Affiliate Analytics

#Understand Affiliate through three time axes

Affiliate orders can arrive long after the first invitation. In Affiliate → Analytics, do not reduce the business to one overall conversion rate. The page separates Reachout, Approval, and Post-approval performance, each with its own starting event and observation window.

Choose the Shop scope and time window, then refresh. Portfolio counts at the top describe the current operation and are not necessarily limited by the selected historical window.

APP CAPTUREThe English light-theme Affiliate Analytics overview with Reachout cohorts, responses, and Sample application trends.

#1. Reachout begins on the real invitation date

Reachout cohorts begin when invitations were actually sent. A “response” here means a Creator submitted a Sample application, not any chat reply.

Read:

• invitations sent;
• the day applications begin to appear;
• the cumulative response curve for the same cohort;
• whether differences across dates or Campaigns persist.

Recent cohorts have not completed the full observation period and should not be compared directly with a mature cohort’s final curve. When the page says a window has no mature cohort, more time is required; it does not mean the business produced no result.

#2. Approval begins on the application date

Approval analytics group Sample applications by application date, including Approved, Merchant Rejected, Overdue, and In-flight outcomes.

APP CAPTUREThe English light-theme Affiliate Approval analytics with application outcomes and AI versus non-AI decision origins.

Also inspect decision origin:

• AI decision: completed within an authorized automation boundary;
• Non-AI decision: completed by the team or another human process.

Origin alone is not a quality verdict. Read it together with approval, overdue, downstream orders, and current policy to decide which cases can be automated and which still need human review.

#3. Post-approval uses a fixed 90-day order window

Orders can lag far behind Sample approval, so Post-approval performance uses a fixed 90-day window. It reports:

• approved applications that produced orders;
• actual units sold;
• Sample performance on a shipped-Sample basis;
• sales performance on the full Affiliate-channel basis.

These bases answer different questions and must not be treated as one denominator. Use the shipped-Sample basis to evaluate Sample strategy; use the Affiliate basis to understand the Product’s wider channel performance.

#4. Use Explore for deeper breakdowns

Switch to Explore, select a dataset and contract, then add metrics, dimensions, filters, time granularity, and a chart.

APP CAPTUREThe English light-theme Affiliate Explore builder with dataset, metrics, dimensions, filters, and chart configuration.

Follow three rules:

1. recompute rates from numerator and denominator instead of adding daily percentages;
2. keep time windows and filters consistent when comparing Campaigns, Products, or Shops;
3. do not add GMV from different datasets, because they may describe different event scopes.

Useful questions include which Campaign produces more mature applications, which Product has steadier post-approval orders, and whether one Shop’s overdue reviews cluster in a particular stage.

#5. Establish a review rhythm

Daily

• Confirm invitations are progressing.
• Check application and approval backlog.
• Resolve reviews or Escalations that block AI work in Agent Workbench.

Weekly

• Compare candidate, invitation, and application funnels by Campaign.
• Inspect approval, rejection, and overdue outcomes.
• Feed recurring questions back into Product Knowledge or Business Instructions.

Monthly

• Evaluate Campaigns and Products with mature cohorts.
• Compare downstream outcomes from AI and human decisions.
• Adjust thresholds and Approval Policies using Sample cost and order performance.

Analytics narrows the next investigation; it should not replace Campaign and relationship context. When a result looks unusual, return to Campaigns for filters and decision records, then inspect the Creator relationship for the actual conversation history.