Methodology
Data-led Digital PR methodology: how I find and test a data story before it becomes a headline.
Data-led Digital PR uses original analysis of public data, surveys or a client's own figures to create findings that journalists can report. My job is to find a story worth telling and make sure it holds before it reaches a journalist.
In my experience, most weak data stories don't fail in the maths. They fail earlier: the question was too loose, a source counted something different or the headline claimed more than the numbers could support. This page explains how I find a story and what I check at each step.
- A story idea that connects your client’s work to a public question.
- An early answer on whether the idea will hold up.
- Every number traced back to its source.
- Answers ready when a journalist asks how it was done.
Why this matters now
AI can produce a convincing statistic in seconds. It won't reliably notice that a blank cell was read as a zero or that a count should have been a rate. Someone still has to check.
Journalists feel the difference. In Cision's 2026 State of the Media survey of 1,899 journalists across 19 markets, including Australia and the UK, half cited accuracy, fact-checking and combating misinformation as one of their biggest challenges. A number whose source and method cannot be explained creates a risk for the agency that pitches it.
1. The public question: find a story worth telling
A client's product is rarely a story on its own. A journalist needs a question that matters to readers. So the first thing I look for is a public question the client can credibly help answer. The story should lead and the brand should support it.
For Independent Steel, a steel supplier, the product was not the story. So I asked a question a trade journalist would care about: how much more metal could Australia recover if every state matched the best-performing state? The supplier then appeared as the company that did the analysis, not as an advertiser. Waste Management Review reported the findings.
That question is a “what if” scenario and not a forecast. The method note says so. Keeping what the data shows separate from what we assume is part of the story.
I do this with your team. You know the client and the journalists you pitch. I bring a public question and check whether the data can answer it.
How I choose between story ideas
How an idea becomes a story angle
- 1What people care about
- 2Several ideas
- 3Score each idea
- 4Pick one we can deliver well
- 5Angle from the analysis
Already have an idea? We start with the test: can the data hold it?
I rarely start with one idea. I start with what people already think and worry about. I read surveys of public concern and surveys or reports on the client's sector. For a travel company, that means the country's travel surveys and reports. AI now helps a lot with this research.
Then I list several ideas that connect the client's work to a wider public question people care about. I also look at the journalists the story is for: what they cover and what they have written recently. When I work with your outreach team, I ask for their research on journalist beats. If there is none, I do it myself or help the team do it.
Then I score every idea. The exact criteria vary by project, but broadly they are these:
- Does it fit the client?
- Does it add something new?
- Are people searching for it now? Is there a time of year when they do?
- Is there public data that can answer it? How granular is it: national, state or local?
- Can it be shown in a simple chart or table that still says something?
- How much time will it take to build?
- Can it be refreshed next year?
Finally, I pick the idea we can deliver well within the budget. That is not always the idea with the highest score.
One frame I often use is the gap between where the country is and a goal it has set. Another is the gap to its best performer. The brand then becomes part of the answer, not the subject.
The final angle often comes out of the analysis itself. For Independent Steel, the angle that ran was one of 21 angles the analysis produced. Sometimes the angle comes even later. For Aquatots, the finding ABC News reported came from an updated version of the report, after it was first published.
Good ideas rarely come from staring at a spreadsheet. Many come from conversations with your team and from reading what journalists cover. That is why I like to be involved early.
If your team already has an idea, we start one step later: I test whether the data can hold it before anyone commits hours.
The question a journalist asks:"Why would my readers care about this?"
2. The research question: make it testable
Before I touch the data, we agree on what the study needs to answer, for whom, across which places and years and what you need at handover.
“Do people in India have digital skills?” sounds like a question. It isn't testable yet. For a data story in The Hindu, my co-author and I measured digital skills through specific tasks from a national survey: using the internet, sending an email, copying and pasting, doing arithmetic in a spreadsheet and banking online. Once the question was that specific, the data could answer it. Only 37.8% of people aged 15 and above could bank online in 2022–23.
When I'm not sure whether the data can support an idea, I start with a feasibility review. The answer can be: proceed, revise the question or stop. Stopping early is cheaper than defending a weak story later.
The question a journalist asks:"What exactly did you measure, where and over which years?"
3. Source definitions: find out what each source really counts
Two datasets can use the same word and count different things. So before I combine anything, I check the definitions, years covered, units and collection methods. With surveys, I also check the sample size behind each group I report on. I check whether questions allowed more than one answer, because the percentages may then add up to more than 100%. I also look at how the sample was chosen and weighted. A difference smaller than the margin of error is not a headline.
Sometimes the thing we want isn't measured at all. Then I use a proxy: a measure close to it. I say plainly what it counts and what it doesn't. In the Aquatots study, I used the number of sports-instruction businesses as a rough indicator of access to swimming lessons. It's a useful clue, but it isn't a count of lesson places. The report says so.
The question a journalist asks:"Does your source actually count what your headline says?"
4. Missing data: look at the data before calculating anything
Problems often hide here. I check for duplicate records, missing values, place names spelt in different ways, numbers that look wrong and totals that don't add up.
Never assume a blank is a zero. Check what the source says a blank means.
In the drowning research for Aquatots, the Royal Life Saving database hides any count from one to four to protect privacy. In a small state or territory, that can hide many years. But the annual National Drowning Reports print some of these numbers. So I took each number from the report that printed it and noted where it came from.
| What the database shows | What I do |
|---|---|
| A published count | Use it as published |
| A hidden count (1 to 4) | Use the exact number if an official report prints it. Note the page. |
| Hidden, with no printed number | Leave it blank. Never read it as zero. Never guess. |
On one national ranking, gaps in the climate records would have quietly removed some coastal towns. I found this on the first day of work. Before building the ranking, I proposed a fix and the agency accepted it.
Every exclusion and change goes into the method note. Nothing just disappears inside a spreadsheet.
The question a journalist asks:"What happened to the places with no data?"
5. Counts vs rates: compare fairly
A count tells you how many. A rate tells you how many relative to the size of the group. A large group will have more of almost everything, so a count alone can mislead.
In the Aquatots research, one question was whether people born overseas make up a bigger share of drowning deaths than of the population. Counts alone could not answer it, because the two groups are different sizes. The fair comparison is each group's share of deaths against its share of the population.
People born overseas, Australia, 2018–19
share, %
A bigger share of drowning deaths than of the population.
So I check what each number is divided by, which years are compared and which groups sit side by side.
Small numbers need care too. In the Independent Steel study, the Northern Territory's waste volumes are small, so its recovery rate can move a lot with small changes. When groups have different age profiles, a fair comparison may also need age-adjusted rates.
And when two things move together, that alone doesn't prove one causes the other.
The question a journalist asks:"Is that a count or a rate?"
6. Rankings and index weights: show the choices behind a ranking
A simple ranking starts with at least three decisions: what is measured, which places are included and what happens to missing data.
An index that combines several measures needs more decisions: which measures are included, how they are put on a common scale, which direction counts as better and how much weight each measure receives. Each of those choices can change the ranking.
So I test at least one sensible alternative to see whether the order changes. In the Aquatots index, giving every measure equal weight did not change the ranking order. That is exactly the kind of question a journalist may ask, so I make sure the answer is ready.
I also look for near-ties where places are so close that their order means little. I check for measures that count the same thing twice.
In the Aquatots index, a score compares states. It is not a percentage or a measure of anyone's chance of drowning.
The question a journalist asks:"Would a different weighting change your number one?"
7. Headline wording: make the words match the numbers
Overclaiming usually shows up in the headline. I check that each finding matches the calculation behind it, that chart labels describe what was actually measured and that each limitation sits next to the claim it affects rather than being buried in a footnote. I also record how many ways I cut the data, so a striking result isn't just the best of many attempts.
For one example road-safety brief, the proposed angle was to tell drivers their chance of hitting a kangaroo. The data could show only recorded strikes. So I recommended reporting the recorded strike rate instead. It is a narrower claim, but one a journalist can verify and report. Read the feasibility review of the kangaroo brief (PDF).
✗ The first idea
"Your chance of hitting a kangaroo"
The data cannot support this claim. It contains only recorded strikes.
✓ What the data supports
"The recorded strike rate"
A narrower claim that a journalist can verify.
The question a journalist asks:"Can your data really say that?"
8. The handover: the findings and how I got there
You receive the agreed findings, charts, reusable tables, sources, calculations and a short method note. The proposal sets out exactly which files you will receive, how many review rounds are included and what support I provide during journalist outreach. The method note records each source's version and the date I downloaded it.
I hold every handover to one test, the same one StudioHawk's brief set: could another analyst reproduce the result from my files without asking me a single question? If not, it isn't finished.
This page describes my general approach. For each project, the method note records the actual sources, decisions and exclusions.
If you find an error on this site, email me. I will correct it and note the date on the page.
The question a journalist asks:"Can you send me the method?"
Where I fit
I do the research
The story idea, the question, the sources, the analysis, the charts and tables, the method note and answers when a journalist asks how it was done.
Your agency runs the campaign
The client relationship, design, media lists and pitching. We shape the story idea together.
Read the full case studies
StudioHawk: checking the data before the ranking went public →
Aquatots: drowning data and an access-gap index →
Independent Steel: recovery comparisons and a benchmark scenario →
“Samiran supported our data team on several Digital PR projects. His data for an El Niño summer campaign landed coverage across seven Seven West Media titles, including The West Australian and PerthNow. His work is clean, easy to use and well received by our team and journalists. Very knowledgeable, highly recommend.”