Usable Data Is What Most Banks Are Missing
This is Article 1 of 4 in the Q-Lana Data and Knowledge Management series. Usable data is what most banks are missing. Four questions test whether credit decisions rest on trusted information. A regional lender shows the hidden cost of scattered data.
Usable data is what most banks are missing, even though they hold plenty of data. Every credit decision rests on three inputs. What the numbers say. What the people who know the client say. What the policy permits. When all three are present, current and consistent, a committee decides in one sitting and can defend that decision two years later. When one input is missing, the committee still decides, but with less than it thinks it has.
This article sets out four questions that test for usable data and shows what good looks like.
Usable Data is a Decision Problem
Ask a lender about data and the conversation reaches software inside a minute. Which platform, which warehouse, which integration, which vendor. That is the wrong end of the problem.
Start instead with the decisions good information should support. Take a credit committee reviewing a renewal for a grain trader. It needs the latest financials, the RM's view of the harvest and the current sector limit. If the financials are a year old, the committee is guessing with a quorum. Seen this way, the technology question becomes smaller, later and much cheaper. The real problem is decision quality, not storage or integration.
Four Questions That Test for Usable Data
Four simple questions sit behind every data discussion in a lending institution. In many banks, nobody can answer them with confidence.
i. Do we trust this information?
Not "is it in the system", but would a senior credit officer act on it without calling someone to check?
ii. Is it complete and consistent?
The same client, the same exposure, the same leverage ratio, wherever in the institution it is read.
iii. Can we reproduce this decision?
Not the outcome. The reasoning. The inputs as they stood on the day, the policy as it applied, the exception as it was argued.
iv. Can we explain it to a regulator?
In minutes, from the record, without a project.
Reading the answers
An institution that answers all four without hesitation does not have a data problem. It may have other problems, but not this one. An institution that hesitates on any question has located its problem precisely. More software licenses will not fix it.
The Familiar Symptoms of Scattered Data
The symptoms are common enough to feel normal. Client information sits in pieces across several systems. Several versions of the same financials circulate, each defensible and none authoritative. The RM's sharpest observation of the year sits in an email. A decision taken eighteen months ago cannot be reconstructed, because the file records the conclusion but not the argument. Early warnings arrive after the problem instead of before it. A bank can show all five symptoms and still describe itself, accurately, as well run.
None of that is unusual. It is close to the industry default. Which is exactly why it is worth naming: an institution can carry all five symptoms and still describe itself, accurately, as well run.
A Regional Lender With Plenty of Data

Consider a regional SME lender with a book of roughly USD 3 billion, spread across three sectors and two countries. Competent management. Reasonable systems. A credit culture people are proud of.
One client, three identifiers
The same client appears under three identifiers, because it was onboarded three times through three channels. Two teams quote different leverage ratios, each from a spreadsheet each maintains correctly. Neither team is wrong. They read different sources, and nobody has decided which one is authoritative.
There is no villain here. There is no failure of diligence. The situation built up one reasonable decision at a time.
The Cost of Misusing Data Never Looks Like a Crisis. Here's the Reason
Scattered data does not announce itself. There is no incident, no loss event, and no morning where someone discovers the damage. That is why the problem survives.
It shows up as slowness
The credit package that takes eleven days instead of four, because the analyst is reconciling rather than analyzing. It shows up as rework. The same financials keyed three times, in three formats, for three purposes. It shows up as decisions that are hard to defend. Not wrong, necessarily. Just hard to evidence when someone asks six months later why the exception was granted.
Slowness, rework and weak defensibility do not appear on any dashboard. They appear in the cost-income ratio and in the time-to-decision that a client experiences and quietly compares against a competitor. By the time the effect is visible in the numbers, it has been compounding for years.
That is the case for treating this as a priority rather than as housekeeping. The cost is real, it is already being paid, and it is invisible on the reports that management actually reads.
What Usable Data Looks Like in Daily Work

Good data is best defined by behaviour, not architecture. An institution can test itself against each of these four roles in about a minute.
The relationship manager does not hunt
No searching for the current financials, no reconciling two versions, no asking a colleague which collateral valuation is the live one. The client's profile is in one place: recent financials, collateral valuations, repayment patterns, ESG self-assessment, the record of what was discussed and agreed. Preparation for a review takes minutes, and the time saved goes back into the client conversation rather than into the file.
The risk team works from consistent inputs
PD, LGD and EAD calculated from the same source data everyone else is reading, aligned to the risk appetite framework rather than to a parallel model maintained separately. When the risk view and the business view of a client differ, the difference is a matter of judgment rather than a matter of arithmetic.
The committee receives a decision-ready package
Validation has already happened. The committee spends its time on the credit question rather than on establishing whether the pack is internally consistent. Meetings shorten. More importantly, they change character, from verification to judgment.
Compliance and audit reconstruct the reasoning in minutes
Any approval, any exception, any waiver, traced back to the inputs as they stood and the policy as it applied. Not days of retrieval. Not a project. Minutes.
The commercial effect
Together, these changes shorten credit cycles. Risk differentiation sharpens, because the institution can actually see which segments perform. Pricing discipline improves, because the inputs to price are trustworthy. Client relationships deepen, because the institution behaves like an organization that remembers. The institution grows without losing control of what it is growing.
Usable Data is a Strategic Asset

Every client relationship, credit proposal and monitoring report creates information no competitor holds. It cannot be bought or licensed, because it records how this bank has lent, to whom and with what result.
How does the record compound?
Managed well, that record gains value over time. Five years of repayment behaviour make the next scoring model sharper than any bought-in model. Structured RM notes give an early warning system something to learn from. Recorded committee reasoning helps a new credit officer become productive in months instead of years.
Waiting for scale costs more
Managed late, the same record decays. Definitions drift. Records scatter. The institution reaches scale and discovers that its most valuable asset was never assembled.
This is the argument for acting early rather than at scale. Lenders often postpone systematic data management until the portfolio justifies it. By then inconsistent definitions and scattered records have already eroded the integrity of the thing they were waiting to build. The institutions that win take the opposite path. They define the data model early, standardize the quantitative and the qualitative inputs together, and assign ownership for each domain from the start.
None of that requires a large budget. It requires a decision, and then the discipline to hold it.
What Comes Next in This Series
Naming the problem is the easy part. The harder question is what to build, in what order, and how to prove it is working before patience runs out.
The next three articles
The next article covers the eight data domains every lender needs and the smallest architecture that can hold them. The third sets out the governance and quality rules that keep the system durable. It also covers the point where most implementations give up. The fourth shows what becomes possible once the foundation is real. That runs from data products to five levels of AI, with a ninety-day plan.
Take the Usable Data Diagnostic
The four questions above open a short diagnostic. It places your institution between fragmented and good, and names the behaviours you are missing. It takes about three minutes.
The Q-Lana white paper Data & Knowledge Management for SME and Corporate Lenders sets out the full argument. It covers the domain definitions, the architecture, the governance rules and the ninety-day playbook.
Download the white paper
All Articles in This Series
- Usable Data Is What Most Banks Are Missing (this article)
- Minimum Viable Data Architecture Rests on Eight Domains
- Data Governance: Small Rules, Ruthlessly Enforced
- AI in Lending Amplifies Whatever Data Is Already There
Follow Q-Lana
One short email a week on credit risk, lending and impact finance, free. Or browse more guides in the Q-Lana Knowledge Center.
Get Q-Lana Weekly
One short email a week on credit risk, lending and impact finance, free. Or browse more guides in the Q-Lana Knowledge Center.