The last link in the chain

• Reporting is the sixth stage of the research process we mapped in session 1 — and the stage where studies most often fail their users
• Research that is never communicated, or communicated badly, might as well not have been done
• The Meridian team now faces this stage: months of evidence must become something the board can act on
• One study usually produces several documents: a full report for the record, a presentation for the decision, a summary for readers who will only ever read one page
• Good reporting is a design problem: start from the audience, not from the study

The research report: structure

• The report is the study’s complete, permanent record. A typical organisational research report runs:
  1. Title page and executive summary
  2. Background and aims — the problem and the research question
  3. Methods — design, sampling, and data collection, in enough detail to be scrutinised
  4. Findings — organised by question or theme, not by the order things happened
  5. Discussion — interpretation, and the study’s limitations
  6. Conclusions and recommendations — clearly separated and labelled
  7. Appendices — instruments, full tables, technical detail
• The academic version of the same logic is IMRaD: Introduction, Methods, Results, and Discussion

The managerial presentation: a different structure

• A presentation is not the report projected onto slides — it has its own architecture
• Board attention is scarce, so the order inverts: answer first, evidence after (the “pyramid principle”)
  1. The question the board asked
  2. The answer — the headline findings
  3. The key evidence behind each finding
  4. The recommendations, with options and costs
  5. Methods in reserve — backup slides ready for questions
• Methodology gets one slide at most; the detail exists, but it waits to be asked for

Report vs presentation: a comparison

The executive summary

• A craft of its own — the hardest writing per word in the whole report
• Findings first: it opens with what was found, not with the study’s history or a methodology lecture
• One page at most, and self-contained — many readers will read nothing else
• Plain language, the few numbers that matter, and recommendations clearly labelled as recommendations
• Written last, even though it appears first — you cannot summarise what is not yet finished

Tailoring to the audience

• The same findings reach at least three audiences, and each needs a different document:
  • The board — decision-focused: the answer, the implications, the costs and risks; minimal method
  • The academic reader — method-focused: needs design, measures, and analysis to judge whether the claims are warranted
  • All-staff communication — trust-focused: what we found, what will change, and what happened to your data
• What varies is order, detail, and vocabulary; what must never vary is the findings themselves
• Reporting back to respondents closes the loop — and protects the response rate of every future survey

Report uncertainty honestly

• Every estimate from a sample carries uncertainty; accurate reporting shows it rather than hiding it
• Report the margin of error or confidence interval around survey estimates — “64%, plus or minus 5 points” — so small differences are not mistaken for real ones
• Report the response rate and who is likely missing: non-response bias matters, and leavers are the hardest people to reach
• In qualitative work, report how many people were studied, how they were selected, and what settings the findings can speak to
• Avoid false precision: “about two-thirds” is more honest than “66.42%” from a sample of 300

Limitations are part of the findings

• Session 1’s closing lesson, now applied to the page: acknowledging limitations is a mark of good research, not a confession of failure
• Limitations belong in the discussion section — visible, not buried in an appendix — and echoed wherever a claim depends on them
• The usual suspects: coverage (who was missed), measurement (self-report and social desirability), timing (one point in time), and generalisability (which settings the results travel to)
• A report that claims no limitations is not a stronger report — it is a less honest one
• Stated limitations do productive work: they define what the next study should examine

Findings are not recommendations

• A finding is what the evidence showed; a recommendation is what the organisation should do about it
• The recommendation always adds something the data cannot supply: judgements about values, costs, priorities, and risk
• The classic failure is to slide from one to the other without noticing — accurate reports label and separate them
• Evidence narrows the options; it rarely dictates the choice

Do not overclaim causality

• A cross-sectional survey establishes association, not causation — correlation is not causation, however managerially convenient it would be
• The damage is usually done by language: “drives”, “leads to”, “because of”, used for merely correlational results
• Honest wording: “is associated with”, “predicts”, “is consistent with”
• Always ask what else could produce the pattern: reverse causation and confounding variables
• A defensible causal claim wants an experimental or at least a longitudinal design

Report what did not support expectations

• Reporting only the results you hoped for — cherry-picking — distorts the picture as surely as inventing data would
• Null and negative results are findings: “pay was not the main driver” redirects real money away from a raise that would not have worked
• Report every analysis you planned, and label exploratory, after-the-fact analyses as exploratory
• The pressure to tell the sponsor what it wants to hear is real; resisting it is what makes the research worth paying for

One message per chart

• A chart is an argument, not decoration — know the message before you design the figure
• Put the message in the title: “Turnover is concentrated in first-year store staff”, not “Figure 3: Turnover data”
• One chart, one comparison: if a figure needs a paragraph of explanation, split it or simplify it
• For every chart, ask: what comparison is the reader supposed to make — and does the design make that comparison easy?

Choose the chart for the comparison

• The comparison you want the reader to make dictates the form:
  • Comparing categories (turnover by division) → bar chart
  • Trend over time (satisfaction by month) → line chart
  • Parts of a whole → a pie only for two or three shares; beyond that, humans compare bar lengths far better than angles — use bars
  • Relationship between two variables (engagement vs turnover across stores) → scatterplot
• When exact values matter more than the pattern, a small, well-formatted table beats a chart

Honest axes

• Bar charts must start at zero: a bar encodes its value by length, so a truncated axis makes the bar lie
• Example: engagement scores of 3.9 vs 3.6 plotted on an axis starting at 3.0 turn an 8% difference into bars that differ by 50%
• Line charts may zoom in on a range to show change — but the axis must be clearly labelled, and the zoom acknowledged
• Never extrapolate a line beyond the data as if the projection were observed
• Keep scales consistent across charts the reader is meant to compare

Declutter and label directly

• Every element must earn its place: drop heavy gridlines, borders, backgrounds, and decorative colour
• Never use 3-D effects — perspective distorts every length and angle it touches
• Label directly: put values on or beside the bars and names at the ends of lines, rather than forcing the reader to shuttle to a legend
• Use colour to carry meaning — highlight the one bar the message is about, mute the rest
• Round the numbers on labels; put the source, the sample size, and the scale in a footnote

Visualisation: quick reference

One connected research process

• The course has been a single journey in disguise — the research process from session 1, taken one stage at a time:
  • From problem to question (session 2), hypotheses and design (session 3), ethics (session 4)
  • Literature and secondary data (sessions 5–6)
  • Choosing approaches — quantitative, qualitative, mixed (sessions 7–9)
  • Sampling and recruitment (session 10)
  • Collecting data (sessions 11–12)
  • Analysing it (sessions 13–14)
  • Reporting it (today)
• The chain metaphor holds to the end: each choice constrains the next, and the study is only as strong as its weakest link

The distinctions that carry the course

• If you hold these securely — and can apply them to a scenario — you are ready:
  • Qualitative vs quantitative — and matching the method to the question
  • Induction vs deduction — building theory up vs testing it
  • Probability vs non-probability sampling — and what each licenses you to claim
  • Validity vs reliability — accuracy vs consistency
  • Correlation vs causation — association vs a defensible causal claim
  • Findings vs recommendations — what the evidence showed vs what to do
• Every one of these has appeared at least twice in the course, because every one of them recurs in real research

The examination: format

• A written examination, consisting of open and/or closed questions
• Scored out of 100 points; the pass mark is 51
• Closed questions test recognition of core concepts; open questions test whether you can explain and apply them
• Everything examinable was covered in the sessions and is summarised in the handouts
• Bring nothing but a pen and your understanding

The grading scale

How to revise

• The handouts are your study guides — each one was written to summarise its session completely, with every definition and distinction you need
• Understanding beats memorising: expect application questions — “which design fits this scenario, and why?”, “which sampling strategy, and why?”
• For each distinction, close the handout and reconstruct the comparison table from memory, then check
• Attach your own organisational example to every concept — an example you built is far more retrievable than one you read
• Practise on scenarios: today’s exercise is a rehearsal for exactly the kind of thinking the exam rewards

Example exam-style questions

• Closed: What is a key characteristic of non-probability sampling? (a) it guarantees a representative sample; (b) participants are selected by convenience or accessibility; (c) it requires a complete list of the population; (d) it is used mainly for large-scale surveys
• Closed: What is the primary aim of the explanatory sequential mixed-methods design? (a) to generate quantitative data from qualitative findings; (b) to run both strands simultaneously; (c) to use qualitative data to explain quantitative results; (d) to replicate quantitative findings
• Open: A company’s survey shows that stores with lower engagement have higher turnover. Explain why this alone cannot establish that low engagement causes turnover, and name one design that could support a causal claim
• Open: A firm’s customer satisfaction is falling. Plan a small study to diagnose why: state a research question, a design, a sampling strategy, and one limitation

Conclusion

• Reporting is honest and audience-shaped: findings first for decision-makers, methods complete and in reserve for sceptics, and the executive summary as a one-page craft of its own
• Accurate reporting means showing uncertainty, owning limitations, separating findings from recommendations, refusing to overclaim causality, and reporting the results you did not expect
• Honest charts choose the right form, start bars at zero, carry one message, and label it directly
• The course was one connected chain — problem, question, design, ethics, literature, approach, sample, data, analysis, report — and the exam rewards those who can apply its distinctions, not just recite them
• Revise from the handouts, practise on scenarios, pass at 51, aim well above it — and thank you: it has been a pleasure taking this journey with you

Exercise

Study guide