Running receipt verification for multiple agency clients

An agency rarely runs one receipt campaign. It runs a grocery cashback offer, a gift-with-purchase launch, and a loyalty acquisition program for different clients, often at the same time. The operational challenge is not simply reading receipts. It is keeping each client’s evidence, rules, reviewers, delivery configuration, and reporting separate while still giving the agency a repeatable way to work.

Steve and Open Loyalty can support that division of labor. Steve verifies submitted evidence and routes exceptions. Open Loyalty runs campaign and earning mechanics. A multi-client agency setup still needs discovery: do not assume that every access-control, reporting, branding, or cross-client isolation requirement is already packaged as an agency product.

Start with isolation, not templates

Treat each client’s organization context, access controls, workflows, and destination configuration as acceptance criteria. A reviewer working on Client A should not see Client B’s receipts. Confirm that separation in the proposed deployment rather than inferring it from a workflow configuration. Duplicate history also needs an explicit scope. Steve’s documented duplicate scope is one workflow’s submission history; comparing evidence across unrelated clients would require separate product, contractual, privacy, and false-positive analysis.

A reusable agency template should describe the process, not copy client data. It can include:

• required campaign intake fields;
• a default receipt schema;
• a capture-quality policy;
• standard reviewer reason codes;
• delivery acceptance tests;
• a reporting dictionary.

Then define each client workflow’s eligible products, retailers, dates, participant identity, confidence threshold, Open Loyalty target, and retention decisions. Validate each item against the actual workflow configuration available to that client.

Separate evidence rules from campaign rules

Suppose two clients both run “buy two, earn a reward.” The phrase hides different jobs.

Steve can extract configured receipt fields, return per-field confidence, and run configured duplicate and verification checks. Which retailer, line-item, quantity, and total fields are reliable must be established on representative receipts. Open Loyalty can decide whether approved observations satisfy a campaign and what the member earns. Keeping this boundary makes change control safer. A brand manager can increase a bonus without changing receipt extraction. An operations lead can adjust a configured confidence threshold without rewriting the loyalty campaign.

Question	Best home
Is the receipt readable?	Steve
What products and quantities appear?	Steve
Has this evidence or purchase identity appeared before?	Steve
Is this member eligible for the campaign?	Open Loyalty
Does two qualifying units earn 500 points?	Open Loyalty
Has the member reached a reward cap?	Open Loyalty

Design the reviewer operation before launch

“Human review is available” is not an operating plan. Decide who reviews, when, and with what authority.

Create a client-specific guide for common exceptions: cropped total, unreadable date, ambiguous product abbreviation, duplicate match, return line, and total mismatch. Define which cases a reviewer may approve or reject and when to escalate to the client. If the operation requires field correction, scope that capability separately; the current review actions should not be described as a general-purpose field editor. Use the available rule and decision reasons for reporting, and validate any additional taxonomy the agency requires.

A synthetic daily queue might contain 320 approved automatically, 18 waiting for product-match review, seven duplicate flags, and five images that need a better capture. Those numbers are examples, not expected rates. The useful question is whether the agency can explain every queue and assign ownership before the campaign starts.

Review capacity should be planned from a range. If expected volume is 10,000 receipts per week, model 2%, 5%, and 10% review rates. At three minutes per case, those scenarios require 10, 25, and 50 reviewer hours. The figures are assumptions for planning, not Steve performance claims.

Make launches repeatable

Use a launch checklist for every client:

1. Collect representative receipts from every participating retailer and channel.
2. Define the output fields and product catalogue.
3. Label difficult examples and expected verdicts.
4. Configure evidence checks and review reasons.
5. Configure Open Loyalty campaigns, earning rules, caps, and member segments.
6. Map the approved transaction or custom event.
7. Test member matching, retries, and duplicate delivery behavior.
8. Train reviewers with the labelled examples.
9. Agree reporting definitions and escalation times.
10. Run a limited pilot before media volume arrives.

If invoices are accepted, note the present input boundary: images and single-page PDFs. Multi-page PDFs require preprocessing or a separately scoped design.

Report facts at the right layer

Agency reports often collapse submission, approval, delivery, and reward into one “conversion” number. That makes problems hard to diagnose. Report the funnel in stages:

• uploads received;
• captures stopped for quality;
• processed submissions;
• approved, reviewed, rejected, and blocked outcomes;
• deliveries succeeded, skipped, or failed;
• Open Loyalty campaign outcomes, such as points or challenge progress.

This also prevents an approved receipt with a delivery failure from disappearing inside an “approved” total. Likewise, a successfully delivered purchase that earns nothing under campaign rules should not be described as a verification failure.

Avoid unsupported white-label promises

An agency may use its own campaign site, app, or chat experience and integrate by API. That does not automatically mean every Steve surface is a white-labelled agency product. Define the participant-facing capture experience, reviewer interface, support ownership, email templates, domains, and data controller responsibilities during discovery.

The strongest agency proposition is a repeatable operating model with explicit client separation, configurable evidence workflows, controlled review, and native delivery into campaigns. Confirm the required access and reporting boundaries in a pilot rather than presenting a universal agency tenancy package.

Explore the agency use case, receipt verification, and Open Loyalty integration. To scope a multi-client demo, bring one campaign brief and five difficult receipts; we can use them to design the first repeatable operating template.