Why Limit Trees Fail in Multi-LOS Environments

The Promise — What Limit Trees Were Supposed to Do

Every bank swears by its limit tree. Every system claims — “We can’t breach limits.”

And yet… breaches happen. Quietly. Daily.

Not because technology failed —but because governance did.

LIMIT TREES

The Reality — Too Many LOS, Too Little Linkage

Think of the limit tree as a family tree for credit. The root node is the borrower’s total sanctioned limit. Each child node — CC, OD, LC, BG, WCDL — draws from that root.

If one branch over-grows, the system should stop it.

In theory, that’s flawless. In practice, today’s banks don’t have one tree. They have a forest — each Loan Origination System (LOS) growing its own version.

Retail LOS, SME LOS, Agri LOS, LAS, Co-Lending LOS…Each creates its own customer record, its own limit structure, its own reality. And when all these roots push data into Core Banking, something breaks.

 The Illusion of Control — When the System Says “All OK”

Because here’s the catch…The limit tree only works if the borrower is the same person across systems.

But what if:

• SME LOS creates CIF 101 for ABC Traders

• Digital LOS creates CIF 204 for ABC Traders again

• CBS now sees two customers, each with ₹5 crore exposure

  
  

On paper, both are within limit. In reality — your exposure is ₹10 crore. The system smiles. The breach hides.

The tree didn’t fail. It grew in two gardens.

The Proof — RBI’s Group Exposure Rule vs. Ground Reality

RBI’s Master Circular on Exposure Norms (2022) makes this unambiguous:

“All exposures on a single borrower and its related entities shall be reckoned on a group basis for compliance with prudential limits.”

Yet, multiple-LOS banks routinely track at the account level, not group level. CIF duplication rates of around 8–12 % are common in internal data audits. That’s how a ₹50 crore sanctioned exposure quietly becomes ₹56 crore overnight.

No alarm. No alert. Just absence of a “Golden Source” of truth.

The Real Culprit — Not Tech, But Trust Architecture

The real issue isn’t technology — it’s trust architecture. Banks are layering LOS after LOS, each promising speed and analytics…but skipping the boring backbone of data discipline —code management.

No unified customer master. No enterprise limit registry. No real-time parent-child validation.

And so, while the board discusses AI and embedded finance, the basement leaks with silent breaches.

This isn’t just a credit control story. It’s a metaphor for digital governance everywhere —speed without structure, innovation without integration.

Let’s Talk About the Roots

If your institution runs more than one LOS, ask a simple question this week:

“Do we have one borrower master — or many?”

You might be surprised by the answer.

Let’s start that conversation —not about technology stacks, but about trust stacks.

🔖 Disclaimer

The views expressed in this article are personal and intended for educational discussion. They do not represent the policies, systems, or data of any specific bank or financial institution. References to “CIF duplication,” “limit tree,” “LOS,” or related processes are based on publicly available knowledge and broad industry observations. Actual practices and figures may vary across banks depending on their core banking systems, data governance maturity, and internal audit frameworks. Readers are encouraged to interpret this content as conceptual insight — not as an audit finding, regulatory view, or institutional disclosure.