A real star schema
Facts and dimensions per platform rather than a pile of denormalised report rows. Adding a metric does not mean re-pulling three years of history to reshape it.
All capabilities
Data Warehouse
History that does not roll off at 60 days.
Most tools cache a recent window and call it reporting. BidHarmony keeps a star-schema warehouse per platform, backfilled as far as each API will go, so the trend line has a past and year-over-year means something.
- days backfilled where the API allows
- 914days backfilled where the API allows
- grains: campaign, ad group, keyword, product
- 4grains: campaign, ad group, keyword, product
- Amazon Marketing Stream ingest
- HourlyAmazon Marketing Stream ingest
In detail
What Data Warehouse covers
Hourly, where it exists
Amazon Marketing Stream delivers Sponsored Products, Brands and Display hourly over SQS. Dayparting is only honest when the data underneath it is finer than a day.
Backfilled on connect
Authorizing a platform starts a historical pull, so the first chart already has a past. Each API's ceiling is different and the app says which one it hit.
Down to the product
Keyword and product grain, not just campaign totals — which is the level the actual decisions get made at, and the level most exports flatten away.
Async snapshot sync
Platforms that answer report requests slowly are handled as request-then-collect rather than a blocking call, so a slow Walmart snapshot cannot stall the rest of the sync.
Honest year-over-year
Comparison periods come out of stored history rather than being extrapolated. Where history does not reach back far enough, the comparison is absent rather than invented.
Why the warehouse exists at all
Every ad platform will answer "what happened in the last 30 days". Very few will answer "what did this keyword do last November", and none of them will answer it across five platforms at once.
Keeping the data locally is what makes the rest of the product possible: rules that compare against a trailing window, blended metrics that span platforms, and charts that do not change shape because an API quietly expired its history.
Each API's ceiling is different, and named
Amazon's sales history tops out at roughly a year. Pinterest allows considerably more. Walmart's snapshots arrive asynchronously and on their own schedule.
Rather than average those into a vague promise, the app records what it actually retrieved per platform, and the charts start where the data starts.
Questions
Data Warehouse, asked and answered
Where does the data live?
In BidHarmony's own Postgres, per brand. It is not a pass-through to the platform APIs at render time, which is why the pages stay fast and why history survives an API's retention policy.
How far back does a new connection go?
As far as that platform's API permits, pulled on connect. The ceiling varies by platform and the app shows the range it actually holds.
Ready when you are
BidHarmony access is limited to active DEFY Creative & Co. clients. If you already have an account, sign in — otherwise your account manager can set one up.