The daily marketplace check with AI: Which tasks take priority today?

A daily marketplace check with AI starts with orders and returns that are currently open. Match each result to a specific order or item, add documented deadlines and missing evidence, and define the next check. This creates a worklist for your team. Today’s retrieval should include older open cases too; looking only at new orders is not enough.

With Backlane, you can query the data made available by your connected seller accounts. The documentation explains access, and the practical guide to marketplace analysis covers the basics. This guide describes a check that you start in the chat. It does not set up a scheduled run or notification. The prompts are templates and have not been run against a seller account.

Which four checks belong in this daily routine?

Four clearly defined tasks are enough to get started. Choose the checks that fit your connected marketplaces. This is a shared routine; it does not assume that every platform provides the same data or statuses.

Check	Specific question	Detailed guide
OTTO orders	Which items are ready to process and not yet marked as shipped?	Check OTTO order status
OTTO returns	Which returns are announced or misdirected, and what evidence is missing?	Check OTTO returns
Kaufland returns	Which return units are awaiting inspection, clarification or repair?	Analyse Kaufland returns
Mirakl orders	Which orders are awaiting acceptance or shipment, and which have an incident?	Check Mirakl order status

This selection complements sales and advertising analysis. A revenue chart does not tell you which individual return item needs to be checked against warehouse records today. Equally, an operational checklist is not a complete assessment of your business’s performance.

How do you define the scope clearly?

Before retrieval, record the connection, marketplace and, where relevant, shop, Kaufland storefront and fulfilment type. Use one reporting time zone, such as Europe/Berlin. Each source retains its own retrieval time: several interfaces queried in sequence do not provide a snapshot taken at precisely the same moment.

For open tasks, you need the entire available open workload. Avoid a lower boundary such as “created since today” if it would exclude older open cases. The four specialist guides explain the relevant filters and complete pagination. If a retrieval is partial, label that area “incomplete” rather than presenting an apparently complete total.

Also define what each entry represents. OTTO position items, Kaufland return units and Mirakl orders are different units. Keep their counts separate. The same identifier on two marketplaces does not mean the same case.

How do you turn a status into a justified priority?

A priority needs a traceable reason. For example, your team could use the following working rule:

Observation	How to classify it in the checklist	Next check
A supplied shipping deadline has passed or is imminent.	Review first; include the time and source.	Compare dispatch evidence, warehouse records and the marketplace confirmation.
There is an open incident or a discrepancy between shipment and processing status.	A case needing clarification, with specific evidence missing.	Identify the affected item and the person responsible.
A processable item has been open for some time.	Sort by the documented time field.	Check progress; do not invent a deadline from its age.
Retrieval fails or returns only part of the data.	The checklist’s coverage is uncertain.	Restore access or complete the retrieval.

This is a suggestion for organising work, not a priority order imposed by the marketplaces. Known agreements and specific promises to customers can change the order. Any assumption about urgency should be labelled as an assumption in the report.

The differences matter: Mirakl provides a dedicated shipping_deadline field; the expected delivery date of an OTTO item is a different piece of information. The definitions appear in Mirakl’s order schema and the OTTO API Index. Do not combine these values into an apparently uniform shipping deadline.

Which prompt creates the combined worklist?

The following request brings the four checks together. Include only the areas relevant to your accounts. The output is limited to ten suggestions; the underlying open lists must still be retrieved in full.

Create today's marketplace check using read-only access.
Scope: [OTTO connection], [Kaufland connection with storefront
and fulfillment_type], [Mirakl connection with shop].
Skip areas that have not been selected. Report in Europe/Berlin.

Check OTTO PROCESSABLE items using AT_LEAST_ONE and then filter
the individual items; also check ANNOUNCED and MISDIRECTED returns.
Check Kaufland return units in need_to_be_returned, return_arrived,
return_in_clarification and return_in_repair;
add their return status.
Check Mirakl WAITING_ACCEPTANCE and SHIPPING, plus a separate
has_incident=true query without restricting it to those order statuses.
Use the available interface definitions to implement the checks.

Do not lose older open cases through lower date boundaries.
Read every page while respecting request limits. Keep counting units
and retrieval times separate for each source; flag partial retrievals.
Use connection, shop/storefront and the original identifier for every match.

Suggest no more than ten next checks, each with:
marketplace, case/item, reported finding, documented deadline
or other time field, missing evidence and next step.
Explain the order: a specific deadline, need for clarification or age.
State what each time field means; do not infer time in the current
status from a creation or update date. Flag missing values.
Count additional cases not displayed separately for each check area.
Combine the same case with several reasons for checking,
without losing individual items or reasons.

Do not invent deadlines, causes or responsibilities. Do not display
names, addresses, contact details or customer free-text content.
Do not change anything or send messages.

Restricting the output does not replace access controls. If you want to enforce read-only use technically, you can disable write access across your Backlane organisation; the setup documentation explains permissions.

What does a good daily report look like?

Fictional example: Three checks appear in the report for 30 October at 09:00. All identifiers and findings are invented.

Case	Finding and interpretation	Next step
Mirakl EXAMPLE-M1	SHIPPING, reported shipping deadline of 08:00 has passed	First compare dispatch evidence with the marketplace confirmation.
OTTO EXAMPLE-O1	Return announced, shipment DELIVERED, warehouse receipt unverified	Check goods receipt; a carrier scan alone is insufficient.
Kaufland EXAMPLE-K1	Return unit return_in_clarification	Identify the unresolved question and person responsible in your own process.

The fact that OTTO does not necessarily interpret DELIVERED as warehouse receipt is explicitly documented in the ReturnShipments documentation. Sources like this make the difference between a useful checklist and a premature decision.

How do you later check what has been resolved?

Keep the first list, including its identifiers and retrieval times. A new query provides the current state. That alone cannot reliably tell you what happened since the morning; you need to compare it with the saved snapshot.

Recheck the cases from this earlier worklist using read-only access:
[list with connection, shop/storefront, identifier, finding and retrieval time].

Compare each earlier finding with the current data.
Distinguish: status unchanged, status changed, evidence still missing,
not found, or retrieval failed.
A case that is no longer found is not automatically resolved.

For each change, state the old and new values and both retrieval times.
Confirm completion only for the specific step supported by the evidence.
For outstanding questions, propose the next check.
Do not display buyer data or make any changes.

This keeps the daily check manageable: current findings, a justified order and a follow-up that can be verified. The detailed status and data rules remain available through the four linked specialist guides.