The benchmark question set for LeadInsight's Ask-the-Data semantic retrieval (EmbeddingGemma + pgvector over the gold data dictionary). Ground truth is derived from the tables behind the published dashboards, phrased three ways — plain business language, banker jargon, and casual/vague wording. A question scores a hit when any accepted table ranks inside the retrieval window that fills the SQL prompt. Current scores (2026-07-12): hit@1 90% · hit@3 97% · hit@5 98.5% · MRR 0.938 · 0 misses in top-10 (from a 2026-07-11 baseline of 69% / 85% / 93% / 0.782 before description tuning).
Maintenance: when a new dashboard ships, add 2-3 questions for its tables and re-run the eval before and after any description change.
| # | Question | Accepted table(s) |
|---|---|---|
| 1 | How many active loan customers do we have? | ops_loan_portfolio |
| 2 | What is our total principal outstanding? | ops_loan_portfolio |
| 3 | Show me the active loan book by branch | ops_loan_portfolio |
| 4 | Which branches have the biggest loan portfolio? | ops_loan_portfolio, ops_branch_summary |
| 5 | What is our PAR30? | ops_par_rollup |
| 6 | Portfolio at risk by zone | ops_par_rollup |
| 7 | Which branches have the worst arrears? | ops_par_rollup, ops_loan_portfolio |
| 8 | How much of the portfolio is overdue more than 30 days? | ops_par_rollup |
| # | Question | Accepted table(s) |
|---|---|---|
| 9 | How much did we disburse last month? | ops_loan_disbursement |
| 10 | Loan disbursement by zone for 2026 | ops_loan_disbursement |
| 11 | How many new loans were booked in June? | ops_loan_disbursement |
| 12 | Month on month disbursement trend | ops_loan_disbursement |
| # | Question | Accepted table(s) |
|---|---|---|
| 13 | How much did we collect in repayments this month? | ops_loan_repayment_events |
| 14 | Collections by branch last week | ops_loan_repayment_events |
| 15 | What repayments actually came in yesterday? | ops_loan_repayment_events |
| 16 | How much repayment is due this month? | ops_loan_expected_repayment, ops_loan_schedule_aging |
| 17 | Expected collections for next week | ops_loan_expected_repayment |
| 18 | What is our collection rate? | ops_loan_expected_repayment, ops_loan_repayment_events |
| 19 | Cash flow forecast from loan repayments | ops_loan_expected_repayment |
| 20 | Forward ageing of amounts falling due | ops_loan_schedule_aging |
| # | Question | Accepted table(s) |
|---|---|---|
| 21 | How many new clients did we onboard this month? | ops_customer_onboarding, ops_customer_onboarding_summary |
| 22 | Client acquisition trend by zone | ops_customer_onboarding, ops_customer_onboarding_summary |
| 23 | How many customers joined last quarter? | ops_customer_onboarding |
| # | Question | Accepted table(s) |
|---|---|---|
| 24 | Total savings balance by branch | ops_savings_listing, ops_deposit_class_summary |
| 25 | List savings accounts with balances over 1 million | ops_savings_listing |
| 26 | How many savings accounts do we have? | ops_savings_listing, ops_branch_summary, ops_deposit_class_summary |
| 27 | Term deposit portfolio size | ops_td_mobilisation |
| 28 | Show me all fixed deposits maturing this month | ops_td_mobilisation |
| 29 | Deposit mobilisation against target | ops_deposit_mobilisation |
| # | Question | Accepted table(s) |
|---|---|---|
| 30 | How many branches do we have? | dim_branch, ops_branch_summary |
| 31 | List all branches in Lagos zone | dim_branch |
| 32 | Branch performance summary | ops_branch_summary |
| # | Question | Accepted table(s) |
|---|---|---|
| 33 | KPI scorecard for zones | ops_kpi_scorecard |
| 34 | Which RMs are hitting their targets? | ops_cso_scorecard, ops_kpi_scorecard |
| 35 | Account officer performance against budget | ops_cso_scorecard |
| 36 | What are the 2026 budget targets per zone? | ops_budget_target |
| 37 | Target versus actual for disbursement | ops_budget_target, ops_kpi_scorecard |
| # | Question | Accepted table(s) |
|---|---|---|
| 38 | Mobile app transaction volume this month | ops_digital_channel_daily_summary, ops_digital_channel_postings |
| 39 | Digital channel activity by branch | ops_digital_channel_daily_summary, ops_digital_channel_postings |
| 40 | USSD versus mobile banking usage | ops_digital_channel_daily_summary, ops_digital_channel_postings, ops_channel_scorecard |
| 41 | Digital channel scorecard by branch | ops_channel_scorecard |
| 42 | How many ATM cards have we issued? | ops_card_register |
| 43 | POS terminals deployed by branch | ops_pos_terminal |
| # | Question | Accepted table(s) |
|---|---|---|
| 44 | Which customers closed their accounts this year? | ops_closed_clients |
| 45 | Customer age and gender distribution | ops_customer_demographics, ops_customer_profile_summary |
| 46 | Loans that were fully paid off last month | ops_loan_liquidation |
| 47 | When did each delinquent client last make a payment? | ops_loan_last_payment |
| 48 | Group lending summary | ops_group_summary |
| # | Question | Accepted table(s) |
|---|---|---|
| 49 | Show me the trial balance for June | fin_trial_balance, fin_trial_balance_asreported |
| 50 | What was our profit last month? | fin_trial_balance, fin_income_statement |
| 51 | Total interest income year to date | fin_trial_balance, fin_income_statement |
| 52 | Balance sheet as at end of June | fin_balance_sheet, fin_trial_balance |
| 53 | Operating expenses by GL account | fin_trial_balance |
| 54 | Trial balance by branch | fin_trial_balance_branch |
| 55 | Has the June period been closed? | fin_period_close_status |
| 56 | CBN returns for savings and loans | fin_cbn_saving_and_loan |
| 57 | Loan loss provisioning per CBN ageing buckets | fin_cbn_loan_expected_ageing |
| 58 | What is our total asset position? | fin_balance_sheet, fin_trial_balance |
| 59 | Income versus expenses trend over the last 6 months | fin_trial_balance, fin_income_statement |
| 60 | GL postings for a specific account | ops_account_postings, fin_trial_balance_asreported |
| # | Question | Accepted table(s) |
|---|---|---|
| 61 | Who are our dormant savings customers? | ops_savings_listing |
| 62 | Number of clients per account officer | ops_loan_portfolio, ops_cso_scorecard |
| 63 | Value of transactions passing through our channels daily | ops_digital_channel_daily_summary, ops_digital_channel_postings |
| 64 | Write-offs in the loan book | ops_loan_writeoff, ops_loan_portfolio, ops_loan_schedule_aging |
| 65 | Interest collected versus principal collected | ops_loan_repayment_events |
| 66 | Which loan products perform best? | ops_loan_portfolio, ops_loan_disbursement |
| 67 | Average loan size by product | ops_loan_disbursement, ops_loan_portfolio |
| 68 | Overdue loans needing follow up today | ops_loan_portfolio, ops_par_rollup, ops_loan_last_payment |