Run a needs assessment

What You Will End Up With

You will have a documented picture of where resources and services fall short, and of what the community already has, so that assistance can be prioritized and your response strategy rests on evidence. That is the main goal of a needs assessment.

The work also gives you baseline data: information from the period before a program starts that shows the size of the problem you set out to address. To make your findings mean something, compare your community's figures with other communities, the national picture or the community's own past. A number on its own tells a reader very little.

Before You Start

Settle a few things before you collect anything. It saves you from gathering data nobody can use.

• The purpose and audience. Agree on why your group needs this information and how it will be used. Decide whether results will be shared publicly or kept inside your organization.
• The exact questions. Decide what you want to know, and whether you need statistics only or also qualitative material such as life stories. Precise questions make the search far easier.
• What already exists. Take stock of existing information first, so you do not repeat assessments or ask the same people the same things again and again. In the early stages of an emergency, when time, staff, resources and access are limited, it may be impossible to collect primary data, so existing information matters most.
• What the community already has. Asset mapping is a systematic way of cataloging services, benefits and resources: people's skills, organizations, physical space, institutions, associations and the local economy.

Collect your own data only when what you need does not already exist, and set limits on how much, because too much is as much of a problem as too little.

Steps

1. Review the context. Define the scope and scale of the problem, the affected population and the locations. Check that your planned data collection meets ethical standards.

2. Review existing data and map assets. Look at secondary sources such as vital statistics, records, mortality reports and literature reviews before going anywhere near the community. Then map assets by cataloging what the community already has: people's skills, organizations, physical space, institutions, associations and the local economy.

3. Match the method to the question. Numbers cannot express motives, opinions, feelings or relationships, so you will usually want both kinds of method. Use this as a guide.

   • How many, or how often? Use a survey or written questionnaire. Decide the purpose first, because it shapes the questions, the population and how you deliver it. If your population has limited literacy, do not use a written survey (see step 4).
   • Why, or how? Use interviews. They suit perceptions and in-depth information, and their style can be highly structured, semi-structured or unstructured. For a valid structured result, ask everyone identical questions.
   • What does a group think, and what should we do? Use a focus group, a small discussion guided by a trained leader.
   • What are the broad barriers and possible solutions? Use a public forum, which can cover issues and concerns, barriers and resistance, community resources for change, and recommended alternatives.
   • What is actually happening? Use observation: go to a place or event, or take part in community life while recording what you see. In a rapid assessment, primary data usually comes from observation, key informant interviews and community group discussions.

4. Collect the data, carefully. Train your collectors first. Who is asked, how and when all affect the answers. For a representative sample, every member of the target group needs an equal chance of being chosen, or the sample should be fairly large. A typical focus group has about 6 to 10 people, sometimes chosen so members do not know one another and social pressure is lower. Before one, recheck your goals, pick an experienced facilitator, arrange a recorder and write your topics and questions in advance.

   Do not give a written survey to a population with limited literacy. Surveys delivered cold bring back only two to three percent unless the topic matters a great deal to the community, and the more personal the delivery, the higher the return.

   Handle consent and safety as part of the work, not as an afterthought:

   • Explain the purpose, risks and benefits, and tell people they can withdraw at any time.
   • Do no harm. Work out what unintended negative effects could arise, such as tension, worse inequalities or added risk, and plan to reduce them.
   • Keep participants anonymous, store electronic data password-protected and encrypted, and keep paper copies in locked locations.
   • Explain the purpose of the collection and the response capacity you actually have, and make sure response and referral systems exist. Otherwise you create unrealistic expectations and assessment fatigue.

5. Clean, verify and analyze. Check for inconsistencies, errors and gaps, and cross-reference with other reliable sources. Then look for key trends, needs, challenges and vulnerabilities, and flag anything that needs immediate attention. As a rule of thumb, do not calculate a rate from fewer than 30 people, and even 30 is probably far too few for an accurate picture. After this first round, look at what is missing. Either search again for sources you overlooked or collect the missing data yourself.

6. Validate, share and use the findings. Talk with community members to check what you found. A focus group, for example, can explain why outcomes changed over several years. Share the findings and their implications openly, including with the affected communities. Then use them: ask staff how the new information changes their projects, or revise the plan. Plan to keep collecting, because needs data goes stale and one source says data more than a year old should be treated as unreliable.

Using AI Safely

Some teams use AI or machine-learning tools to transcribe interviews or analyze text. If you do, the same duties to participants apply, and they need to be spelled out.

• Say so in the consent form. The form must explain how participants' data will be used, analyzed and protected when such tools are involved.
• Keep the safeguards. Anonymity, encryption and do no harm apply to everything a tool touches or produces.
• Verify what comes out. The cleaning and verification in step 5 apply to anything a tool produces.

Common Mistakes

• Collecting more than you can use. Set limits before you start.
• Sending out untrained collectors. Train people first.
• Relying on old data. Treat data more than a year old with caution, since needs data goes stale.
• Computing rates from tiny samples. As a rule of thumb, do not calculate a rate from fewer than 30 people.
• Using unrepresentative samples. Do not let the few who answer a cold survey stand for everyone.