Custom enterprise software · 2017
Branch operations scorecard: three assessment levels, rules the bank maintains itself
Operations platform for a bank branch network: teller, supervisor and lead-staff assessment, maintainable error-scoring rules, star-teller selection and certification exams — built so the business owns the rules and the system owns the arithmetic.
The situation
A commercial bank’s branch network needs to know how well each outlet is running, and it needs that answer to be acceptable to the people being measured. The Beijing branch assessed three levels — tellers, supervisors and lead staff — across its outlets, using indicators with weights, error and penalty scoring rules, and a points system. It also ran a star-teller recognition programme, handled appeals, and certified staff through examinations.
The scheme changes every period. Indicators are added or retired, weights are rebalanced, error categories are refined, and the emphasis moves as the branch’s priorities move. The assessment also has to survive contact with the assessed: a teller who receives a low score will ask why, and the answer has to be a computation they can follow, not a decision someone made.
The call we made
Separate who keeps the rules from who runs the score. The obvious build is a scorecard for this year’s scheme: hard-code the indicators and weights, produce the report, done. That version is obsolete in a quarter and the branch is back to a spreadsheet — and once the spreadsheet is back, the scores that matter are the ones in it.
So the deliverable is a rule-maintenance tool with a scoring run attached. The operations department maintains indicators, weights, error-scoring detail and points rules themselves, in the system, with no release involved. The system runs the scoring over the period’s data — assessment inputs, recorded errors, examination results, attendance — and produces per-person results. The division of labour is the design: the business owns the policy, the platform owns reproducibility.
Every adjustment is a record, not an overwrite. Scores get adjusted — a disputed error, a correction, a documented exception. Allowing an editable score makes the number meaningless and the appeal unanswerable. Each adjustment carries a reason, an author and an approval, and the original computation stays visible underneath. That is what makes an appeal process survivable: you can show what the number was, what changed it, and who decided.
Star-teller selection is a rule, not a meeting. Recognition inside a branch network runs on credibility. If the winner is chosen in a room, the programme costs more in resentment than it returns in motivation. Selection runs on the published scoring detail, so the result can be explained to the person who did not win — which is the only test that matters.
Examinations belong to the assessment, not to a separate learning system. The certification exam was already a requirement. Putting it in a separate product would have meant two records of the same competence. Question bank, paper assembly and results live in the platform, and exam scores feed certification and the assessment from one place, so training and measurement cannot disagree about what someone knows.
What we built
- Leadership overview — assessment position across all outlets on one screen: where each outlet stands, what is moving, what needs attention. This is the view branch leadership actually opens.
- Assessment management — indicators and weights, error and penalty scoring detail, points rules, and the scoring run for the period, with approval of results and an appeal path.
- Star-teller selection — rule-based selection over the published scoring detail, with the computation retained.
- Online examination — question bank, paper assembly, timed sitting, marking and results, feeding certification.
- Career and training records — training history and development records attached to the person rather than stored separately.
- Audit trail — every score, adjustment and points change logged with reason and approver.
Result
The branch runs a closed loop across three assessment levels: rules are maintained by the operations department, scoring is computed from period data, results go through approval, and appeals are answered from the record. Error scoring is flexible enough to be maintained without a release, which is what keeps it in the system instead of in a spreadsheet. Star-teller selection is transparent because it is calculated, and points and adjustments are traceable to the person who made them.
The broader effect is the one the branch cared about: assessment stopped being an annual administrative event that consumed a month, and became a standing view that leadership can look at any week. That only works when the rules can change without a project — which was the whole argument for building it this way.
Technology
| Stack | B/S web application · Java + Spring Boot · Oracle · ECharts · Question bank and online exam module |
|---|---|
| Delivery model | Private deployment, B/S application |
| Client | Bank of Nanjing — Beijing branch |
| Status | Delivered |
What we cannot show
Employee names, individual scores, outlet performance and examination content are not published. Screens show structure and configuration only.
Similar problem?
If this looks like something you are dealing with — a legacy system that must keep running, a workflow no product covers, an integration nobody wants to touch — send a short description. We will tell you honestly whether we are the right team, and what the first two weeks would look like.