Build or buy your Amazon data pipeline?
You can own the parsers, reconciliation and every Amazon change. Or use the maintained data layer Nova already runs.
Above the waterline
Connect the SP-API ยท week one
Below the waterline
- 1Ingest every Amazon source
- 2Reconcile numbers that keep changing
- 3Maintain the layer after launch
What do you want your team to own?
- seller_partner_idvarchar
- marketplace_idvarchar
- skuvarchar
- metric_datedate
- unitsbigint
- product_salesfloat
- net_profitfloat
- campaign_idbigint
- child_asinvarchar
- metric_datedate
- spendfloat
- salesfloat
- stock_regionvarchar
- skuvarchar
- snapshot_datedate
- fba_sellablebigint
- awd_onhandbigint
- marketplace_idvarchar
- skuvarchar
- is_buyablebool
- offers_buy_box_lostbool
- parent_asinvarchar
- search_queryvarchar
- search_volumebigint
- your_purchasesbigint
- cost_keyvarchar
- export_namevarchar
- cost_levelvarchar
- shipment_idvarchar
- skuvarchar
- shipment_statusvarchar
- quantity_receivedbigint
- parent_asinvarchar
- search_queryvarchar
- search_volumebigint
- your_purchasesbigint
- campaign_idbigint
- metric_datedate
- spendfloat
- campaign_idbigint
- keyword_idbigint
- search_termvarchar
- spendfloat
- campaign_idbigint
- placementvarchar
- metric_datedate
- spendfloat
- campaign_idbigint
- target_idbigint
- targetingvarchar
- spendfloat
- campaign_idbigint
- ad_group_idbigint
- negative_idbigint
- keyword_textvarchar
- keywordvarchar
- snapshot_datedate
- search_term_impression_sharefloat
- child_asinvarchar
- metric_datedate
- product_salesfloat
- net_profitfloat
- child_asinvarchar
- snapshot_datedate
- vendor_sellable_unitsbigint
- child_asinvarchar
- week_start_datedate
- mean_forecast_unitsbigint
- purchase_order_numbervarchar
- child_asinvarchar
- ordered_unitsbigint
- received_unitsbigint
- invoice_idvarchar
- invoice_statusvarchar
- invoice_amountdecimal
- payment_idvarchar
- invoice_idvarchar
- net_amount_paiddecimal
Drop the pipeline build. Build what people use.
The data layer is already maintained
- 15-minute refresh
- Published data refresh
- Seller, Ads and Vendor
- Amazon sources maintained together
- 99.9% accuracy
- P&L reconciled against settlements
- Maintained by Nova
- Parsers, mappings and revisions
The return is where your team spends its time
Engineering
Maintain extraction and backfills
Build the app and analysis
Product
Own Amazon edge cases and QA
Own user workflows and decisions
Time to value
Starts after the data layer works
Start application work today
A maintained subscription replaces permanent pipeline ownership.
"Nova has become a core part of how we run our Amazon agency. Data Share puts clean, structured Amazon data straight into our own stack, so we no longer maintain those pipelines ourselves. The hourly updates mean we spot changes during the day instead of managing accounts on yesterday's data."
Stephen Harris, SitrunaUK agency managing 100+ Amazon accountsWhat your team must maintain
Ingest every report without losing a day
Marketplace formats, practical rate limits and silent failures all become your teamโs responsibility.
See examplesHide examples
Ingest every report without losing a day
Marketplace formats, practical rate limits and silent failures all become your teamโs responsibility.
Singapore
The Singapore settlement report starts with Amazon's postal address. Five lines before the data. For months, no Singapore settlement made it in anywhere.
Languages
Your listings report speaks the language of your account, not the marketplace's. A German seller gets "Hรคndler-SKU", a Japanese one "ๅบๅ่ SKU", a Saudi one Arabic written right to left. Six languages, six translation tables.
Dates
The same date is written "2026-10-08" in the US, "2026/10/08" in Japan and "08.10.2026" in Germany. In India, amounts come in quotes.
Time zones
Amazon.ca counts its days in Los Angeles time. An order placed at 11:30 pm lands on the wrong day, and your "yesterday" no longer matches Seller Central.
Bursts
The docs promise 15 report requests in a row. Amazon accepts 5. The sixth is refused, and if you insist, Amazon blocks you for five minutes.
FATAL
Request your inventory report again less than 90 seconds after the last one and it fails almost every time. No message. Just "FATAL".
Empty reports
An empty report is not necessarily empty. Requested again the next day, the same report holds 270 rows.
Missing block
One missing block in a traffic report is enough to get the whole report rejected. No error. Everything is green. One day is just missing.
Token
The Amazon connection key expires after one hour. Some downloads take longer. The error looks exactly like "access denied". The job is green. The end is missing.
SKU traffic
Ask for traffic by SKU and sessions disappear for one seller in five. Amazon cannot tie a visit to a SKU.
Queue
Another tool connected to your account can block your reports. Amazon handles requests one at a time. A tool flooding the queue makes your orders report wait seven hours.
Reconcile numbers that keep changing
Your team must reopen history, translate Amazon's financial dictionary and avoid counting shared stock more than once.
See examplesHide examples
Reconcile numbers that keep changing
Your team must reopen history, translate Amazon's financial dictionary and avoid counting shared stock more than once.
Revisions
Yesterday's sales are not final. Amazon revises them for two weeks. Monday's number is not Thursday's.
Cancellations
An order can be cancelled six months later. Your March P&L changes in September.
Renamed fees
One day in June, Amazon renamed a family of fees without warning. Refunds disappeared from P&Ls. A customer is the one who noticed.
Shifted timestamps
Download a settlement report again and Amazon can send it back with every timestamp shifted by two hours. Your tool sees new fees. Everything counts twice.
Fee names
More than 400 fee names in the Finances API. 134 others in the settlement report. For the same things.
Tax
The same fee arrives tax included in one API and as net plus VAT in the other. Miss it, and you count VAT as a cost.
Duplicates
Two identical lines in a settlement are two real fees. Deduplicate "cleanly", and one settlement in two no longer adds up to the payout.
Mexico
In Mexico, the FBA price excludes tax and the merchant-fulfilled price includes it. Same column, two meanings.
Pan-EU
Your pan-European stock appears once per country. Add it up and you have five times more stock than you really do.
UAE and Saudi Arabia
Stock in the Emirates and Saudi Arabia is reported in both countries at once.
"Available"
In 2026, Amazon changed what "available" means. Units moving between warehouses vanished overnight. The result: stockout alerts everywhere, for nothing.
Maintain the layer after launch
Your first connection is not the product. Your team owns API overlap, attribution differences and every change that follows.
See examplesHide examples
Maintain the layer after launch
Your first connection is not the product. Your team owns API overlap, attribution differences and every change that follows.
Legacy API
The legacy advertising API returns the whole account, whichever seller asks. Two brands on one account: spend counted twice.
New API
The new API forgets older Sponsored Brands campaigns. For some sellers, that is half the spend.
Attribution
Sponsored Products attributes a sale over 7 days, Sponsored Brands over 14. Comparing the two is comparing apples and oranges.
The work Nova owns
The pipeline is checked again every morning
One missing order breaks trust. Nova compares every day of the available history against two other Amazon sources. When a day does not match, it becomes an alert for our team, not a discrepancy for your customer to find.
- 81
- report feeds merged
- 139
- money movements reconciled
- 23
- Amazon marketplaces
- 23
- months checked again daily
Use the clean layer wherever you work
Stream it into your warehouse with Nova Data Share, ask it in Claude through Nova MCP, or read it in the Nova App.
- seller_partner_idvarchar
- marketplace_idvarchar
- skuvarchar
- metric_datedate
- unitsbigint
- product_salesfloat
- net_profitfloat
- campaign_idbigint
- child_asinvarchar
- metric_datedate
- spendfloat
- salesfloat
- stock_regionvarchar
- skuvarchar
- snapshot_datedate
- fba_sellablebigint
- awd_onhandbigint
- marketplace_idvarchar
- skuvarchar
- is_buyablebool
- offers_buy_box_lostbool
- parent_asinvarchar
- search_queryvarchar
- search_volumebigint
- your_purchasesbigint
- cost_keyvarchar
- export_namevarchar
- cost_levelvarchar
- shipment_idvarchar
- skuvarchar
- shipment_statusvarchar
- quantity_receivedbigint
- parent_asinvarchar
- search_queryvarchar
- search_volumebigint
- your_purchasesbigint
- campaign_idbigint
- metric_datedate
- spendfloat
- campaign_idbigint
- keyword_idbigint
- search_termvarchar
- spendfloat
- campaign_idbigint
- placementvarchar
- metric_datedate
- spendfloat
- campaign_idbigint
- target_idbigint
- targetingvarchar
- spendfloat
- campaign_idbigint
- ad_group_idbigint
- negative_idbigint
- keyword_textvarchar
- keywordvarchar
- snapshot_datedate
- search_term_impression_sharefloat
- child_asinvarchar
- metric_datedate
- product_salesfloat
- net_profitfloat
- child_asinvarchar
- snapshot_datedate
- vendor_sellable_unitsbigint
- child_asinvarchar
- week_start_datedate
- mean_forecast_unitsbigint
- purchase_order_numbervarchar
- child_asinvarchar
- ordered_unitsbigint
- received_unitsbigint
- invoice_idvarchar
- invoice_statusvarchar
- invoice_amountdecimal
- payment_idvarchar
- invoice_idvarchar
- net_amount_paiddecimal
Start the app, analysis or agents today
Nova maintains the Amazon data layer underneath. Your team builds the work customers and operators use.