Skip to main content
Xalicon

FinTech

Cutting reconciliation break investigation from hours to minutes

Placeholder case study. Model-assisted investigation of reconciliation breaks, with every suggestion evidenced and every decision recorded for audit.

  • Python
  • TypeScript
  • PostgreSQL
  • Kafka
  • Temporal
  • ClickHouse
  • AWS
Start a similar projectAll case studies

Sample case study — illustrative, not a delivered client engagement

Orbit Payments is an invented company used to demonstrate our case-study format and our approach to this kind of work. The technical detail reflects how we would genuinely deliver it; the client name is fictional and the figures under Results are design targets rather than measured outcomes.

Client

Orbit Payments

Illustrative payments platform — PLACEHOLDER CONTENT

Industry
FinTech
Year
2026

The business challenge

PLACEHOLDER — replace with the real engagement narrative. The illustrative situation: a finance operations team spent most of each morning matching ledger entries against processor and bank statements by hand. Breaks were resolved eventually, but the reasoning lived in spreadsheets and inboxes, so audit requests meant reconstructing decisions months after the fact.

Objectives

  • PLACEHOLDER — Reduce time spent investigating each reconciliation break
  • PLACEHOLDER — Record the reasoning behind every resolution for audit
  • PLACEHOLDER — Keep the resolution decision with a human analyst
  • PLACEHOLDER — Surface systemic causes rather than only individual breaks

Discovery and strategy

What we did before writing code

  • PLACEHOLDER — Sampled three months of breaks and categorised root causes
  • PLACEHOLDER — Timed analysts through a full investigation cycle
  • PLACEHOLDER — Reviewed audit requests to learn what evidence is actually asked for
  • PLACEHOLDER — Agreed that the system would suggest and evidence, never auto-resolve

Solution

What we built

  • PLACEHOLDER — Automated matching pass

    Replace with real detail. Deterministic rules clear the clean majority first, so analysts only see genuine exceptions.

  • PLACEHOLDER — Evidenced suggestions

    Replace with real detail. For each remaining break the system proposes likely explanations and shows the records behind each one.

  • PLACEHOLDER — Decision audit trail

    Replace with real detail. The analyst’s choice, the evidence shown and the reasoning are stored together as an immutable record.

  • PLACEHOLDER — Pattern reporting

    Replace with real detail. Recurring break causes are aggregated so the underlying integration fault can be fixed rather than re-investigated.

Architecture

How it fits together

PLACEHOLDER — Reconciliation architecture

Replace with the real architecture. The illustrative shape: deterministic matching first, model assistance only on what remains, and a human decision at the end of every path.

  1. Ledger eventsAppend-only internal record
  2. External feedsProcessor and bank statements
  3. Matching engineDeterministic rules pass
  4. Break queueExceptions with evidenced suggestions
  5. Analyst reviewHuman decision, always required
  6. Audit storeDecision, evidence and reasoning

Design approach

Decisions in the interface

  • PLACEHOLDER — Put the supporting evidence next to each suggestion, never behind a click
  • PLACEHOLDER — Made confidence explicit so analysts calibrate rather than trust blindly
  • PLACEHOLDER — Designed the queue around ageing, so nothing sits unresolved

Development process

How the work ran

  • PLACEHOLDER — Evaluation set built from historically resolved breaks
  • PLACEHOLDER — Suggestions scored against the resolution analysts actually chose
  • PLACEHOLDER — Ran in shadow mode alongside manual work before analysts saw output

Technology stack

What it runs on

  • Python
  • TypeScript
  • PostgreSQL
  • Kafka
  • Temporal
  • ClickHouse
  • AWS

Challenges resolved

What went wrong, and what we did

Every project has these. A case study that omits them is a brochure.

  • PLACEHOLDER — Suggestions risked becoming defaults

    Replace with real detail. Accepting a suggestion requires the same explicit action as any other resolution, so it stays a decision rather than a click-through.

  • PLACEHOLDER — Timing differences looked like breaks

    Replace with real detail. Settlement windows were modelled explicitly so expected timing gaps stopped entering the queue at all.

Results

Objectives and design targets

Because this is a sample case study, these are the objectives and design targets agreed in scope — not measured business results.

  • PLACEHOLDER — Investigation time

    Target only

    Placeholder. Replace with a measured, verifiable outcome or remove.

  • PLACEHOLDER — Audit evidence

    Recorded per decision

    Placeholder design target from the engagement scope.

  • PLACEHOLDER — Resolution authority

    Always human

    Placeholder. A constraint agreed during discovery.

Client feedback

In their words

Placeholder quote

This quote is sample content shown to demonstrate the layout. It is not attributable to a real person or organisation and will be replaced only with a genuine, approved testimonial.

  • “PLACEHOLDER — replace with a real, approved client quote. This text exists only to demonstrate the layout.”
    Placeholder namePlaceholder role (illustrative)

Related services

The practices behind this work

More work

Other case studies

All case studies

Similar situation?

Start a project like this one

Tell us where you are. We will tell you how we would approach it and what we would want to understand first.

Start Your ProjectTalk to us first

Prefer email? contact@xalicon.co