Why another dashboard isn't enough
Imagine you're running paid traffic for several products. Your advertising dashboard reports conversions, your store reports purchases, and your payment provider records refunds and disputes. You can open each report separately, but understanding why the numbers differ still requires connecting the records. TraceKit Intelligence is designed to help teams investigate those business questions using available evidence, rather than treating one dashboard as the complete answer.
A useful question often crosses more than one system.
Start with a question you would actually ask
A business owner might ask: Which affiliates are associated with refunded orders? Why did reported conversions rise while captured payments stayed flat? Which campaigns brought customers who purchased again? The answer depends on the data available, but a focused question helps identify which records need to be reviewed.
Example: which affiliates are linked to chargebacks?
Let's use a fictional example. Your business received 12 disputes last month. Seven have confirmed attribution references associated with Affiliate A, three with Affiliate B, and two have no reliable affiliate reference. A responsible analysis shows those groups separately and makes the unmatched cases visible. It should not claim Affiliate A caused seven chargebacks merely because the recorded attribution points there.
7 linked to A + 3 linked to B + 2 without confirmed attribution = 12.
Example: which campaigns are actually making money?
Suppose Campaign A generates $10,000 in recorded sales and Campaign B generates $8,000. It may seem obvious that A performed better. But a profitability investigation also needs refunds, Product COGS, advertising spend, fulfillment, payment fees, and any relevant disputes. If some of those costs are missing, the result should explain what is known and avoid claiming verified net profit.
Sales alone cannot establish which campaign has the better margin.
Example: why don't the sales reports match?
Imagine an affiliate platform lists 100 conversions, your store shows 96 orders, and your processor reports 94 captured payments. Intelligence should help investigate whether different event definitions, delayed records, missing order references, pending payments, or reversals explain the gap. It shouldn't force the three totals to match without evidence.
What should a trustworthy answer include?
A useful investigation should show the question, the reporting period, which source records were used, how transactions were linked, and which relationships remain uncertain. The original provider references matter because they let an authorized user inspect why an order was connected to a particular acquisition source. A plausible explanation without supporting records is not enough.
Show the evidence behind the answer, not just the conclusion.
Why missing information is important
Suppose a purchase exists in your store but no matching ad click is available. The purchase still happened. TraceKit's intended approach is to preserve it as a commercial record and show that its attribution is unknown, rather than deleting it or guessing which campaign deserves credit. The same principle applies to unmatched refunds, rebills, and disputes.
Unknown is an honest status, not a failure to tell a complete-looking story.
Can Intelligence make changes to my account?
Reading information and taking action are different things. An investigation might be allowed to inspect available evidence, while changing a connection, running a sensitive operation, or modifying settings requires separate permissions and product support. User roles, client context, and production policies determine which capabilities are available. Never assume that because a tool exists, every user can run it.
Permissions and approval boundaries matter as much as the answer.
Built-in Intelligence versus external AI tools
TraceKit Intelligence is intended to be accessible inside the product. MCP is an optional way for compatible external AI assistants to access supported TraceKit tools when configured and authorized. You do not need to know what MCP stands for to understand the benefit of evidence-backed business investigation. Specific Intelligence and MCP capabilities must be confirmed against production readiness.
Built-in Intelligence is the primary experience; MCP is optional.
How to get started with a useful investigation
Begin with one concrete business question. Identify the advertising, commerce, and payment sources that hold the necessary records. Confirm that the relevant integrations and permissions are available, then inspect a small set of representative transactions. Check the explanation against source records before expanding the analysis. If something is missing, document the gap rather than assuming the answer.
Start with one question, verify the evidence, and expand from there.
Explore the connected picture.
Learn how customer journeys and supported integrations contribute evidence to business investigations.