the loop into optimization rules rather than staying a chart.

What happens if I ask for a window longer than 24 hours?

Each prompt explicitly names the AMS report type so the analyst does not guess. For windows longer than roughly 24 hours, the analyst routes to the daily sponsored_products_campaigns report automatically.

Is Amazon Marketing Stream billed separately?

Amazon Marketing Stream is enabled and billed per advertising profile, so you can turn it on only for the profiles where intraday data changes a decision.

Back to Home

Marketplace Ad Pros

Amazon Marketing Stream
(AMS) Playbook

Hourly Sponsored Products data through the AI analyst. Pacing, intraday performance, budget alerts, and the dayparting → optimization-rules loop that turns AMS findings into Amazon-native bid rules.

Click any prompt to copy Three AMS datasets Hour-grain data

How to Use This Playbook

Enable AMS on the profiles you want hourly data for — toggle AMS on per profile from the Integrations page. Billing is per profile, so pick the ones where intraday matters.
Wait ~10-15 minutes after enabling for the first hourly file to land. After that, every closed hour shows up within minutes.
Add MAP to your AI client — Claude, ChatGPT, Codex, or any MCP-compatible tool. See setup.
Copy a prompt below, replace <YOUR_BRAND> with your brand name, paste into your AI client, and go.

Each prompt explicitly names the AMS report_type (ams_sp_traffic, ams_sp_conversion, ams_budget_usage) so the analyst doesn't guess. For windows longer than ~24 hours, the analyst routes to the daily sponsored_products_campaigns report automatically.

Pacing & Budget

The clearest win for AMS: catching budget problems before the day is gone. ams_budget_usage fires real-time events when a campaign crosses 80% or 100%.

Campaigns near their daily budget

For <YOUR_BRAND>, use the `ams_budget_usage` report to list the top 10 Sponsored Products campaigns where the latest budget_usage_percentage exceeds 80%. Show campaign_id, budget_amount, budget_usage_percentage, and usage_updated_timestamp. Take the most recent record per campaign — do not sum across hours.

READ The "latest per scope" rule matters — budget-usage emits multiple events per day. Always filter to the most recent per campaign.

Today's spend pace vs daily average

For <YOUR_BRAND>, use `ams_sp_traffic` to sum cost by hour for today (UTC). Compare today's cumulative spend at this hour to the average cumulative spend at the same hour over the previous 7 days from the daily `sponsored_products_campaigns` report. Tell me if we're pacing high, low, or on track.

CROSS Joins AMS (today, hourly) with the daily report (prior week baseline). Best run mid-afternoon when half the day is in.

Budget-burn velocity

For <YOUR_BRAND>, use `ams_sp_traffic` to compute cost per hour for the last 6 hours by campaign. Flag any campaign whose latest-hour spend is more than 2× its 6-hour average — those are accelerating and worth checking against the budget.

READ Catches "we just lit up" moments: a sudden conversion spike that pulls Amazon's auto-bidding into a higher pace.

Intraday Performance

"How's it doing right now?" questions. The daily report doesn't know — these go straight to ams_sp_traffic.

Last 6 hours by hour

For <YOUR_BRAND>, use `ams_sp_traffic` to sum impressions, clicks, and cost for each of the last 6 hours of available data. Group by date_trunc('hour', time_window_start) and order chronologically. Tell me which hour had the highest CTR.

READ The hour that just closed lands within minutes — this is the fastest read you can get on Sponsored Products.

Day-part performance shape

For <YOUR_BRAND>, use `ams_sp_traffic` over the last 3 full days. Bucket each hour into morning (06-11), afternoon (12-17), evening (18-23), and overnight (00-05) UTC. Show impressions, clicks, cost, and CTR per bucket per day. Where is the strongest CTR window?

READ The daily report can't show this — it rolls everything up. AMS lets you see day-part shape without an AMC workflow.

Bid-change impact check

I just raised bids on campaign <CAMPAIGN_ID> one hour ago. Use `ams_sp_traffic` to compare its hourly impressions, clicks, and cost in the hour before vs the hour after the change. Did the change start moving traffic yet?

READ The classic "did my change do anything?" question. Don't wait until tomorrow — the next hour tells you.

Hourly Conversions

ams_sp_conversion ships attributed sales and conversions across 1-day, 7-day, 14-day, and 30-day windows — same dimensions as sp-traffic, so they join cleanly.

1-day attributed sales by hour

For <YOUR_BRAND>, use `ams_sp_conversion` to sum attributed_sales_1d and attributed_conversions_1d by hour for the last 12 hours. Order chronologically. Where are conversions clustering?

READ The 1-day window is the freshest — what's converting right after the click.

Hourly ACoS shape

For <YOUR_BRAND>, join `ams_sp_traffic` (cost) with `ams_sp_conversion` (attributed_sales_7d) by hour and campaign for the last 24 hours. Compute hourly ACoS per campaign. Show the 5 campaigns with the most volatile hourly ACoS.

CROSS The two AMS datasets share campaign / ad group / targeting dimensions — joins are straightforward.

Joining AMS + Daily

AMS is the right tool for "now" and "the last 24 hours." The daily reports are still the source of truth for cross-day rollups and historical analysis. These prompts use both.

Today (AMS) vs last week (daily)

For <YOUR_BRAND>, use `ams_sp_traffic` to sum today's impressions, clicks, cost so far. Then use the daily `sponsored_products_campaigns` report to get the same metrics for last Tuesday (same weekday). Compare side-by-side and show the day-over-week change.

CROSS Use AMS for partial-day "so far" totals; use the daily report for completed historical days.

Spotting late-arriving conversions

For <YOUR_BRAND>, take 3-day-old click cost from `ams_sp_traffic` (so the conversion windows have time to fill in) and join with `ams_sp_conversion` for the same window. Compare attributed_sales_1d, _7d, and _14d for that same day. Which campaigns have the biggest gap between 1d and 14d attribution? Those are slow-burn converters.

READ AMS is the only place you see all four attribution windows alongside each other at hour grain.

Dayparting & Optimization Rules

The closed loop: AMS finds the hour-of-day patterns, then you ship Amazon-native sp_optimization_rules to bid up during those windows. Amazon executes the rules on its own clock — no Claude polling required.

End-of-week dayparting scan

For <YOUR_BRAND>, use the past 7 days of `ams_sp_traffic` and `ams_sp_conversion`. For each campaign, find the top 3 hours of the day by CTR and the top 3 by lowest ACoS. Show the hour, CTR, ACoS, total clicks, and total attributed sales. Skip hour buckets with fewer than 20 clicks (too noisy to trust).

READ Run this Sunday evening. The clicks-floor filter keeps the analyst from over-fitting to a single conversion at an odd hour.

Draft SCHEDULE rules from the findings

Based on the dayparting scan above for <YOUR_BRAND>, draft a `sp_optimization_rules` payload (SCHEDULE type) for each top-performing window. Use +25% as the default boost, more if the CTR/ACoS gap is large. Show me the payloads grouped by campaign before creating anything — I want to review.

WRITE Always preview the payload before approving. Optimization rules only INCREMENT — there's no "lower bids" option, so a too-large boost only over-bids; it can't underbid.

Verify rule impact a week later

For the `sp_optimization_rules` I created last weekend on <YOUR_BRAND>, use `ams_sp_traffic` and `ams_sp_conversion` to compare the rule's target hours this past week vs the same hours the previous week. Did impressions, clicks, and 7d attributed sales go up? Is ACoS staying in range?

CROSS The whole point of the AMS → rules loop. Run this every Monday to see whether the dayparts are still the right ones.

Budget-cap hours → sp_budget_rules candidates

For <YOUR_BRAND>, use the past 7 days of `ams_budget_usage` to find campaigns that crossed 100% of their daily budget before 18:00 local time on at least 3 of those days. These are candidates for either a daily budget increase or a `sp_budget_rules` SCHEDULE boost during their peak hours.

CROSS Budget rules can SCHEDULE-boost a campaign's budget during specific time windows (same pattern as optimization rules but for budget). Useful when peak hours land before the campaign cap.

Diagnosing AMS

When the analyst says "no AMS data" or you're not sure how fresh the feed is, ask these.

How fresh is each AMS dataset?

For <YOUR_BRAND>, tell me the latest observed timestamp for each AMS report (ams_sp_traffic, ams_sp_conversion, ams_budget_usage). If any is more than 2 hours stale, flag it. Use the report metadata, not a fresh query.

READ The analyst stamps observed_latest_data_ts on every AMS sync — this is the cheapest health check.

Which AMS columns are available?

List the columns available in the `ams_sp_traffic` report for <YOUR_BRAND>'s profile, plus one sample row. I want to see what dimensions and metrics I can ask about before writing a longer query.

READ Schema and a sample row are stamped at merge time — surface them before building a complex prompt.

Ready to Try AMS?

Enable Amazon Marketing Stream on the profiles where intraday data is worth it. $5/profile/week on AI Connect or $15/profile/month on monthly plans.

Enable AMS

Per-profile billing · Disable anytime · Same MCP endpoint

Copied to clipboard

Amazon Marketing Stream(AMS) Playbook