Research impact

Two things are called impact

They get used interchangeably and they are not the same thing.

Academic impact is what happens to your work inside the research system: it gets cited, it gets built on, it lands in a journal people in your field read. It is countable, it appears on your CV, and it is what promotion committees have historically looked at.

Real-world impact is what changes outside the research system because the work exists: a clinician does something differently, a device gets built, a standard gets revised, a policy shifts, a piece of software becomes the thing everyone in the field uses. This is what funders now ask you to write about, and what most people mean when they say the work mattered.

The uncomfortable relationship between them: the first is easy to measure and the second is what you actually want. So people measure the first and talk about it as if it were the second. Don’t. Almost everyone reading your application knows the difference, and the ones who don’t will be replaced by someone who does.

Outputs are not impact

The single most common mistake in an impact statement is describing outputs and calling them outcomes. It helps to have the chain in your head:

A paper is an output. A paper that changed what a clinic does is impact. Most PhDs finish somewhere around outputs and early outcomes, and that is completely normal — the honest move is to be precise about where you are, not to inflate an output into an impact and hope nobody checks.

Four kinds of impact

The NHMRC framing, which UQ uses too, and a useful checklist because it stops you defaulting to “more citations”:

• Knowledge — the field understands something it didn’t. New method, disproved assumption, a dataset others now use.
• Health — patient outcomes, clinical practice, diagnosis, guidelines, health service delivery.
• Social — public understanding, education, equity, quality of life, community practice.
• Economic — cost saved, productivity, commercialisation, jobs, a licence, a spin-out.

For most of the work that happens in this building, the honest answer at thesis submission is knowledge, with a plausible pathway sketched toward one of the others. That’s a real answer. Write that.

Plan it early, because reconstructing it is miserable

Impact evidence is cheap to collect on the day it happens and expensive to reconstruct three years later. The email from the clinician who used your protocol is trivially findable this week and effectively gone by the time you’re writing a fellowship application.

So the actual skill here is not “generating impact.” It is noticing and recording the impact that your work is already having, and doing a small amount of deliberate planning about who it’s supposed to reach.

Map who it’s for

Before the reach, the audience. For each group: who are they, what do they already read or attend, and what would they have to see for this to be useful to them?

Most researchers can only realistically serve two or three of these. Picking them deliberately beats gesturing at all six.

Write the statement before you need it

Grant applications, confirmation milestones, job applications and progress reviews all ask a version of the same question, usually in about 150 words. Drafting it early is useful even if nobody reads it, because it forces you to answer so what while you can still change the research to have a better answer.

Metrics, used honestly

You will be asked for numbers. Here is what each one actually is.

Two documents are worth knowing by name, because citing them is how you signal you understand this:

• DORA — the Declaration on Research Assessment, out of a 2012 meeting of the American Society for Cell Biology. Its core ask: stop using journal-level metrics as a proxy for the quality of individual papers or individual researchers. Thousands of institutions and funders have signed it.
• The Leiden Manifesto — ten principles for research metrics, published in Nature in 2015 by Hicks, Wouters and colleagues. The one to remember: quantitative evaluation should support qualitative expert assessment, not replace it.

Keep the receipts

This is the part that pays off, and it takes about two minutes a month.

Open one file — a spreadsheet, a note, anything you’ll actually reopen — and log things the day they happen. What goes in it:

• Uses of your work. Someone emails asking for your protocol, your code, your data. Someone forks the repository. A group asks you to run their samples.
• Invitations. Talks, reviews, panels, committees. Note who invited you and why.
• Citations that matter. Not the count — the specific ones: a clinical guideline, a policy document, a systematic review, a patent, a competitor building on your method.
• Media and attention. An interview, a news write-up, a post that travelled.
• Collaborations begun, and what started them.
• Teaching and training. Students supervised, workshops run, materials others adopted.
• Numbers with dates. Downloads, GitHub stars, dataset accesses — captured at a point in time, because these are hard to reconstruct backwards.

For each: what happened, when, who, and a link or a saved PDF. Screenshot the guideline page. Save the email. Evidence beats recollection, and a link that worked in 2024 often doesn’t in 2028.

Set the record up once

Half of “tracking impact” is just making sure the systems can find your work and attribute it to you. This is a one-time job that everybody wishes they’d done in year one.

• ORCID — a free, permanent 16-digit ID that stays yours through every name change and institutional move. It solves the problem of being one of four people with your name. Connect it to Crossref and DataCite and new works push themselves onto your record. If you do only one thing from this session, do this one.
• UQ eSpace — where UQ’s record of your outputs lives, and what feeds your UQ Researchers profile. If it’s wrong there, it’s wrong everywhere the university looks. The library’s guide to managing your outputs covers the mechanics.
• Google Scholar profile — imperfect and over-inclusive, but it’s the first thing most people find when they search your name, and you control it. Claim it, prune the wrong entries.
• CRediT — the 14-role contributor taxonomy, an ANSI/NISO standard since 2022 and increasingly required by journals. Agree roles with your co-authors before submission. It is far easier than renegotiating author order afterwards, and it gives you something specific to claim later.
• Make the work findable. A preprint, an open-access version in the repository, code with a licence and a DOI via Zenodo. Nothing raises the ceiling on impact more cheaply than being readable by someone without a subscription.

PROBLEM   Who has this problem, and what does it cost them?
WORK      What did you do about it? One sentence.
CHANGE    Who does something differently, or could, because of it?
EVIDENCE  What shows that — or, honestly, what would show it?

Templates you can steal

Asking someone who used your work what came of it — send this the moment you find out:

Hi [Name] — I saw you’d been using [the protocol / the dataset / the code] from [paper]. That’s great to hear. If you get a chance, I’d love to know what you ended up doing with it and whether it worked for your setup — partly out of interest, and partly because I’m keeping a record of where the method’s been useful. No rush at all.

Agreeing contributions before submission:

Before we submit, can we agree CRediT roles? My read is: [Name] conceptualisation and supervision; me methodology, software, formal analysis, writing — original draft; [Name] investigation and writing — review & editing. Shout if that doesn’t match how you saw it.

Offering your work to the people who could use it — a cold email that isn’t self-promotion, because it leads with their problem:

Dear [Name], I read [their guideline / report / paper] and noticed [specific point]. We’ve just published work that speaks to that directly: [one sentence on what you found and what it would mean for them]. Happy to send the paper, the data, or to talk it through if it’s useful. Either way, thought you’d want to know it exists.

The honest summary

You cannot manufacture impact, and a research degree is mostly too short to demonstrate much of it. What you can do is: be clear from the start about who the work is for, make it findable and attributable, and write things down as they happen. That’s it. Do those three, and in four years you will have a real story to tell with evidence attached, while someone equally good is spending a fortnight trying to remember who emailed them in 2026.

Tools and reading

• ORCID — free, permanent researcher ID. Fifteen minutes, once, forever.
• DORA — the case against journal-level metrics for individual assessment, and a database of institutions changing their policies.
• The Leiden Manifesto — ten principles for using metrics responsibly. Two pages.
• CRediT — the 14 contributor roles. Read it once, use it on every paper.
• UQ Library: research impact evidence — what counts as evidence, organised by type. The most directly useful page on this list if you’re at UQ.
• UQ Library: metrics for grants and promotions — which numbers to quote, and how, for ARC and NHMRC applications.
• Altmetric — tracks news, policy and guideline mentions. Treat the score as noise and the policy citations as gold.
• Zenodo — free DOIs for code, data, slides and posters, so the things that aren’t papers can still be cited.

Where to go next

Reproducible research is the practical other half of this: work nobody can re-run is work nobody can build on, and code with a licence and a DOI is the cheapest impact lever there is. Surviving peer review covers getting the paper out in the first place.

Download the slides (.pptx)