You are validating a daily transaction report for a shared service that processes vendor payments across multiple regions. The system's automated count shows 1,250 transactions processed, but the manual log you maintain separately shows 1,287. How would you investigate the discrepancy and identify the root cause?
RoleManagement Trainee
DifficultyMedium
TopicTechnical
Asked at
Why This Is Asked
ING's online assessments and business panels test analytical rigor. This question simulates a real shared-services scenario where inconsistent reports must be reconciled quickly. It evaluates logical problem-solving, attention to detail, and the ability to handle ambiguity, core for roles in Analytics, Risk, or Finance.
General Approach
First, verify the manual log's accuracy: re-check counts for obvious errors, like missed or duplicated entries, and ensure the period covers the same time window. Second, compare transaction IDs or reference numbers line by line if feasible, or use data comparison tools (e.g., Excel VLOOKUP). Third, check for system filters (e.g., processed vs. settled) that might exclude certain transactions. Fourth, consult the team to see if any manual adjustments or batch updates occurred. Finally, if discrepancies persist, trace a sample of mismatched transactions to source documents to determine whether the system or log is correct. Communicate findings clearly, proposing corrections and process improvements to prevent recurrence. Avoid jumping to conclusions; methodically rule out each possible cause.
General Answer
Investigating a discrepancy like this, 1,250 in the automated count versus 1,287 in the manual log, a gap of 37 transactions, should follow a structured, escalating sequence rather than guessing at a cause. First, confirm both records cover the exact same scope: the same date range and cutoff time, the same regions, and the same transaction status definitions, since 'processed' in the system may not mean the same thing the manual log is counting if pending or exception-handling transactions are treated differently. Second, check the manual log itself for basic errors, duplicate entries, or transactions logged twice across a shift handover. Third, once scope and manual-log accuracy are confirmed, run a line-by-line reconciliation using transaction ID or reference number, for example a VLOOKUP-style match in Excel or a proper reconciliation tool, to isolate exactly which 37 transactions are unmatched rather than treating the gap as a single unexplained number. Fourth, look for a pattern in the unmatched transactions, such as concentration in one region, one payment channel, or one time window, since a pattern usually points to the root cause, a batch job that ran late, a system filter excluding certain transaction types, or a manual override not reflected in the system. Fifth, trace a small sample of the unmatched transactions back to source documents to confirm which side, system or manual log, is actually correct. Throughout, the goal is to rule out causes systematically and document the trail, since the appropriate fix, a process change, a system configuration correction, or additional training, depends entirely on which root cause the evidence points to.
PH Interview Gotcha
Candidates in the Philippines often start with 'Baka po mali yung manual count ko' (maybe my manual count is wrong) as a first instinct, undercutting confidence in their own data. Another pitfall is over-relying on the system as the single source of truth without questioning it. Instead, calmly assert a step-by-step verification without self-deprecation.
Sample STAR Answer▾
Key Takeaway
Systematic, step-by-step reconciliation beats jumping to conclusions when investigating a data discrepancy.
Attempt Your Answer
Write your own answer, then get instant AI feedback graded against:
- •STAR Structure
- •Specificity & Numbers
- •Ownership Language
- •PH Workplace Context
Get AI feedback on your answer — free.
3 free AI-graded answers + 1 free mock interview, no card needed.
Sign Up FreeAlready have an account? Log in
Share Your Answer (0)
Sort
Sign in to join the conversation.
No answers shared yet — be the first to show how you'd approach this.