This page documents how the 2025–2026 Impact & Farm Assessment dashboard is produced. It is intended to help MEAL and programme staff understand the source data, transformations, assumptions, matching rules, indicators, quality checks and interpretation of results.

1. Analytical framework

The dashboard combines the Impact Assessment, Farm Assessment and Rwanda Rural CPI time series. Survey-round results are organised chronologically and used to compare the first valid cohort round with the latest valid round. Household welfare indicators come primarily from the Impact Assessment; coffee production indicators come from the Farm Assessment.

Impact Assessment Income · household size · savings · welfare

→

Analysis Engine Clean · match · CPI adjust · PPP convert · validate

→

Dashboard Overview · progress · poverty · farm · quality

2. Data sources

Impact Assessment

Primary household-level source for Fellow identification, survey round, household income, household size, savings and other welfare measures. The analysis standardises variable names before creating the dashboard indicators.

Farm Assessment

Source for coffee-Fellow identification and farm indicators including harvest, productive trees, unproductive trees, rejuvenated trees, seedlings and district-level farm performance.

Rural CPI

Annual Rural Consumer Price Index series used to convert nominal household income into constant 2021 RWF. Future survey rounds should be accompanied by an updated CPI file containing the relevant year.

3. Population definitions

All Fellows

Distinct Fellows represented in the Impact Assessment after the dashboard's name/identifier cleaning.

Coffee Fellows

Fellows identified through the Farm Assessment and linked to the Impact Assessment. Coffee-specific analysis should use this population rather than assuming that every Fellow has coffee-farm data.

Matched Fellows

For longitudinal progress analysis, the same Fellow must be identifiable across the required survey rounds. Unmatched records remain useful for round-level summaries but should not be interpreted as individual longitudinal change.

4. Three income measures

A. Real Income (RWF)

Purpose: compare household purchasing power across survey years using 2021 as the common price base.

Real Income_t = Nominal Income_t × (CPI_2021 / CPI_t)

Interpretation: an income reported in a later year is deflated into the amount of 2021 RWF with comparable purchasing power. This is not the same as calculating year-to-year inflation growth using CPI_t / CPI_(t-1).

B. Real Income (USD)

Purpose: express constant-2021 RWF income in a common USD unit.

Real Income USD = Real Income RWF / 988.62

Interpretation: the conversion uses the fixed 2021 nominal exchange rate so changes across rounds are not driven by changing exchange rates.

C. Income / Person / Day (USD PPP)

Purpose: compare household economic resources with an international purchasing-power benchmark.

PPP/day = Real Income RWF / Household Size / 365 / 291.6

Interpretation: annual real household income is first expressed per household member per day and then converted using the fixed 2021 household-consumption PPP factor.

5. Constants and analytical parameters

Parameter	Value	Use in dashboard
Common price base	2021	All real-income comparisons
Rural CPI — 2021	140.80	Base CPI in the real-income formula
Rural CPI — 2024	209.50	Used when analysing 2024 nominal income
Rural CPI — 2025	220.30	Used when analysing 2025 nominal income
2021 nominal exchange rate	988.62 RWF/USD	Real RWF → Real USD
2021 PPP conversion factor	291.60 RWF / international $	Real RWF → USD PPP
Poverty benchmark	$3.00 / person / day	2021 PPP welfare benchmark
Days per year	365	Annual household income → daily income

For future rounds, update the Rural CPI source file. The fixed 2021 exchange-rate and PPP parameters remain part of the common-base methodology unless the methodology is formally revised.

6. Poverty & household welfare

A household is classified as below the dashboard poverty benchmark when its calculated Income / Person / Day (USD PPP) is less than $3.00.

Below benchmark = PPP/person/day < 3.00

The dashboard reports the share of analysable Fellows below this threshold. A valid household-size value is required because household resources are divided across household members.

Important: this dashboard indicator is a benchmarking measure. It should not be interpreted as a complete multidimensional poverty assessment.

7. Household size and missing PPP values

Household size enters directly into the PPP/person/day denominator. Missing, zero or otherwise unusable household-size values cannot produce a valid per-person result and should be flagged during data-quality review.

The poverty denominator should therefore be the population with valid information required for the calculation, not automatically every record in the raw survey.

8. Savings

Average monthly savings are summarised by survey round from the Impact Assessment. Savings provide an additional household-resilience indicator alongside income.

Savings should not be interpreted as equivalent to income. Changes can reflect household behaviour, seasonality, VSLA participation, shocks and the timing of the survey.

9. Coffee-Fellow identification

Coffee Fellows are identified through the Farm Assessment. This avoids classifying all programme Fellows as coffee producers when a Fellow does not have a corresponding farm record.

When comparing household outcomes specifically for Coffee Fellows, the dashboard uses the coffee-Fellow keys to filter the Impact Assessment population.

10. Farm & harvest indicators

Average harvest

Average Harvest = mean(Fellow harvest kg)

Average kilograms of coffee harvested among records with usable harvest information in the survey round.

Productive trees

Average Productive Trees = mean(productive trees)

Average reported number of productive coffee trees for the relevant Farm Assessment round.

Coffee productivity

Kg/tree = Average Harvest kg / Average Productive Trees

The dashboard summary uses the ratio of the round-level average harvest to the round-level average number of productive trees.

Headline farm KPIs use the same latest valid Farm Assessment round. Tiny accidental rounds with fewer than five harvest observations are excluded from headline comparisons so they cannot replace a valid cohort round.

11. Fellow Progress methodology

Fellow Progress is a longitudinal analysis. It attempts to compare the same Fellow across survey rounds rather than interpreting differences between two independent samples as individual progress.

Standardise Fellow identifiers/names used for matching.
Identify Fellows represented in the required survey rounds.
Reshape matched income records into a Fellow-level wide structure.
Compare nominal household income between rounds.
Compare constant-2021 real household income between rounds.
Calculate PPP/person/day where valid household-size data are available.
Apply the programme's progress thresholds to the relevant nominal-income measure.
Flag cases where nominal income improves but inflation-adjusted real income declines.

12. Nominal vs real progress

Nominal income change describes the change in the number of RWF reported by the household.

Real income change describes the change after adjusting both rounds to constant 2021 purchasing power.

A Fellow can therefore have positive nominal income growth but negative real-income growth when prices increase faster than nominal income. The dashboard surfaces this divergence because nominal growth alone can overstate improvements in purchasing power.

13. Progress targets

Programme progress thresholds are kept on the nominal-income basis used by the reference 2022–23 workbook so that programme target reporting remains comparable.

Real-income and PPP indicators are shown alongside those programme thresholds as complementary welfare measures. They should not silently replace the programme's established target definition.

14. Survey-round selection

Round-level summaries are ordered using the dashboard survey-order field. Headline comparisons use the first valid round and the latest valid round.

For Farm Assessment headline indicators, a round with fewer than five usable harvest observations is not allowed to replace a valid cohort round. This rule was introduced after detecting an isolated Farm Assessment record coded as collection period '2'.

Internal review rule: unexpected collection-period values should still be corrected in the source data where possible; the dashboard filter is a protection against misleading headline results, not a substitute for data cleaning.

15. Data quality and exclusions

Duplicates: multiple records for the same Fellow/round should be reviewed before longitudinal interpretation.
Unmatched Fellows: may contribute to round-level summaries but cannot automatically contribute to matched progress analysis.
Missing income: cannot contribute to the corresponding income summary or progress calculation.
Missing household size: prevents valid PPP/person/day and poverty-benchmark calculation.
Zero baseline income: percentage growth from zero is mathematically undefined and requires careful treatment.
Unexpected survey rounds: should be investigated and corrected rather than accepted automatically.
Farm denominators: harvest, productive-tree and productivity measures must refer to the same valid round for headline reporting.

16. Data Quality Metrics 1–7

The Internal Data Quality table uses generic column names metric_1 to metric_7 because the meaning of each metric depends on the section of the table. Use the definitions below when interpreting the Data Quality results.

A. Survey-round quality

Rows in the Survey-round quality section describe the quality of each survey round independently.

Metric	Meaning	How to interpret it
Metric 1	Unique Fellows	Number of distinct Fellows represented in that survey round after the dashboard's identifier/name cleaning. This is the main participant count for the round.
Metric 2	Total Records	Total number of survey records in the round. If Total Records is larger than Unique Fellows, there may be duplicate submissions or multiple records for some Fellows.
Metric 3	Duplicate Extra Records	Calculated as Total Records minus Unique Fellows. A value above zero signals that one or more Fellows have multiple records in the same round and those records should be reviewed.
Metric 4	Zero-Income Records	Number of records where household income equals zero. These cases require review because a true zero income is possible, but zero may also reflect missing or incorrectly entered data.
Metric 5	Missing-Income Records	Number of records with no usable household-income value. These records cannot contribute to the corresponding income averages or Fellow-level income-change calculations.
Metric 6	Missing or Invalid Household Size	Number of records where household size is missing, zero or otherwise unusable. These records cannot produce a valid Income / Person / Day (USD PPP) value or poverty-benchmark classification.
Metric 7	Missing CPI	Number of records for which the dashboard cannot identify a valid Rural CPI for the assigned survey year. Real Income (RWF) and Real Income (USD) should not be interpreted until the CPI series is updated or corrected.

B. Baseline matching quality

Rows in the Baseline matching quality section describe whether the same Fellow can be compared longitudinally between Baseline and a follow-up round.

Metric	Meaning	How to interpret it
Metric 1	Matched Fellows	Number of Fellows with usable records at both Baseline and the selected follow-up round. This is the preferred denominator for Fellow-level progress analysis.
Metric 2	Missing Baseline	Fellows found in the follow-up round but without a usable Baseline income record. They can contribute to follow-up round summaries, but not to Baseline-to-follow-up percentage change.
Metric 3	Missing Follow-up	Fellows present at Baseline but without a usable record in the selected follow-up round. They are excluded from that matched progress denominator.
Metric 4	Baseline-Zero Fellows	Matched Fellows whose Baseline household income equals zero. Percentage growth from zero is mathematically undefined, so these cases should be flagged rather than forced into an ordinary percentage-change calculation.
Metric 5	Moved from Zero to Positive Income	Fellows with zero Baseline income and positive income at follow-up. This is an important positive transition, even though a conventional percentage-growth rate cannot be calculated.
Metric 6	Not used for Baseline matching	Metric 6 is intentionally blank / not applicable in the Baseline matching section.
Metric 7	Not used for Baseline matching	Metric 7 is intentionally blank / not applicable in the Baseline matching section.

Recommended internal review: Do not interpret a change in the matched-Fellow denominator as programme performance. The denominator can change because of missing records, unmatched identifiers, duplicate submissions or unavailable follow-up data.

17. Data Refresh workflow

1. Upload Impact · Farm · CPI

→

2. Validate Structure · rounds · IDs

→

3. Recalculate Income · PPP · farm · progress

→

4. Review Data Quality first

→

5. Report Internal / public

Upload the latest Impact Assessment export.
Upload the latest Farm Assessment export.
Upload the updated Rural CPI time series when a new survey year is introduced.
Run Save files & recalculate.
Review Data Quality before interpreting headline results.
Review Fellow Progress and compare selected records against the source exports.
Only after validation, use Overview and the analytical modules for reporting.

18. Internal vs public reporting

Internal mode

May include Fellow-level progress, matching diagnostics, duplicate checks, source-file refresh controls and other quality information.

Public mode

Should contain aggregate results only. Participant-level records, data-quality details and source-file controls remain hidden.

19. Interpretation cautions

A change between survey rounds is not automatically proof that the programme caused the change.
Inflation adjustment improves comparability of purchasing power but does not control for every economic or household factor.
PPP/person/day depends strongly on reported household size and income quality.
Farm indicators may be affected by seasonality, weather, coffee cycles and timing of data collection.
Always review sample size and data quality before communicating large percentage changes.
External reporting should use validated aggregate results, not raw internal records.

20. Recommended validation checklist before reporting

□ Confirm survey-round labels and chronological order.

□ Review duplicates and unmatched Fellows.

□ Check missing income and household-size fields.

□ Verify Rural CPI contains the survey year.

□ Compare selected real-income calculations manually.

□ Confirm PPP/person/day against a manual example.

□ Check the $3/day denominator and percentage.

□ Confirm harvest, productive trees and kg/tree use the same round.

□ Review unusual outliers and very small round samples.

□ Compare headline figures with source exports before publication.

21. Rural CPI currently loaded

This table shows the CPI series currently available to the analysis engine. When future data are collected, confirm that the relevant survey year appears here before interpreting real income.

← Home

2025–2026 Cohort

Impact & Farm Assessment

Update source files

Upload replacement exports and click Recalculate. The files are stored on the server under the standard names used by the analysis engine.

Impact Assessment (.xlsx)

Browse...

Farm Assessment (.xlsx)

Browse...

Rural CPI (.xls or .xlsx)

Browse...