Skip to content
Documentation
All docs

Analyze

ROI Report

Dollar-impact summary of what RefundSentry has saved you.

The ROI Report is the dollar-impact view of RefundSentry. It estimates how much money you’ve kept in your pocket by acting on high-confidence risk scores, and how that compares to what you pay for the app.

Hero card

One headline number: the ROI ratio (savings divided by app cost) and the dollar value of refunds prevented. Designed to be the screenshot you paste into a Slack message or a board deck.

Breakdown sections

  • Savings breakdown— Money saved by detection type (auto-holds, manual reviews, customer tags, chargeback predictions).
  • Cost-benefit comparison— Side-by-side: what RefundSentry costs versus what the prevented refunds would have cost you.
  • Monthly trend— Savings over time so you can see whether the value compounds month over month.
  • Score accuracy— A reality check on the savings estimate — how often our HIGH scores were validated by a subsequent chargeback or merchant label.
  • Phase breakdown— Real-time scoring vs. backfill-derived savings, so you can separate the install bump from ongoing value.

Conservative vs. Realistic

The estimate toggle at the top lets you switch between two assumptions. Conservative counts only the dollars tied to refunds that were actually blocked, held, or downgraded. Realisticadditionally credits soft outcomes — customers we tagged who then stopped returning, manual reviews that saved restock costs, etc. Conservative is the number to quote externally; Realistic is closer to the truth.

Methodology

A collapsible section at the bottom shows exactly how every dollar figure is computed: which subset of returns / orders is included, what counts as “saved”, and the assumptions behind the Realistic estimate. Open it before you forward the report to anyone who’ll question the numbers.

Period & export

Period selector covers Monthly / Quarterly / Yearly. The whole report exports to PDF for finance and board reporting. If your shop is too new for a meaningful estimate (fewer than ~50 scored events), the empty state explains what’s missing instead of showing misleading numbers.

Next steps