Data health
How much to trust each number.
- 34 of 34
- breakdown tables that add up exactly to their totals
- 81.0%
- of 12,673 indicator × year cells have a value
- 2,402
- cells are blank in the source (never filled or estimated)
- 194
- indicators whose unit the source does not state
Completeness
Blank cells, sheet by sheet
| Sheet | Indicators | Values | Blank | Filled |
|---|---|---|---|---|
| Revenue by TaxHead | 68 | 640 | 40 | 94% |
| MacroEconomic | 31 | 292 | 18 | 94% |
| Tax Compliance | 49 | 373 | 117 | 76% |
| SectorRevenue | 33 | 330 | 0 | 100% |
| Customs Trade | 55 | 468 | 82 | 85% |
| Enforcement | 31 | 297 | 13 | 96% |
| Admin HR | 51 | 423 | 87 | 83% |
| Excise VAT Detail | 54 | 453 | 87 | 84% |
| StampDuty Vehicles | 24 | 223 | 17 | 93% |
| Audit Enforcement | 95 | 663 | 192 | 78% |
| DTS Digital Tax Stamps | 64 | 380 | 324 | 54% |
| EFRIS Digital Invoicing | 18 | 125 | 19 | 87% |
| Presumptive Tax Register | 261 | 2,325 | 285 | 89% |
| Taxpayer Registration Detail | 400 | 3,279 | 1,121 | 75% |
Blank is measured against every indicator × fiscal-year column the sheet declares. Many blanks are structural, not errors: a series that starts later (system tracking began, a table was added), a partial FY2025/26, or long-tail district and country rows that report intermittently. A blank is shown as “not available” everywhere, never as zero.
Known caveats
Flags carried beside the data
Re-exports jump to UGX 13,392.73bn in FY2022/23, roughly ten times the neighbouring years, while imports, exports and the trade balance show no comparable movement. It looks like a data anomaly, not a real trade event, and should be checked against the original URA Data Book table before it is used in any headline claim.
Source: STORY_DISCOVERY.md #20
The EFRIS registered-taxpayer series spikes to 677,914 in the mandatory-rollout year (FY2020/21) and then falls sharply. This most likely reflects a change in what “registered” counts (new registrations that year against cumulative active users), not a literal collapse in adoption. It is held out of any growth claim until the source definition is confirmed.
Source: STORY_DISCOVERY.md #14
Fees & Licenses (and its sub-items) report values for FY2015/16–FY2020/21 and then go blank; in the same year Non Tax Revenue jumps from UGX 504.81bn to UGX 1,232.37bn. A reclassification of the fee lines into NTR is plausible but is an inference, not confirmed by the source. Do not chart these as one continuous series without marking the break. Appropriation in Aid reports only FY2019/20 and FY2020/21.
Source: DATA_AUDIT.md §4.2
FY2025/26 is a partial year in the operational-system sheets, so it is not comparable with the full years before it.
Source: Workbook README (coverage)
Total Invoices Issued is blank from FY2023/24 onward: the figure was not available at extraction. A blank is not a zero.
Source: DATA_QUALITY_REPORT.md
EFRIS carries a small “Pre-2019/20 (Undated)” bucket (0.02%–0.56% of totals depending on the measure), kept as an explicit extra column rather than dropped.
Source: Workbook README (data-quality flags v)
The EFRIS “by access channel” and “by VAT rate” tables total to registration counts, not invoice counts. They are not invoice-level breakdowns.
Source: DATA_AUDIT.md §4.5
Twelve of eighteen commodities show no data before FY2024/25: that is the system-tracking start date, not a processing error.
Source: DTS sheet source note; DATA_AUDIT.md §4.6
In Source A, the registration and cancellation series are numerically identical in the raw export (verified; a source-system data-quality artefact), retained as provided. Do not present cancellations as an independent series without this caveat.
Source: DATA_AUDIT.md §4.5
The F/M field is interpreted as gender; this interpretation is not independently confirmed.
Source: DATA_AUDIT.md §4.5
GDP and the ratios built on it are undefined for FY2015/16 because there is no prior-year base in the source.
Source: DATA_QUALITY_REPORT.md
Reconciliation
34 of 34 breakdown tables reconcile
For every table that has a Total row, the other rows in that section are summed and compared with the Total for each fiscal year (tolerance: the larger of 0.5 or 1% of the total). Every group passes with no mismatch. It is strong evidence the category-level figures were transcribed correctly from URA’s source systems.
Show all 34 checks
| Sheet | Section | Rows summed | Years | Result |
|---|---|---|---|---|
| Audit Enforcement | AUDIT CASELOAD BY TAX OFFICE (COUNT) | 6 | 9 | Adds up |
| Audit Enforcement | AUDIT ADJUSTMENT VALUE BY TYPE (UGX BN) | 3 | 9 | Adds up |
| Audit Enforcement | AUDIT CASELOAD BY STATION (COUNT) | 30 | 9 | Adds up |
| Audit Enforcement | AUDIT CASELOAD BY WORKFLOW STATUS (COUNT) | 13 | 9 | Adds up |
| Audit Enforcement | AUDIT CASELOAD BY AUDIT TYPE (COUNT) | 4 | 9 | Adds up |
| Audit Enforcement | AUDIT ADJUSTMENT VALUE BY AUDIT TYPE (UGX BN) | 4 | 9 | Adds up |
| Audit Enforcement | AUDIT ADJUSTMENT VALUE BY TAX HEAD (UGX BN) | 12 | 9 | Adds up |
| Audit Enforcement | AUDIT CASELOAD BY TAX HEAD (COUNT) | 12 | 9 | Adds up |
| DTS Digital Tax Stamps | EXCISE DUTY BY COMMODITY (UGX BN) | 18 | 11 | Adds up |
| DTS Digital Tax Stamps | DELIVERY QUANTITY BY COMMODITY (UNITS) | 18 | 11 | Adds up |
| DTS Digital Tax Stamps | DUTY PAYABLE BY COMMODITY (UGX BN) | 18 | 11 | Adds up |
| EFRIS Digital Invoicing | EFRIS REGISTRATIONS BY ACCESS CHANNEL (COUNT) [CORRECTED LABEL — see data_dictionary] | 5 | 8 | Adds up |
| EFRIS Digital Invoicing | EFRIS REGISTRATIONS BY VAT RATE CLASSIFICATION (COUNT) [CORRECTED LABEL — see data_dictionary] | 5 | 8 | Adds up |
| Presumptive Tax Register | BY REGISTRATION RECORD TYPE (COUNT) | 2 | 10 | Adds up |
| Presumptive Tax Register | BY TAXPAYER STATUS (COUNT) | 3 | 10 | Adds up |
| Presumptive Tax Register | BY TAXPAYER CATEGORY (COUNT) | 2 | 10 | Adds up |
| Presumptive Tax Register | BY DISTRICT (COUNT) | 124 | 10 | Adds up |
| Presumptive Tax Register | BY TAX STATION / OFFICE (COUNT) | 82 | 10 | Adds up |
| Presumptive Tax Register | BY ECONOMIC SECTOR — ISIC CLASSIFICATION (COUNT) | 22 | 10 | Adds up |
| Presumptive Tax Register | BY FILING / OBLIGATION FLAG (COUNT) | 3 | 10 | Adds up |
| Presumptive Tax Register | BY TAX REGION CATEGORY (COUNT) | 3 | 10 | Adds up |
| Presumptive Tax Register | BY BUSINESS CATEGORY (COUNT) | 8 | 10 | Adds up |
| Taxpayer Registration Detail | TAX-TYPE REGISTRATIONS BY STATUS CODE (COUNT) | 1 | 10 | Adds up |
| Taxpayer Registration Detail | TAX-TYPE REGISTRATIONS BY TAX TYPE (COUNT) | 13 | 10 | Adds up |
| Taxpayer Registration Detail | TAX-TYPE CANCELLATIONS BY TAX TYPE (COUNT) | 13 | 10 | Adds up |
| Taxpayer Registration Detail | CUMULATIVE TAX TYPES HELD, BY TAX TYPE (SUM OF COUNT) | 13 | 10 | Adds up |
| Taxpayer Registration Detail | NEW TIN REGISTRATIONS BY STATUS (COUNT) | 3 | 11 | Adds up |
| Taxpayer Registration Detail | NEW TIN REGISTRATIONS BY ECONOMIC SECTOR — ISIC, DESCRIPTIVE LABELS (COUNT) | 22 | 11 | Adds up |
| Taxpayer Registration Detail | NEW TIN REGISTRATIONS BY ECONOMIC SECTOR — ISIC, SINGLE-LETTER CODE (COUNT) [ALTERNATE SOURCE FIELD] | 22 | 11 | Adds up |
| Taxpayer Registration Detail | NEW TIN REGISTRATIONS BY TAXPAYER CATEGORY (COUNT) | 3 | 11 | Adds up |
| Taxpayer Registration Detail | NEW TIN REGISTRATIONS BY DISTRICT (COUNT) | 124 | 11 | Adds up |
| Taxpayer Registration Detail | NEW TIN REGISTRATIONS BY COUNTRY OF ORIGIN (COUNTRY CODE, COUNT) | 164 | 11 | Adds up |
| Taxpayer Registration Detail | NEW TIN REGISTRATIONS BY FILING / OBLIGATION FLAG (COUNT) | 3 | 11 | Adds up |
| Taxpayer Registration Detail | NEW TIN REGISTRATIONS BY GENDER, AS RECORDED (COUNT) [F/M INTERPRETED AS FEMALE/MALE] | 3 | 11 | Adds up |
Method
The rules every chart follows
- 1.Key on sheet, section and indicator. Labels such as “Kenya” or “PAYE” repeat across sibling tables, so every indicator has a stable id (its sheet and source row) and is never matched on its label alone.
- 2.A blank is not a zero. Missing values stay missing: lines break at a gap, tables show “—”, and nothing is interpolated or estimated.
- 3.Fiscal years run July to June. Years are labelled like 2015/16. FY2025/26 is a partial year in the sheets that carry it. 414 indicators are documented in the dictionary.
- 4.Units come from the dictionary. Where it has none, the unit is read from the label or section text (for example “UGX Bn”); if neither states one, the unit is shown as “not stated”, and the values are never rescaled.
- 5.Series breaks are marked. Where the source changes definition (Fees & Licenses into NTR), charts draw a dashed break and say so.
- 6.Raw codes stay raw. Business-category and tax-region codes are shown as in the source, not translated speculatively.
- 7.Documentation gaps are known. Four sheets have indicators the dictionary does not describe (27 in all, mostly commodity lines in Excise_VAT_Detail); they are listed in the audit and shown here without a description.
- 8.Not externally validated. The workbook is treated as the source of truth: it has not been checked against published URA Data Books or UBOS releases beyond its own reconciliation.
Dataset v1.0 · audited 18 September 2026 · 10,271 observations across 14 sheets.