Grade 12 · Unit 7: Communicating with data¶

Track: A/B · Tools: charts, writing, optional Streamlit · Lessons: 3 Datasets: census-districts

Big idea. Analysis only matters if people understand and can act on it: action titles, honest charts, short memos and clear limits.

What you will be able to do

I can design a chart with an action title, honest scales, highlighting and a complete caption.
I can write a one-page memo with the bottom line first, evidence, uncertainty and a recommendation.
I can present findings and uncertainty in plain language, and publish work with credits and licences.

Words to know: audience, action title, caption, highlight, decluttering, bottom line up front (BLUF), memo, uncertainty, plain language, licence, attribution. Full list: curriculum/framework/glossary.md.

Notebook: notebooks/data_story.ipynb (runs offline on the synthetic table).

Before you start¶

This unit gathers communication skills you have practised since Grade 8 (Unit 7 posters, claim–evidence–reasoning) and raises them to a professional level. Setup:

import pandas as pd
import matplotlib.pyplot as plt

districts = pd.read_csv("districts_synthetic.csv")       # SYNTHETIC
by_province = districts.groupby("province")[["absentee_population", "population"]].sum()
by_province["abroad_per_1000"] = 1000 * by_province["absentee_population"] / by_province["population"]
national = 1000 * districts["absentee_population"].sum() / districts["population"].sum()    # 80.7

Lesson 1: Action titles and honest charts¶

Hook (5 min). Two slides show the same chart. One is titled "Figure 3: Absentee population by province". The other says "Province 6 loses the most people abroad: 94 in every 1,000". Which one will a busy official remember?

Explanation¶

Start from the audience and the message. Before drawing, write one sentence: what should the reader know or do after seeing this? That sentence becomes the action title.

Topic title (weak)	Action title (strong)
Absentee population by province	Province 6 has the highest share of people living abroad: 94 per 1,000
Literacy by belt	Mountain districts trail Hill districts by 11 points in literacy
PM2.5, 2019–2020	Kathmandu's air breaks the national standard almost every winter day

Design an honest, clear chart (a checklist): 1. Right chart for the message (Grade 9 Unit 6): bars to compare categories, lines for time, scatter for relationships. 2. Sort bars by value (unless there is a natural order). 3. Honest scales: bars start at zero; consistent axes when comparing charts; no 3D. 4. Highlight what matters (one colour), grey the rest; label values directly instead of a legend where possible. 5. Declutter: remove borders, heavy gridlines, unnecessary decimals. 6. Caption: what is shown, units, source and date, "synthetic" if so, and one limit.

s = by_province["abroad_per_1000"].sort_values()
colors = ["tab:red" if p == s.idxmax() else "lightgrey" for p in s.index]

fig, ax = plt.subplots(figsize=(7, 3.5))
ax.barh([f"Province {p}" for p in s.index], s.values, color=colors)
ax.axvline(national, color="black", linestyle="--", linewidth=1)
ax.text(national + 1, -0.6, f"national {national:.0f}", fontsize=9)
for i, v in enumerate(s.values):
    ax.text(v + 1, i, f"{v:.0f}", va="center", fontsize=9)
ax.set_xlabel("People living abroad per 1,000 residents")
ax.set_title("Province 6 has the highest share of people abroad (synthetic data)", loc="left")
for side in ("top", "right"):
    ax.spines[side].set_visible(False)
fig.text(0.01, -0.04, "Source: districts_synthetic.csv (invented numbers, CC0). Rates weighted by population.", fontsize=8)
plt.tight_layout()
plt.show()

The highlighted bar, the national reference line and direct labels let a reader get the message in seconds: province 6 (94 per 1,000) is highest, province 2 (56) lowest, national 81.

Activity (20 min) — PRIMM¶

Rewrite five topic titles from earlier units as action titles.
Predict what the highlighted chart will look like; run it.
Modify: highlight the lowest province instead and change the title to match.
Make: redraw a cluttered chart from an earlier project using the checklist; show before and after.

Common mistakes¶

Titles that describe the axes instead of the message.
Rainbow colours on every bar.
Missing source, units or "synthetic".

Check (5 min). 1. What is an action title? 2. What goes in the caption? 3. Why a zero baseline for bars?

Reflect (2 min). What surprised me? Who or what is missing from this data?

Lesson 2: The one-page memo¶

Hook (5 min). A ward chairperson has three minutes between meetings. You have ten pages of analysis. What do you give them?

Explanation¶

Decision-makers read the first lines and maybe the chart. Put the bottom line up front (BLUF).

Memo structure (≈ 250–400 words, one page): 1. Title = the key finding. 2. Bottom line (2–3 sentences): what we found and what we recommend. 3. Evidence: 3–4 bullet points with numbers, units and comparisons; one chart. 4. How sure are we: sample, uncertainty (intervals), limits. 5. Recommendation(s): specific, feasible, with who should act. 6. Details: data source, method in one line, contact; full analysis in an appendix.

Example memo (synthetic data):

Province 6 should be first for migrant-family support services

Bottom line: In our (synthetic) district table, 94 of every 1,000 residents of Province 6 live abroad, the highest rate of the seven provinces and above the national 81. We recommend piloting family-support desks for migrant households there first.

Evidence: * Province 6: 94 per 1,000; Province 4: 90; Province 2: 56 (lowest). * Within Province 6 the rate varies widely: Hill-02 (107) and Hill-06 (101) exceed 100 per 1,000, while Mountain-16 is at 53. * Rates are weighted by population, so large and small districts count fairly.

How sure are we: These are invented numbers for practice; real decisions need the 2021 census absent-population tables. Absentee counts rely on household reports and may miss whole households that left.

Recommendation: Validate with census data; if the pattern holds, start in Province 6 and Province 4.

Data: districts_synthetic.csv (CC0). Method: absentee ÷ population by province. Contact: …

Plain language. Short sentences, everyday words ("people living abroad" not "absentee population"), numbers rounded sensibly (94, not 94.0721), one idea per paragraph. Read it aloud: if you run out of breath, the sentence is too long.

Test with a reader. Give the memo to someone outside the project for 60 seconds, then ask them to tell you the main point. If they can't, rewrite the first lines.

Activity (20 min)¶

Draft a 250-word memo for a finding from your Grade 11 or 12 work, using the structure.
Swap with a partner; they summarise it back in one sentence after 60 seconds.
Revise the bottom line until their summary matches what you meant.

Common mistakes¶

Starting with the method ("We downloaded… we cleaned…").
Hiding uncertainty or limits in the appendix.
Recommendations nobody can act on ("more research is needed" alone).

Check (5 min). 1. What goes first? 2. Why say what you don't know? 3. Who is your audience?

Reflect (2 min). What surprised me? Who or what is missing from this data?

Lesson 3: Presenting and publishing¶

Hook (5 min). "The trend is 0.28 °C per decade (95% CI 0.18–0.39)." How would you say that to a farmers' group, a newspaper and a scientist?

Explanation¶

Presenting uncertainty in plain words

Technical	Plain language
95% CI 0.18–0.39 °C/decade	"warming by about a quarter of a degree every ten years — very likely between a fifth and two-fifths"
p = 0.03	"a difference this big would be unusual if it were just chance"
R² = 0.09	"these factors explain only a small part of why babies' weights differ"
recall 0.74	"the list finds about three in four of the women who need a visit"

Use natural frequencies ("3 in 4", "94 in every 1,000") rather than percentages of percentages, and ranges rather than false precision. Show uncertainty visually: error bars, shaded bands, or a few plausible lines.

A short talk (3 minutes): one sentence of context → the main finding (action title) → one chart, pointing at the key part → how sure we are → recommendation → invite questions. Practise with a timer.

Publishing your work responsibly * Credit and licences: data sources with licences (CC BY requires attribution; BY-SA requires share-alike; some data cannot be republished at all, e.g. DHS); your own text and code under an open licence if you choose (this curriculum uses CC BY 4.0 for content and MIT for code). * Privacy: no personal data; aggregate; small-group rule (Grade 8 Unit 8). * Reproducibility: link the notebook or repository (Grade 11 Unit 8). * Formats: a PDF report, a poster, a web page, a Streamlit dashboard (Grade 10 Unit 7). Students who want to build full websites can explore web frameworks such as Django or Flask, but they are not required: a clear memo and chart are what decision-makers use. * The student gallery: with consent, publish an anonymised write-up in community/student-gallery/ of this repository.

Activity (20 min)¶

Translate four technical statements from your earlier projects into plain language for two audiences.
Give a 3-minute talk with one chart and one limit; collect two stars and a wish.
Write the credits and licence section for your capstone (Unit 8) publication.

Common mistakes¶

Reading slides aloud; too many charts.
Dropping uncertainty to sound confident.
Publishing without licences, credits or privacy checks.

Check (5 min). 1. How to show uncertainty plainly? 2. What do you publish with data work? 3. Is Django required?

Reflect (2 min). What surprised me? Who or what is missing from this data?

Unit summary¶

Message first: action titles; right chart; sorted, highlighted, decluttered, honest scales; captions with units, source, date, limits.
Memos: bottom line up front, evidence with numbers, how sure we are, specific recommendations, details last; plain language; test with a reader.
Uncertainty in plain words: natural frequencies, ranges, visuals.
Short talks: context → finding → chart → certainty → recommendation → questions.
Publish with credits, licences, privacy checks and links to reproducible work.

Practice exercises¶

★ Rewrite as an action title: "Monthly PM2.5, 2019–2020".
★ What three things should every caption contain?
★ Turn "OR = 1.41 per visit" into plain language.
★★ Draw the province chart sorted from highest to lowest with a vertical bar chart instead. Which is easier to read? Why?
★★ A memo starts: "This report presents the methodology and results of an analysis of district-level data…" Rewrite the first two sentences in BLUF style.
★★ Make a chart with error bars showing the bootstrap 95% interval of the facility-delivery rate for each belt (Grade 11 Unit 3). Give it an action title.
★★ Write the credits for a map that uses OSM roads, HDX boundaries and census 2021 data.
★★★ Critique a chart from a Nepali newspaper or government report using the six-point checklist and redraw it.
★★★ Your finding is uncertain (interval includes 0). Write a memo bottom line that is honest but still useful.
★★★ Write two versions of the same finding: for a ward chairperson and for a newspaper headline plus first paragraph. Discuss what changed.

Exercises with real data¶

R1. A real migration memo (RD-13). Using the Census 2021 absent population by province (or district), redo the lesson chart and memo with real numbers. Compare with the synthetic story — did the conclusion change? Offline fallback: the synthetic table, as in the lessons.

R2. A climate chart for farmers (RD-07 / RD-06). Make one chart and a 150-word note for farmers in your district about temperature or rainfall change, with uncertainty in plain words.

R3. Fix a published chart (RD-02 / RD-01). Find a chart about Nepal in the news. Recreate it from OWID or World Bank data, improve it with the checklist, and write a two-line explanation of the changes.

Unit project: Story and memo¶

Follow Problem · Plan · Data · Analysis · Conclusion. Hand in one finished chart (before and after versions), a one-page BLUF memo for a named decision-maker, a 3-minute talk with feedback notes, and a publication section with credits, licences and privacy check. Full brief and rubric: project.md.

If you use an AI assistant, follow the AI-use policy (explain, don't answer; disclose; be ready for an oral check).

Going further¶

Cole Nussbaumer Knaflic, Storytelling with Data (2015).
Alberto Cairo, How Charts Lie (2019).
Edward Tufte, The Visual Display of Quantitative Information (2nd ed., 2001).
The Economist and Our World in Data chart style guides (search online) — models of action titles and clean charts.