The bank's best customers had been hiding inside its own systems for decades. A single household view surfaced 850 of them and cut a week of monthly KYC reconciliation to one day.
INDUSTRY
Community banking
BUSINESS PROFILE
92,000+ customer relationships · 79 employees · Third-generation family-owned · Chartered 1965
Midwest Community Bank is a family-owned community bank chartered in 1965 with offices in Freeport, Rockford, and Chicago, Illinois. Now in its third generation of family ownership, the bank serves more than 92,000 customer relationships across retail, small business, and agricultural segments in northern Illinois with a team of 79 employees.
When a long-term customer walked into the Freeport branch in October 2024 asking about a HELOC, the relationship manager pulled his record and saw a checking account opened in 1998. What the system did not show was that the same customer also held a business credit card, a commercial line of credit on his auto repair shop, and a 15-year mortgage paid off the previous year. All four products lived in separate systems, and none of them were linked to a single household.
The relationship manager treated him as a new HELOC applicant and quoted standard rates.
That kind of scene played out routinely at Midwest Community Bank, and Anne Hartzell, the bank's Vice President of Operations, had been quietly cataloging examples for almost two years.
The root cause sat in the bank's plumbing. The core banking platform held checking, savings, and CD records. The loan origination system held mortgages, HELOCs, and consumer loans.
A third system, bolted on years earlier to support commercial banking, held lines of credit, business cards, and treasury services. Each platform assigned its own customer ID, and no shared identity layer connected them.
The bank had tried to fix the problem twice. The first attempt used SQL JOIN logic written against name and zip code, which produced obvious matches and missed everything else.
The second attempt used a desktop fuzzy-matching tool the operations team had purchased for around $4,000 per seat. It worked for one-time merge purges on small lists, but it could not connect to the bank's databases, would not handle joint accounts where the same individual appeared in two name positions, and required the operations analyst to export, match, and re-import every quarter.
By late 2024, the issue forced its way out of operations. The bank's BSA officer faced a customer due diligence review that required demonstrating know-your-customer relationships at the household level. The fragmented customer master would not pass examiner scrutiny.
Treating one person as three different people across systems was no longer a service issue alone. It was a compliance posture issue.
Midwest Community Bank evaluated three vendors over five weeks. The same anonymized 142,000-record extract ran through each tool, scored against a gold-standard set of 600 known matched and unmatched pairs the operations team had built by hand over the previous year.
MatchLogic won the evaluation. The bank licensed the API tier so the matching engine could run both as a nightly batch job across the three systems and as a real-time lookup at the point of account opening. Four factors the other tools could not match decided the choice.
MatchLogic combines Jaro-Winkler distance for name similarity with rules-based field comparison for tax IDs and account numbers. Every linkage carries a traceable rule explanation. “The system said so” was never going to convince a federal examiner, but a documented logic trail for every household match decision could.
The team defined three tiers: Tier 1 auto-linked records that matched on tax ID with at least 92% fuzzy match on legal name; Tier 2 linked records that matched on fuzzy name plus exact zip plus at least 85% address similarity, after operator review; Tier 3 surfaced lower-confidence candidate pairs for periodic batch review. Surfacing 80% of the value with minimal false positive risk was worth more than chasing the last 20% with any risk of falsely linking two unrelated customers.
Midwest Community Bank's core processor contract had clauses about where customer PII could be sent. A cloud-based matching service would have required a new vendor risk review, a new data processing agreement, and a regulatory conversation the bank did not want to have with a due diligence review already on the calendar. MatchLogic ran on a server inside the bank's existing data center, with no outbound calls.
The product's drag-and-drop pipeline let Hartzell's three-person operations team build, test, and rerun the matching workflow without filing a ticket with the bank's outsourced IT services provider. Every IT change at Midwest cost money and weeks of lead time. Owning the workflow internally was a meaningful operational advantage.
The rollout ran six weeks from kickoff to production.
Profiling came first. MatchLogic's solutions team ran the visual profiler across extracts from all three systems and found that 27% of records had inconsistent address formatting between systems, 14% had name fields entered in different orders, and 6% had data quality issues serious enough to flag for manual cleanup before any matching ran.
The operations team then built and validated the three-tier match rules against the bank's gold-standard set. MatchLogic identified 31% more true relationships than the bank's prior SQL-based approach while producing one false positive across the 600-pair validated set. Joint account holders, formerly invisible to the bank's relationship view, accounted for a disproportionate share of the lift.
By the final week, the team operationalized the workflow. A nightly job pulls deltas from each system, runs them through the cleansing pipeline, scores candidate matches, and writes household IDs back to a central customer dimension table that feeds the bank's reporting layer. New account openings hit the matching engine in under 400 milliseconds via a REST call from the core's onboarding screen.
The bank's relationship view now connects checking, savings, lending, and treasury products under a single customer identity. The 31% lift in identified relationships over the prior SQL approach came mostly from joint accounts and from small business owners whose personal and business records had never been linked. Every legacy customer ID was preserved and mapped to the unified household ID, so downstream reporting and statement generation kept working without changes.
Once households were unified, Midwest's product team re-ran customer lifetime value reporting at the household level. The exercise revealed 850 households whose combined deposit, loan, and treasury balances exceeded $250,000 but had never appeared in the bank's high-value relationship retention program.
The aggregated under-recognized balance came to $36.5 million. The relationship management team has begun a structured outreach to those 850 households with tailored product reviews.
The bank's BSA officer had previously spent a week each month reconciling customer due diligence records across the three systems. The unified household view now feeds directly into the CDD dashboard, and the monthly review completes in under a day. The recovered analyst capacity has been redirected to enhanced due diligence work on higher-risk customer segments, which the bank's risk committee had been pushing for over the previous year.
Midwest Community Bank's relationship managers now greet customers on the phone with the full picture of their household relationship rather than a single account record. The data governance team has begun extending the matching pipeline to commercial deposit data and trust account records, with agricultural lending records scheduled for the next phase. For a 79-person community bank competing against regional and national banks, a unified customer 360 has moved from aspiration to operating reality.
The difference was immediate. In the first month, we stopped creating about 1,400 duplicate patient records that our old system would have missed. That's 1,400 kids whose medical histories stayed intact instead of getting split across two profiles.
Founded in 2003, MatchLogic has matched over 2 billion records across 4,500+ enterprise installations, powering production-scale entity resolution pipelines.
MatchLogic combines deterministic and fuzzy matching with visual profiling and drag-and-drop cleansing workflows to identify duplicates and maintain trusted golden records.
Built for fintech, healthcare, insurance, and government teams, the platform provides REST API integrations, cross-platform deployment, and full auditability across every match operation.
.svg)
