This guide works the other way. It starts with the decision what do programme managers need to know to manage effectively? and builds the indicator backwards from that question. The result is a set of indicators that are genuinely used, not just dutifully reported.
Step 1: Start With the Decision, Not the Data
Before writing a single indicator, answer this question for each level of your results framework: What specific decision does a manager at this level need this indicator to inform?
For an outcome indicator in a disease surveillance programme, the decision might be: "Is the surveillance system detecting cases at sufficient sensitivity to trigger a timely outbreak response?" That decision requires an indicator that measures detection sensitivity not case counts, not training completion rates, not the number of reports submitted.
If you cannot name the decision that an indicator informs, remove it from the results framework. An indicator that does not drive a decision is overhead data collection cost with no management return.
Step 2: Define the Result Before the Measure
Indicators measure results. So the result must be defined clearly before the indicator can be designed well. This is a prerequisite that is frequently skipped, producing indicators that measure proxy phenomena things correlated with the intended result rather than the result itself.
The result definition should specify:
- What changes: What is different at the end of the programme period compared to the beginning?
- Who changes: In which population, system, or organisation does the change occur?
- By how much: What is the magnitude of change expected, given the programme's resources and theory of change?
- By when: At what point in the programme cycle is this change expected to be observable?
Example: "By the end of Year 2, the percentage of suspected epidemic-prone disease cases at facility level that are investigated within 48 hours of notification will increase from 43% (baseline) to 75%." This is a result definition specific, attributed, measurable, time-bound. The indicator follows naturally: percentage of notified suspected epidemic-prone cases investigated within 48 hours.
Step 3: Apply the SMART Test
SMART Specific, Measurable, Achievable, Relevant, and Time-bound is the standard quality framework for indicator design. Applied rigorously, it eliminates most weak indicators before they enter the results framework.
Specific
The indicator must be defined precisely enough that two different M&E officers, using the same data source, will calculate the same value. Vague indicators "improved surveillance quality" or "enhanced programme performance" fail the specificity test because they cannot be measured consistently.
Specificity requires defining: the numerator (what is counted), the denominator (the total against which it is expressed), the unit of measure (percentage, number, rate per 100,000), and any relevant disaggregation (by LGA, by disease, by sex or age group).
Measurable
Can the indicator be measured using a feasible, affordable data source? An indicator that requires a nationally representative household survey to measure is not feasible for a district-level programme with a quarterly monitoring cycle. An indicator that can be calculated from existing DHIS2 data with no additional data collection is highly feasible.
At the design stage, identify the specific data source for each indicator before finalising the results framework. If no feasible data source exists, the indicator must be revised.
Achievable
Is the target associated with the indicator achievable given the programme's resources, timeframe, and context? This is where target-setting discipline matters. See Results-Based Management for evidence-based target-setting guidance.
Relevant
Does the indicator actually measure the result it is assigned to? A training attendance rate assigned as an outcome indicator is not relevant it measures an output (training delivered), not an outcome (changed practice). Relevant indicators measure what they are assigned to, not a proxy that is easier to count.
Time-bound
The indicator must be measured at defined points in time with a specified baseline measurement date and target achievement date. An indicator with no time reference cannot demonstrate change over a programme cycle.
Step 4: Write the Indicator Reference Sheet
The single most important quality control in indicator design is the indicator reference sheet a document that defines every indicator precisely enough to enable consistent measurement across implementers, partners, and reporting periods. Without it, the same indicator will be measured differently by different people, and the resulting data will be incomparable.
Every indicator in the results framework needs a reference sheet containing:
- Indicator name
- Short, specific, plain-language name that describes what is being measured.
- Definition
- A full written definition of exactly what the indicator measures specific enough that a trained M&E officer can measure it correctly without additional guidance.
- Numerator
- The count in the top of the fraction what is being measured. Define precisely, including any conditions that must be met for a case to be counted.
- Denominator
- The total count in the bottom of the fraction the population or total against which the numerator is expressed. Specify the denominator source and the population year if a census figure is used.
- Unit of measure
- Percentage, number, rate per 100,000, days, or other unit. This should match the unit used to set the baseline and target.
- Data source
- The specific data source from which numerator and denominator values will be collected DHIS2 data set name and data elements, facility register, laboratory log book, or household survey instrument and question reference.
- Collection method
- How data is collected and aggregated routine facility reporting, dedicated data collection exercise, administrative records review.
- Collection frequency
- How often the indicator is measured weekly, monthly, quarterly, annually. Should match the decision frequency for which it is designed.
- Responsible party
- Who collects, calculates, and reports this indicator. Without named responsibility, indicators routinely go unmeasured.
- Disaggregation
- What disaggregation is required by LGA, by disease, by sex, by age group. Disaggregation that is not specified at design stage is rarely available at reporting stage.
- Baseline value and date
- The measured value of the indicator before programme activities began, and the date of measurement.
- Target value and date
- The intended value at the end of the programme period, and the date by which it should be achieved.
- Calculation notes
- Any specific calculation rules, rounding conventions, or handling of missing data that affect the indicator value.
Step 5: Balance the Indicator Set
A well-designed results framework has a balanced indicator set one that covers all levels of the results chain without creating data collection burden or leaving critical results unmeasured.
The test for balance is simple: if you removed all output indicators, would you still know whether the programme is achieving its outcomes? If you removed all outcome indicators, would you know whether the programme is on track to achieve impact? If either answer is no, the framework is unbalanced.
For most global health programmes, a balanced indicator set has:
- Two to four impact indicators directional, long-term, population-level
- Four to six outcome indicators programme-sensitive, medium-term, linked to decisions
- Four to eight output indicators countable, short-term, under programme control
- Two to four process indicators timeliness, quality, and equity of implementation
Total: ten to twenty-two indicators for a results framework. Anything beyond twenty-five should trigger a rigorous review of which indicators are genuinely essential. Every indicator added is a data collection obligation added. Discipline the set to what will actually be used.
The Equity Lens: Disaggregation by Design
An indicator that reports only aggregate performance conceals inequity. A national immunisation coverage of 75% may reflect 95% coverage in urban LGAs and 55% coverage in rural ones. The aggregate figure is technically accurate and practically misleading.
Equity-focused M&E requires building disaggregation into indicator design not as an afterthought, but as a design requirement. Every outcome indicator should specify the minimum disaggregation required for equity monitoring: at minimum, geographic disaggregation to the LGA level and sex disaggregation where relevant.
The principle is that impact is measured by who is reached last, not by average outcomes. The most vulnerable population geographically remote, economically marginalised, underserved by the health system should be explicitly tracked in the indicator set. If it is not tracked, it is not being measured, and what is not measured is rarely improved.
Common Indicator Design Mistakes
Using percentage change rather than percentage point change
"Coverage increased by 20%" is ambiguous. Does that mean from 50% to 70% (20 percentage points), or from 50% to 60% (a 20% relative increase)? Always specify whether a target is expressed in percentage points or as a relative percentage change, and use the same convention consistently across all indicators.
Indicator names that describe activities, not results
"Number of health workers trained in IDSR" is an output indicator. "Percentage of health workers who correctly apply IDSR case definitions at the point of care" is an outcome indicator. The difference in the name signals the difference in what is being measured and the difference in what the programme is accountable for.
No indicator for data quality itself
The M&E system should include an indicator for its own data quality reporting completeness, data accuracy rate from DQA, or timeliness of data submission. A results framework that monitors programme results but not the quality of the data used to measure them is measuring with an uncalibrated instrument. For more on this, see DHIS2 Data Quality.
Setting targets without baseline data
A target is a statement about change the difference between where the programme starts and where it aims to be. Without a baseline, a target is a statement about an absolute level, not a change. A target of "75% of facilities submitting weekly reports" means very different things if the baseline is 40% versus 72%.