Cleaning a distribution dataset and preparing a coverage map

• Row 2: "Barsalgho" corrected to Barsalogho (spelling from the reference list, P-code correct).
• Row 3: "KAYA" in capitals with a trailing space, standardised to Kaya.
• Row 4: "Pissilla" corrected to Pissila; P-code BF1390 does not exist, corrected to BF1309 (transposed digits, name confirmed).
• Row 6: exact duplicate of row 5 (same partner, location, date and figures), deleted.
• Row 8: name (Kaya) and P-code (BF1301) inconsistent; the partner confirms Barsalogho, the name is corrected.
• Row 9: 7,200 people for 120 households, i.e. 60 people per household, an outlier; corrected to 720 after confirmation (6 people per household).
• Row 10: "Mane" corrected to Mané.
• Row 11: Djibo outside the area and without a P-code, excluded from the dataset (already reported elsewhere, avoid double counting).
• Row 13: impossible date (31 September), corrected to 30/09/2026.
• Row 16: "Ksya" corrected to Kaya (P-code BF1304 valid).

Good practice: keep the raw file, record each correction in a cleaning log and require a drop-down list of communes with P-codes in the partners' form.

Raw total of people: 21,900. Overstatement: 21,900 − 13,810 = 8,090 people, i.e. 58.6% more than the real figure. Row 9 alone accounts for 6,480 people of this difference, the duplicate for 560 and Djibo for 1,050.

Choropleth map of the communes of Sanmatenga showing the coverage rate (and not the number of people reached, which would visually favour the most populated communes). Simple, readable classes, for example: below 8%, 8 to 11%, above 11%, with a distinct colour for "no data" (Ziga, Korsimoro, etc.). Optional proportional symbols for the gap in people. Mandatory elements: title with period, legend, source ("Shelter/NFI working group, partner data September 2026"), production date, scale, north arrow, mention of administrative boundaries and disclaimer. Precautions: communes without data are not communes without needs; Dablo and Pensa stand out as the least covered; do not show the location of sites or precise distribution points.

I do not hand over the file. Named and geolocated data on displaced households are highly sensitive: sharing them with an armed party may expose people to reprisals or targeting and compromise the neutrality and acceptance of the NGO. I reply politely that I am not authorised to share these data, that the decision lies with the country management under the data protection policy, and I immediately inform my line manager. If there is a legitimate civil-military coordination need, it goes through the civil-military coordination mechanism and concerns at most aggregated, non-sensitive information (for example general areas of intervention), after a risk analysis.