CASE STUDY

Credit Risk to Collections Transformation.

11 markets. 3 bots. 105 days.
Credit Policy and Collections Transformation
Industry: Food Delivery / E-commerce
Organisation Size: Regional APAC Operations
Geography: Southeast Asia, South Asia, East Asia
My Role: Manager, Transformation and Automation
Engagement Type: Process Reengineering, Policy Implementation, RPA Deployment
THE CASE
Problem

An e-commerce organisation operating across APAC was absorbing a growing volume of bad debt year over year. Payment defaults had been accumulating with no structured mechanism to prevent, manage or recover them. The impact was showing on the balance sheet and the P&L.

The root cause was a governance gap at the point of vendor onboarding. Credit and payment terms had never been standardised. Each market and each vendor team operated on their own commercial terms, extended credit at their own discretion and followed up on overdue payments in their own way. There was no policy, no threshold, no approval framework and no enforcement mechanism.

The problem had escalated to the highest levels of finance leadership. A cross-functional taskforce was assembled involving six senior stakeholders across Finance, Commercial and Legal, led by the Transformation team.

Impact

The consequences were operational as well as financial.

Without standard credit terms, every vendor negotiation was a fresh exercise in improvisation. Without approval thresholds, credit extensions were granted at the discretion of individuals rather than governed by policy. Dunning notices were sent manually with no defined intervals, making follow-up inconsistent and legally unenforceable. Collection agents were not engaged until debt had aged beyond recovery. Legal notices were avoided because there was no documented basis to issue them.

Across 11 markets, the same seven gaps appeared in varying degrees. The problem was not isolated to one market or one team. It was systemic.

Approach

Bad debt write-offs had been escalated to finance leadership before the engagement began. The taskforce was assembled to understand why the problem existed and what needed to change. The diagnostic work revealed seven specific failure points:

  1. No standard credit and prepayment terms
  2. No maximum credit thresholds
  3. No approval framework for deviation from standard terms
  4. Manual dunning with no defined intervals
  5. Hesitation to issue legal notices due to absence of documented basis
  6. Collection agents not engaged at an early stage
  7. Bad debts written off year over year without a structured recovery process

The first intervention was upstream. The Credit, Pre-payment and Collection Policy had been developed by Finance and Commercial leadership. The next challenge was operationalising it. A Credit Application Process was designed from scratch as a greenfield exercise, building the approval architecture that the policy required but did not yet have. Every credit request now had to pass through a defined approval chain before credit could be granted. A Credit Application Request form was built and automated to route requests through the appropriate review and approval stages.

With credit controls established at the point of origination, the downstream Accounts Receivable process was examined end to end. Payment monitoring, AR ageing report generation, reconciliation, SAP write-off posting, dunning, legal escalation and third party debt collection were all mapped, assessed and reengineered. The reengineering exercise ran for 45 days, followed by structured training across Commercial and Vendor teams.

The automation phase began only after the reengineered processes had been observed in operation for 30 days. Three RPA bots were then designed, built and deployed:

  1. Bot 1 automated AR Ageing Report generation across Restaurants and Corporates
  2. Bot 2 automated the reconciliation and compilation of ageing reports into the master file
  3. Bot 3 automated the preparation and dispatch of the collection list to third party debt collectors

SAP dunning notices were configured with defined intervals, replacing the previous manual process.

OUTCOME
What this engagement delivered:
11
APAC Markets Covered
3
RPA Bots Deployed
7 of 7
Process Gaps Closed
3,000+
Hours Saved Annually
105 Days
Full Engagement Duration
What changed for the organisation:
For the first time across all markets, the organisation had a documented and enforceable basis to prevent credit overextension, pursue overdue receivables and engage legal escalation. The governance and process gaps identified in the root-cause analysis were structurally addressed.
Standardised credit and prepayment terms adopted across all 11 APAC markets.
Approval framework established for credit deviations, removing individual discretion from commercial negotiations.
Dunning process automated with defined intervals, replacing inconsistent manual follow-up.
Legal escalation pathway formalised, giving the collections team a documented basis to act.
Collection agents engaged earlier in the process, improving recovery rates before debt aged beyond viable collection.
LEARNING

What this engagement confirmed.

Policy without process is incomplete. The credit policy drafted by Finance and Commercial was necessary but not sufficient on its own. The real work was in translating that policy into redesigned processes that people could actually follow. Without the process reengineering, the policy would have sat in a shared drive and changed nothing.
The value of observation before automation. Automating a broken process produces faster broken outcomes. The decision to observe the reengineered process for a full month before engaging the RPA team meant that the bots were built on stable, validated workflows.
Cross-market standardisation requires trust before prescription. Mapping 11 markets revealed that every deviation had a local justification. The work was not in identifying the gaps. It was in building enough trust with local teams that they were willing to adopt a standard they had not designed. That required listening before prescribing.