How to write a logframe with an example and donor rules

A logframe, short for logical framework, is a table setting out what a project intends to change, how anyone will know whether it did, and what must be true outside the project for the logic to hold. To create a logframe, establish the causal logic before filling in the cells.

Writing a logframe takes five steps, in this order:

1. Separate causes from symptoms
2. Write the result statements
3. Add indicators
4. Set baselines, targets and verification
5. Write the assumptions

Causal logic first, statements second, measurement last. Each step names the donor rules that change it, because there is no single correct layout. The example running through the guide is invented: the district, figures, facilities and project are made up for this piece, and no real programme or grant is described.

Before you start: get the template and its rules

The template decides how many of each level you may have, which columns exist, and what goes inside a cell rather than beside it. Getting it first saves rework discovered at submission.

Three examples of how far these differ. The Asian Development Bank's 2024 guidelines allow one outcome and up to three impact statements, with no indicators or targets for impact. The Green Climate Fund's simplified approval template places the logical framework in Annex 2a and requires the project description and budget to match it. DG ECHO's Single Form results guidance limits indicators to ten per result.

Check the permitted numbers of results and indicators, and whether impact must be measured. Read the guidance for your funding round: the dated documents below illustrate differences, and the current application pack governs your submission. If no funder has given you a template, The National Lottery Community Fund says outright that it does not mind how applicants share this information, and suggests four columns: activities, outputs, outcomes, evidence.

Step one: separate causes from symptoms

Write down the problem, ask what causes it, then ask what causes that. Keep going until you reach something the project could act on. This is the step people skip, and skipping it is how a project ends up addressing a symptom with a coherent-looking framework. An internal problem-tree exercise can expose those gaps; check its causal claims with the people affected before committing to the design.

For the invented example, the observed problem is low use of skilled care during childbirth. Possible causes include unstaffed facilities at night, gaps in pregnancy referrals, unavailable transport and distrust of the care offered. Each suggests different work.

For this example, assume the analysis identifies staffing, referrals and transport as the barriers the project will address. Turn them into proposed changes, then distinguish what the project can deliver from what requires other people to act. Reversing a negative sentence does not establish that causal link. This is the reason to build the theory of change before the logframe.

Step two: write the result statements

A result statement describes a state of affairs after the change, not an activity. "Train midwives" is an activity; "facilities can manage deliveries at night" is a result.

The same result can sit at different levels in different frameworks. Rosenberg and Posner explain this in their 1979 guide to the logical framework: a project needed to achieve a programme's purpose appears as an output on the programme's framework and as the purpose on the project manager's own, and Des Gasper's 2000 review confirms that levels are contextual rather than inherent. Use the funder's definitions, then test the causal relationship between adjacent levels. The distinction between an output and an outcome depends on the scope of the framework; moving a sentence between rows does not make the logic sound.

For the example, the outcome is that more women give birth in a facility attended by a skilled birth attendant. Output 1 is that facilities can manage deliveries around the clock, output 2 that pregnant women receive documented referrals from community health workers, output 3 that night transport exists. Impact sits above: maternal deaths reduced.

Activities are the work needed to deliver each output. Follow the template on where to record them; they may sit in the matrix or in a separate workplan. Give each activity resources, a cost and a timeframe. Group them under the output each serves and number them so activity 1.2 is the second activity under output 1. The worked example keeps its activity plan separate to make both tables readable.

Step three: add indicators

An indicator specifies what will be measured. In the example below, the target sits in a separate column, so the indicator names the measure without a promised value or direction of change.

Two donors disagree openly here. DFID's 2011 guidance recommends at most three indicators per output, with more requiring justification, and separates the measure from its target. It rejects the term SMART indicator on that basis. Defra's Biodiversity Challenge Funds head their column "SMART Indicators (including disaggregated targets)", ask applicants to aim for at least two indicators at outcome level and per output, and require the template to be completed without empty boxes. Under the first the target has its own column; under the second it goes inside the indicator sentence.

ECHO's archived 2014 to 2020 specific-objective guidance distinguishes delivered products and services from their effects on the people using them. In that framework, a delivery count cannot evidence the specific objective. Otherwise prefer something the project can move and somebody can count. Where you count a stand-in instead, say what you are assuming, because a proxy indicator depends on its relationship to the intended result.

Step four: set baselines, targets and verification

A baseline is the value before the project; a target the value expected by a date. Both belong to the indicator, not the statement.

If you do not have a baseline, say when you will have one rather than leaving the cell blank or inventing a figure. UK Aid Match requires baseline data on impact, outcome and output indicators within three months of the project start, and asks for a collection plan at start-up where new data are needed. That provision is specific to UK Aid Match. The same guidance rules out logframe changes within three months of an annual review assessment, so a target you expect to renegotiate has a narrower window than the project calendar suggests.

The verification column names where each number comes from, and it is the cheapest check on whether an indicator is real. The test is whether you can say who holds the record, how often it is produced, and whether it exists today. Norad's 2008 guidance illustrates four fields: means of verification, collection method, frequency, responsibility. Even where your template has one cell, answering all four is what makes the entry worth having.

One caution on the fourth. WWF's 2012 conservation standards name an individual responsible in the workplan. DFID's 2011 guidance says individuals' names should not appear in its published logframes. Distinguish the internal assignment of work from the version you submit or publish.

Step five: write the assumptions

An assumption is a condition outside the project's control that must hold for one level to produce the next.

The worked example uses assumptions about the link to the next level. The assumption on an output row is what must hold for the outputs to deliver the outcome. On the outcome row it is what must hold for the outcome to contribute to the impact. Its impact row has no assumption, because this example models no further level above it. A condition needed for an output to be delivered at all belongs a level lower, beside the activities.

Check the form before copying that placement. DFID's 2011 frame has assumption fields at outcome and output level, while the CDC's 2025 evaluation framework handles the same content as contextual factors sitting underneath the logic model as foundational to the whole thing, rather than as a per-row cell.

The test: if this condition failed, would the level above still be reached? If yes, it is background, not an assumption. If no, and the project cannot act on it, that is what belongs in the cell.

Worked logframe example with a completed matrix

All figures, costs and dates below are invented. The district has eleven facilities and a project running from January 2027 to December 2029. This is an illustrative format, not a completed submission to any of the donors compared above. It keeps indicators and targets separate and records impact as context. Adapt it to the actual application pack, including any required disaggregation and annual milestones.

The activity plan supplies the work underneath all three outputs. These illustrative budgets cover the activities listed, not the project's full staffing, monitoring or overhead costs.

The impact row needs care. Recording a district mortality rate does not establish that the project caused a change in it. This example monitors the rate without a project target. ADB goes further and omits both impact indicators and targets; another funder may require both. Copy the reasoning, then follow the required format.

Define the denominator before collecting the outcome indicator. Here it is all live births to district residents in the reporting period, including home births. Using facility registers for both numerator and denominator would answer a different question. The written measurement instructions also need to specify how district birth estimates are obtained and updated.

Output 2 counts referrals delivered, not attendance or subsequent births. Count each woman once over the project, rather than adding quarterly totals that include repeat referrals. Its zero baseline belongs to the new project scheme; it does not claim that no referrals existed in the district before the project.

Output 3 uses a percentage so the denominator remains valid in months of different lengths. The invented 95% target demonstrates the format, not a recommended service standard. A real target needs evidence about demand, resources and acceptable interruptions. Availability also says nothing by itself about response time or the quality of a journey.

Midwife postings sit beside the activities because they affect whether output 1 can be delivered. Whether women choose to use a staffed facility sits on output 1 because it affects the link to the outcome. If the project can deliver a condition itself, budget it as work rather than leaving it as an assumption.

With a draft in front of you, read it upwards. If every output is delivered, is the outcome plausible? If not, either an output is missing or an assumption is carrying more weight than it should. Then read down and check that no indicator at one level also evidences the level above, which Rosenberg and Posner named as one of the most common weaknesses in logframe designs. Fix the numbering before anybody reports against it, because logframe numbering becomes a reporting reference.

---

Sources

Donor templates and their cardinality rules

• Guidelines for Preparing and Using a Design and Monitoring Framework: Sovereign Operations and Technical Assistance, Asian Development Bank (December 2024). Source for one outcome statement, up to three impact statements, and no indicators or targets at impact level
• Simplified Approval Process Funding Proposal Template, Green Climate Fund (linked template accessed 18 September 2026). Source for the logical framework in Annex 2a and consistency with the project description and component budget
• 7.3 Results, DG ECHO Partners' Helpdesk, international organisations guidance for 2021 to 2027 (accessed 18 September 2026). Source for the limit of ten indicators per result
• 4.2 Specific objective, DG ECHO Partners' Helpdesk, archived 2014 to 2020 NGO guidance. Source for distinguishing delivery of products or services from effects measured at specific-objective level
• More information about your Reaching Communities application, The National Lottery Community Fund (accessed 18 September 2026). Source for the refusal to mandate a format and the four suggested columns

Indicators, baselines and targets

• Guidance on using the revised Logical Framework, DFID (January 2011). Source for the recommendation of at most three indicators per output, justification for more, separation of indicators and targets, assumption fields, and exclusion of personal names from published logframes
• Supplementary Logframe Guidance, Biodiversity Challenge Funds, Defra approved (accessed 18 September 2026). Source for the target-bearing indicator column, the advice to aim for at least two indicators at outcome level and per output, and the no-empty-box rule
• Developing a logframe, UK Aid Match (linked guidance accessed 18 September 2026). Source for baselines and collection plans within the stated start-up process, and no logframe changes within three months of an annual review

Verification and assumptions

• Results Management in Norwegian Development Cooperation, Norad (2008). Source for the example monitoring framework with separate fields for means of verification, collection method, frequency and responsibility
• WWF Standards of Conservation Project and Programme Management, WWF (2012), Conservation Measures Partnership mirror. Source for assigning an individual responsibility in an internal workplan, distinct from DFID's published logframe
• CDC Program Evaluation Framework Action Guide, CDC (2025). Source for contextual factors positioned underneath the logic model as foundational rather than as a per-row cell

The method itself

• The Logical Framework: A Manager's Guide, Rosenberg and Posner, Practical Concepts Incorporated (1979). Source for an output at one level becoming the purpose at the next, and for indicators at one level being unusable as evidence for the level above. Third-party mirror; PCI no longer exists
• "Logical Frameworks": Problems and Potentials, Des Gasper, Institute of Social Studies, The Hague (2000). Source for levels being contextual rather than inherent