NHS Information Analyst Interview Questions and Illustrative Answers
Practise analyst interview reasoning with original fictional examples covering definitions, joins, quality checks and clear communication.
Applies to: UK-wide
Prepare to explain the analytical decisions
An NHS information analyst interview can explore how you clarify a request, extract or transform data, validate a result and communicate its meaning. Use the actual person specification to decide which tools and level of responsibility are relevant. A junior vacancy and an experienced reporting post can require different depth, even when both include an analytical exercise.
These are original practice questions with fictional worked answers. All datasets and projects below are invented and non-patient. They are not employer exercises, national scoring criteria or claims about real NHS performance. Substitute your own honest experience and explain any training or review involved. Do not bring confidential data or records into a personal portfolio or interview example.
1. Tell us about an analysis you completed
Fictional illustrative answer: I built a synthetic office-supply delivery project to understand why the number of late deliveries had increased. I defined an order and a delivery separately, checked missing promised dates and documented how split deliveries would be treated. I compared both counts and proportions, then produced a short report with the definition beside the chart. Review showed that my first join repeated some order records, so I corrected the relationship and reconciled totals. The useful outcome was an explainable method and a cautious conclusion, rather than an impressive dashboard screenshot.
- Original fictional illustration, not an employer answer or successful candidate testimony.
- Replace the biography and project with your own truthful evidence and actual responsibility.
Why it works: It names a question, checks, a mistake corrected and a limitation. Describe your own project and tools instead of adopting this fictional biography.
2. How would you clarify a request for a dashboard?
Fictional illustrative answer: I would ask which decision it should support, who will use it and how often it needs updating. I would agree the measures, period, population and definitions before designing the display. For a fictional equipment-order dashboard, I would clarify whether an outstanding order means any undelivered line or an order with no delivery at all. I would check source availability and propose a small example for review. If the deadline or data could not support the requested detail, I would make that explicit and agree the next useful output.
3. Two reports disagree. How do you investigate?
Fictional illustrative answer: I would compare their purpose, refresh time, reporting period, filters and definition of one record. I would reconcile at a suitable level and trace the difference to source and transformation steps. In my synthetic delivery exercise, one report counted orders and another counted deliveries, so a split order appeared several times. I would document that finding and obtain agreement on the intended measure rather than force totals to match. If there were a remaining unexplained difference, I would flag it before a stakeholder relied on the comparison.
4. What can go wrong when joining tables?
Fictional illustrative answer: I would establish what one row represents and whether the join keys are unique on either side. A fictional order can have several deliveries, so joining an order value to each delivery can repeat that value and inflate a total. I would check row counts, unmatched keys and aggregate totals before and after the join, and test a small understood example. The right relationship depends on the question. I would not use a blanket duplicate removal to hide a many-to-many problem without understanding which records should be retained.
5. How would you handle missing values?
Fictional illustrative answer: I would establish what missing means in that field and whether it differs from zero, not applicable or an unknown status. I would quantify the gap, check its pattern and discuss it with the source owner where appropriate. In a fictional supplies project, a missing promised date would prevent a reliable late-delivery classification; it should not silently become on time. I would explain how affected records were handled and how that limits the result. Any imputation or exclusion would need a justified, documented approach rather than a convenience choice.
NHS England's data quality resources describe assurance across the use of information. The government's quality framework distinguishes completeness from accuracy: a filled field can still be wrong.
6. How do you know your report is correct?
Fictional illustrative answer: I would check the agreed definitions, source coverage and transformation logic, then reconcile totals and examine exceptions. I would test known synthetic cases, including missing values and split deliveries, and invite appropriate review. I would record the source version, refresh position and checks so another person could understand the result. A chart looking plausible would not be sufficient. After release I would keep a route for questions and corrections, because a dependable output needs maintenance and clear ownership as well as an initial check.
7. Late deliveries rose from 18 to 24. Has performance worsened?
Fictional illustrative answer: I would first ask how many eligible deliveries there were and whether definitions and coverage stayed consistent. If the fictional totals were 60 and 120, the late proportion would fall from 30% to 20%, while the number late increased. Those describe different aspects of performance. I would check the period, missing dates and any change in the mix before offering an interpretation. Two points alone would not establish a sustained trend or explain the cause. I would present the count and denominator together so the audience could assess the comparison.
8. A manager wants the chart to show improvement. What do you do?
Fictional illustrative answer: I would clarify the question and explain what the data supports. I could improve clarity, provide context or show a relevant alternative measure, but would not change definitions, omit inconvenient periods or adjust values to create a preferred impression. If a proposed comparison was misleading, I would explain why and seek an agreed accurate presentation. I would document substantive definition changes and use the appropriate review route if disagreement remained. The aim is information people can rely on, including when the result is unwelcome.
9. How would you explain a technical result to a non-technical colleague?
Fictional illustrative answer: I would start with the question, the main finding and its practical limit. In my synthetic project I explained that more late deliveries did not automatically mean a higher late rate, because total deliveries had doubled. I showed the two denominators beside a simple chart and asked whether that answered the intended question. I kept the join detail available for follow-up rather than begin with it. I would use plain language, accessible labels and the audience's context, while retaining enough method for the conclusion to be checked.
10. How do you prioritise reporting and urgent requests?
Fictional illustrative answer: I would clarify the deadline, decision and consequences, then agree priorities with the responsible manager. I would make existing commitments and required checks visible, suggest a proportionate first output and communicate any impact on other work. I would not promise a fully validated analysis before understanding the sources. If a request changed, I would confirm the revised scope and record what was deferred. In a small project, an explicit task and review log helped colleagues distinguish a draft from a checked output.
11. What if you have not used a named tool?
Fictional illustrative answer: I would state that accurately and explain relevant transferable experience without calling it proficiency in the requested tool. If the specification allowed a similar platform, I would demonstrate the concepts and tasks I had completed there. If prior experience in the exact tool was essential, I would recognise the gap rather than rely on enthusiasm alone. I would describe a focused learning plan and distinguish a guided tutorial from work I had designed, tested and explained independently.
A closed Newcastle Hospitals specification required SQL and Power BI or similar experience while listing R scripting as desirable. That was one employer's requirement, not a universal NHS tool list.
12. What should you ask the panel?
Ask how analytical work is reviewed, which definitions and sources the team owns and how new analysts learn the service context. Clarify the actual tool environment, balance of routine and new work, and any assessment instructions. If the post mentions variation analysis, ask about relevant development and support rather than claim a chart template demonstrates expertise.
NHS England's Making Data Count programme offers resources on understanding variation. Use appropriate learning if that is part of the job, and avoid declaring a trend from a single comparison.
Prepare one genuine example for requirements, quality, communication, organisation and feedback. Practise a short explanation and a deeper follow-up about your checks. The supporting statement guide helps connect this evidence to criteria, while the career guide helps determine which vacancy level and technical requirements match your present experience.
Frequently asked questions
Check the invitation and permitted preparation. Employers choose their own assessment. Ask about required software and adjustments through the stated process; there is no universal format established here.
No. They illustrate reasoning. Use a genuine project you can explain, or develop your own clearly labelled synthetic practice work. Do not claim another example as your employment experience.
Check definitions, period, refresh, record level and transformations. Reconcile against understood cases and flag unresolved differences. Plausibility alone does not establish that either output answers the intended question.