Somebody emails you a spreadsheet. It's got 4,000 rows of scraped review data, columns you didn't design, and a note that says "thought this might be useful for a study." You open it. You close it. You open it again.
That's usually where original research starts—not with a brilliant hypothesis, but with a messy file and a vague feeling that something is in there.
I've built three data studies from scratch and contributed to a handful more. Two earned links. One flopped so badly I still cringe thinking about it. The difference wasn't the quality of the analysis. It was whether I'd asked, before touching a single row, "Who would link to this and why?"
Key Takeaways
- Original research earns links because it gives journalists something they can't get elsewhere: a number nobody else has published.
- Pick your distribution target before you pick your dataset. The link prospect determines the study, not the other way around.
- Sample size matters less than representativeness—200 well-recruited respondents beat 2,000 from a biased panel.
- Most links from a good study arrive six to eighteen months later, not in the launch week.
- An aging study isn't dead. Refreshed data often outperforms the original launch.
Why original research is still the hardest link-building tactic to fake
Link building has a supply problem. Guest posts are saturated. Directory links are worthless. Digital PR stunts get coverage for a week and then vanish into a graveyard of "brand mentions, no link."
Data studies work differently because they create something scarce: a fact that didn't exist before you published it. When a reporter at a trade publication needs a number for a story about, say, how remote teams handle time zones, they can quote an industry report or they can quote your dataset. One of those is a link.
What is research that uses existing data called?
When you analyze data somebody else already collected—government statistics, public APIs, scraped listings, historical archives—that's secondary research (sometimes called desk research). It's the cheapest entry point into this whole game, and it's where I'd start if I had no budget.
Secondary research has one huge advantage: you skip the recruitment problem entirely. Public datasets from statistical agencies, open-source repositories, and company APIs are sitting there waiting. The tradeoff is that you're limited to questions someone else already thought to ask. You can slice and cross-reference, but you can't add a variable that was never measured.
Primary research—surveys, interviews, experiments you run yourself—gives you a blank canvas. It also gives you a recruitment bill, a timeline measured in weeks, and the risk that your sample turns out to be garbage. I've done both. Secondary research got me a link from a national news outlet with about eleven hours of work. Primary research got me three links and cost me roughly $900 in panel fees.
Neither is inherently better. It depends on the question.
What are 5 sources of data you can actually build a study around?
Here's what I've used, ranked roughly by effort-to-reward:
- Public statistical databases — national statistics offices, Eurostat, World Bank open data. Free, citable, and underused because they're boring to browse.
- Platform APIs — most major platforms expose some public data. Rate limits are annoying; the resulting dataset is yours.
- Your own product data — anonymized and aggregated, this is the most defensible research you'll ever publish, because nobody else has it.
- Surveys — 200 to 500 respondents is usually enough for a credible headline finding if the sample is clean.
- Scraped public web data — reviews, job postings, pricing pages. Legally murkier, journalistically very attractive.
Notice what's missing from that list: "asking your Twitter followers." I did that once. Got 87 responses, mostly from people in my own bubble, and the resulting "study" was so obviously skewed that no journalist touched it. Total waste of two weeks.
The methodology that gets you cited (and the shortcuts that get you ignored)
Here's the thing nobody tells you about data studies: journalists don't check your math. They check your methodology section. If it looks sloppy, they move on, because attaching your name to a bad number is a career risk for them.
What does a credible methodology section contain?
- Where the data came from, specifically
- When it was collected, and over what period
- How many observations, after cleaning
- What you excluded and why
- Any known limitations, stated plainly
That last point matters more than people expect. Stating limitations doesn't weaken your study—it signals that you understand it. One of my studies included a line admitting that our sample skewed toward larger companies. A reporter quoted that limitation directly in their piece, then linked to us anyway. Honesty got the link.
How big does your sample need to be?
For a survey-based study aimed at press coverage, 200 respondents is a practical floor, and 500 is comfortable. Below 200 you'll struggle to say anything interesting about subgroups, and any journalist who's seen a few studies will notice.
But size is the wrong thing to obsess over. I've seen a 3,000-person panel produce garbage because it was recruited through a rewards app that skews heavily toward one demographic. A 250-person sample recruited carefully will outperform it every time.
The bias problems that kill studies most often:
- Self-selection — people who volunteer for your survey aren't random
- Survivorship — analyzing only the companies that succeeded
- Recency — data from a single unusual month, treated as a trend
Fix what you can, disclose what you can't.
Where can you find links to research papers?
If you need to cite existing academic work—either to build on it or to position your study against it—the practical starting points are Google Scholar, PubMed for anything health-adjacent, SSRN and arXiv for preprints, and JSTOR if you have institutional access. Semantic Scholar is worth knowing too, because its citation graph helps you find newer papers that build on an older one.
For gray literature—reports from agencies, NGOs, and research institutes—search the organization's own site directly rather than relying on a general search engine. Much of it never gets indexed properly.
Distribution: the part almost everyone gets backwards
Most teams publish, then think about outreach. That order is wrong.
Before you write the first line of your methodology, you should be able to name ten specific journalists or bloggers who would plausibly cover your findings. Not "tech publications." Actual people, with actual beats, who have written about adjacent topics recently.
I learned this the hard way. My failed study was technically solid—clean data, decent sample, nice charts. It got four links. Four. Because I'd built it around a question I found interesting, and it turned out no one else did.
The study that worked was built backwards. I noticed a reporter had written three pieces in one year about a specific operational problem in our industry, each time citing thin data. So I built a study that answered the question they kept circling. They covered it within a week of launch and linked twice.
What outreach conversion rates should you actually expect?
Set your expectations low and you won't be disappointed. In my experience, cold outreach to journalists on a data study converts at roughly 5 to 10 percent—and that's with a personalized pitch, not a template.
The rest of your links come from somewhere else entirely: organic pickup. Aggregators, newsletters, forum threads, and smaller blogs that find the study on their own. That's the compounding part, and it's why studies outperform one-off campaigns. A study published eighteen months ago still brings me a link every few weeks.
| Link source | Typical timing | Effort required | Link quality |
|---|---|---|---|
| Direct outreach | Week 1–3 | High | High |
| Organic pickup | Month 2–12 | None after launch | Variable |
| Syndication / reprints | Month 1–6 | Low | Medium, sometimes nofollow |
| Academic citation | Year 1–3 | None | High authority, slow |
| Internal reuse in your own content | Ongoing | Low | No new links, but strengthens existing pages |
What to do when your study starts to age
Data has a shelf life. A study about platform usage from three years ago reads as historical, not current, and journalists stop citing it.
The fix is cheaper than building something new: refresh it. Re-run the collection with current data, update the charts, change the publication date, and re-pitch. You keep the accumulated links pointing at the URL, and you give it a second launch window.
I've refreshed two studies this way. The first one picked up more links in its refresh month than it had in the previous year combined. The original URL had built enough authority that the updated version ranked and got discovered faster.
One caution: if the findings change substantially, say so prominently. Quietly swapping numbers is the kind of thing people notice, and it destroys the credibility you spent months building.
The honest verdict on data studies as a link tactic
Data studies are slow, expensive relative to guest posting, and there's no guarantee the work pays off. My hit rate is somewhere around two in three—and that's after years of getting the targeting right.
But the links they produce behave differently from campaign links. They arrive without outreach, they arrive years later, and they come from places you'd never have reached with a pitch email. One study I published is still the single largest source of referring domains to my site, and I haven't touched it in over a year.
If you want a tactic that produces predictable links on a schedule, this isn't it. If you want a handful of assets that keep working while you sleep, there's not much that competes.
Start with a secondary-research study. Find a public dataset, ask one question nobody's asked of it, and publish the answer. It won't be glamorous. It might not even get covered. But you'll learn the whole pipeline—collection, cleaning, methodology, pitching—for almost nothing, and the next one will be better.
