The Research Proposal Audit: 12 Questions to Ask Before You Send It to Your Supervisor

A topic is not a problem.

Topic:

Family planning among university students.

Research problem:

University students may have physical access to SRH services while still avoiding them because confidentiality and social visibility make those services feel unsafe.

The second statement gives the study something to investigate.

A literature review should not be:

Study A found... Study B found... Study C found...

It should establish patterns.

For example:

Across settings, confidentiality and provider judgement repeatedly influence adolescent service use, but studies differ in whether the main concern is provider disclosure or being recognised seeking care.

Now we know something.

The gap can follow.

Complete:

Existing evidence shows ______, but cannot yet tell us ______.

If your answer is:

“No study has been done at University X”

ask why University X changes what we need to know.

A gap can concern:

• evidence;
• population;
• context;
• measurement;
• methodology;
• inconsistency;
• explanation.

The gap should create the need for your question.

Gap:

We do not understand how students weigh confidentiality risks during decisions about seeking care.

Question:

What proportion of students use modern contraception?

Mismatch.

The second question may be useful.

It does not answer the gap you established.

The proposal has switched studies.

Read each objective.

Can you tell what evidence would answer it?

Weak:

To understand factors affecting family planning.

Stronger quantitative:

To estimate the proportion of students who used a modern contraceptive method during the previous six months.

Stronger qualitative:

To explore how students describe confidentiality concerns when deciding whether to seek SRH services.

Your objectives should predict the data and analysis.

Objective:

Estimate prevalence.

Design:

Purposive qualitative interviews.

Mismatch.

Objective:

Explore lived experiences.

Method:

Closed questionnaire with six predefined barriers.

Possibly a mismatch.

Objective:

Determine whether an intervention causes a reduction in depression.

Design:

One cross-sectional survey.

Major inferential problem.

Choose the design for the question.

Not because it is fashionable or familiar.

Who exactly does the conclusion apply to?

Check:

• age;
• location;
• eligibility;
• exposure/status;
• time period;
• exclusions.

“Women of reproductive age” may not be enough.

Neither is:

“students.”

A clear target population improves:

• sampling;
• measurement;
• interpretation;
• generalizability.

If you want:

district prevalence

your sample must support population inference.

If you want:

experience of women denied a requested contraceptive method

purposive sampling may be exactly right.

Check:

• sampling frame;
• selection method;
• participant groups;
• clustering;
• non-response;
• qualitative variation;
• sample-size justification.

Your framework says:

access.

What does the dataset contain?

• distance?
• cost?
• perceived access?
• waiting time?
• availability?

Your framework says:

agency.

Does the interview guide actually explore decision-making power?

Your framework says:

quality.

Which dimensions?

A concept that cannot be observed or explored is not ready for data collection.

For quantitative research:

• What primary outcome/parameter drives the sample?
• What assumptions?
• Precision or power?
• Design effect?
• non-response?
• clustering?

For qualitative research:

• How focused is the aim?
• How specific/heterogeneous is the sample?
• Which participant groups?
• What analytic depth?
• What is the initial recruitment target/range?
• How will adequacy be assessed?

And then:

Can your data-collection plan realistically produce the data the objectives require?

A perfect sample size cannot rescue the wrong instrument.

Build the table:

No blank analysis cells.

No analysis with no objective.

This table is one of the fastest ways to expose proposal drift.

Imagine:

• recruitment succeeds;
• no data are missing;
• instruments work;
• analysis runs exactly as planned.

What will you know?

Now return to page 1.

Was that the problem?

If not, the proposal is internally coherent around the wrong question.

That is still a problem.

After the 12 questions, ask:

What is the strongest criticism a reviewer could make of this design today?

Do not answer:

“Time constraints.”

Try harder.

Possible answers:

• temporal ordering cannot be established;
• outcome measurement is weak;
• comparison group is not credible;
• sample cannot support subgroup analysis;
• selection could bias prevalence;
• confounder adjustment is unjustified;
• qualitative sample is too heterogeneous for the planned depth;
• primary objective does not drive sample size.

Now ask:

Can I redesign this before data collection?

This is the cheapest moment to find the flaw.

Before submission, check:

• risk;
• consent/assent;
• privacy;
• confidentiality;
• vulnerable populations;
• sensitive topics;
• referral procedures;
• data security;
• community permissions;
• compensation;
• conflicts;
• dissemination.

Ethical feasibility can change the methodology.

For example:

A plan to interview adolescents about sexual behaviour in a school office with teachers walking past is not merely an ethics issue.

It is also a data-quality problem.

Participants may not answer honestly.

A theoretically perfect study can still fail if:

• the sampling frame does not exist;
• the assay is unavailable;
• follow-up is impossible;
• the target population is too rare;
• the questionnaire is four hours long;
• translation has not been budgeted;
• cluster numbers are insufficient;
• the ethics timeline exceeds the degree deadline.

Methodological quality includes feasibility.

Do not solve feasibility by quietly changing the design after approval.

Plan it.

Before sending, attach or internally complete a one-page audit: