AMC sandbox: test environment with synthetic data, limits and tables
In short: The AMC sandbox is a test environment with synthetic signals that mimic the structure and statistics of real AMC data. It suits learning, query drafts and software development without disturbing live instances. According to Amazon its results must not inform media strategy, because the data is not tied to any real campaign.
What the sandbox is for
Amazon describes the sandbox as an environment with the same functions as an AMC advertiser instance, but with synthetic signals only. This also applies to the paid feature signals from Amazon and third parties.
- Hands-on learning without risk to live executions.
- Signal investigation: with aggregation controls turned off you can look at non-aggregated rows.
- Query iteration: develop queries on smaller synthetic datasets and validate them in the advertiser instance afterwards.
- Exploring paid features, also outside the 30-day trial period.
- Software development for tool developers with or without access to advertiser instances.
Where the sandbox ends
Sandbox data is not tied to any advertiser campaign, customer identifier or ASIN. Amazon states explicitly: do not use sandbox results to inform media strategy or implementation, use them for testing, demonstration or inspiration. According to Amazon it is less suitable for:
- queries that call a specific campaign, an ASIN or custom parameters,
- queries that look for extremely granular conditions that may not exist in synthetic signals,
- queries that detect highly varied distributions or outliers,
- tests of whether compute fails on a query with a certain time window.
According to Amazon, instances created through the instant self-service APIs do not include a sandbox instance.
Which tables exist in the sandbox
The sandbox holds 13 months of backfilled signals; new data is added on a rolling time window. Status of Amazon's table:
| Table | In sandbox |
|---|---|
dsp_impressions, dsp_clicks, dsp_views | yes |
amazon_attributed_events_by_conversion_time, …_by_traffic_time | yes |
sponsored_ads_traffic, conversions, conversions_with_relevance | yes |
dsp_impressions_by_matched_segments, dsp_impressions_by_user_segments | yes |
conversions_all (Flexible Shopping Insights), experian_vehicle_purchases, foursquare_visits | yes (synthetic) |
adserver_conversions, adserver_traffic and variants | no |
audience_segments_* (Audience Segment Insights) | no |
The field lists of the main tables are in our schema reference.
Turning aggregation controls off
In the sandbox the parameter disableAggregationControls decides whether aggregation controls apply. Turned off, the query returns event rows of the synthetic signals so you can see what a report is made of. This view is not available in advertiser instances. With controls on, the sandbox simulates real reporting.
Frequently asked questions
Can I upload my own data in the sandbox?
Yes, Amazon describes uploading advertiser data in the sandbox, split into fact and dimension datasets. According to Amazon the sandbox is integrated with data upload.
May I use sandbox results for media decisions?
No. Amazon explicitly advises against it, because the data is synthetic and not tied to any real campaign.