GOV.UK defines this behaviour as: "Seek out opportunities to create effective change and suggest innovative ideas for improvement. Review ways of working, including seeking and providing feedback." (Success Profiles: Civil Service behaviours, GOV.UK, Open Government Licence v3.0.) The operative word panels listen for is effective — not the size of the idea, but whether the change worked, stuck, and survived you moving on.
This page gives you the question in its common forms, four marked model answers from different settings and grades, and the follow-ups panels use to separate genuine improvers from people who once had a suggestion. The strongest evidence for this behaviour is usually small, finished and adopted — not big and proposed.
One calibration note: at junior grades, panels reward improving your own patch and influencing your team to adopt it. At senior grades they look for changing how change happens — creating the conditions where other people improve things, and handling the resistance that real change generates.
Practise these questions out loud with a live AI interviewer — free. Start with Try Me
How the marking guidance works
Each model answer below is marked against the four criteria a live aurate session scores:
See how a full session is scored
aurate is a practice tool. Marking guidance describes what strong practice answers show — it isn't an employability assessment.
Why it's asked: The headline form. Panels hear a lot of process-tweak answers — the differentiators are a sharp diagnosis, evidence of adoption by other people, and what happened after you stopped pushing it.
In a casework team I noticed our rework pile had a pattern: around half the cases sent back by checkers failed on the same two fields — dates in the wrong format and a declaration box left unticked. Everyone knew rework was annoying; nobody had counted why it happened.
I logged the reasons for a fortnight, which took ten minutes a day, and took the tally to our team leader with a one-page suggestion: move those two fields to the top of the checklist we already used, and add a single example of a correctly formatted date to the case template itself. Not a new process — a change inside the existing one, because I'd seen new processes die in that team before.
She let me trial it with our pod for a month. Rework on those two errors dropped to a handful of cases, the other two pods adopted the template, and the checking team later told us their queue moved noticeably faster. Two years on, the example date is still in the template — which I take as the real test of an improvement.
Marking guide
Why it's asked: The resistance form. Panels use it to find out whether your improvement evidence has ever been tested by someone saying no — and how you behaved when it was. Answers where resistance evaporated after one meeting usually get probed.
I proposed replacing a paper handover sheet between our listing team and the ushers with a shared digital note. The resistance was immediate and reasonable: the senior usher had run the paper system for nine years, it had never failed in a way that mattered to him, and two near-retirement colleagues found the tablet workflow slower than their own handwriting.
I didn't argue the principle; I went after the actual objections. I shadowed the ushers for two mornings and found the genuine problem wasn't the paper — it was that listing changes after 4pm never reached them. So I narrowed my proposal to exactly that gap: paper stayed for the morning handover, and any change after 4pm went to the shared note, which the duty usher checked once at 8:30am. Half the change I'd wanted, all of the benefit that mattered.
Missed late changes went from roughly one a week to one in the first three months. A year later the ushers themselves asked to move the morning handover onto the note as well — which taught me more about sequencing change than any full rollout would have.
Marking guide
Improvement stories invite the sharpest follow-ups
'Did it stick?' 'Who resisted?' 'What did it cost?' — this behaviour draws probing questions because weak versions of it are everywhere. A live aurate session pressure-tests your example with those follow-ups and marks it on the same four criteria used above. Two free sessions. No credit card.
Try it freeWhy it's asked: The definition's second sentence ('seeking and providing feedback') gets its own question surprisingly often. Panels want the loop closed: feedback gathered, something changed because of it, and the change communicated back to whoever gave it.
Our housing repairs line had a complaints pattern we kept misreading: residents weren't mostly angry about slow repairs — the free-text comments, when I actually sat down with 60 of them, were dominated by not knowing when anyone was coming. The service had been improving repair times for a year while satisfaction flatlined, because we were fixing the wrong thing.
I proposed we stop treating the complaint categories as the truth and re-code a sample of raw comments each quarter. On the strength of the first analysis, we introduced a simple commitment: a text the day before any visit with a morning-or-afternoon window, and a call if the operative would miss it. The repairs themselves got no faster.
Complaints about the service dropped by about a quarter in six months, and the tenant panel — who I went back to with a one-page 'you said, we did' — moved from our sharpest critics to the group that proposed the next improvement. The re-coding habit outlived the project and caught a different mis-categorisation the following year.
Marking guide
Why it's asked: The negative form, and a favourite — it tests whether your relationship with change survives losing. Panels look for a fair account of why the idea lost, what you salvaged, and whether you stayed an improver afterwards.
At a charity I proposed consolidating our three volunteer databases into one system. I'd costed it, the inefficiency was real, and the proposal was rejected — partly on budget, but mostly, as the operations director told me plainly, because I'd built the case on staff time saved and never asked the volunteer coordinators what they would lose. Two of the three databases held local notes the central system couldn't represent.
I took the rejection as a review of my method, not just my idea. I went back to the coordinators, mapped what each system actually did for them, and returned six months later with a smaller proposal: keep the local systems, but standardise the one thing that genuinely caused harm — how availability was recorded — so volunteers stopped being double-booked.
That version cost almost nothing and was adopted in a month. Double-bookings fell from a steady trickle to rare, and when funding later allowed a proper systems review, the coordinators asked for me on the working group — because I'd learned to start with what they'd lose, not what the organisation would save.
Marking guide
Why it's asked: A judgement question hiding behind an open one. Panels want a filter — effort against impact, evidence thresholds, who benefits — plus an honest example of an idea you dropped because it failed your own test.
Why it's asked: The self-directed form. The differentiator is specificity of the change: name the piece of feedback, the adjustment you made, and evidence it stuck. Generic growth mindset language scores poorly against one concrete alteration.
Why it's asked: A situational form that tests method under ignorance. Strong answers resist proposing changes and describe a diagnosis routine instead: watch the work, count something, ask the team what wastes their time, find what previous improvers broke.
Why it's asked: A humility-and-judgement form panels increasingly use. The behaviour rewards effective change, not authorship — evidence you found, adapted and credited someone else's working solution can outscore a homegrown idea.
Why it's asked: The strengths-format variant — scored on energy and authenticity, not structure. Keep it under a minute, be honest about the flavour of improvement you enjoy (diagnosing, simplifying, automating, persuading), and anchor it with one quick real moment.
It is the Success Profiles behaviour about creating effective change: spotting opportunities, suggesting improvements, and reviewing ways of working including through feedback. Panels score whether your change was diagnosed from a real problem, adopted by others, and durable — not how large or novel the idea was.
Small and finished beats big and proposed. A two-field checklist change that still exists two years later is stronger evidence than a transformation programme you contributed a workshop to. What panels need is your diagnosis, your action and a number that moved — scale is the least important variable.
Yes — for the failure-form question it is exactly what panels want, and even for the headline form a rejected-then-revised idea can score well. The evidence is in what you did next: how you treated the rejection, what you changed about your approach, and whether an improved version eventually landed.
Authority is not required — influence is. Improving your own work products, evidencing a problem with a simple count and taking it to whoever does have authority, or adapting how you support a process are all scoreable. Panels at junior grades expect exactly this shape: improvement by evidence and persuasion.
Yes for the 'tell us about a time' forms — and give the Result two parts where you can: the number that moved, and the evidence the change stuck (still in use, adopted by another team, requested elsewhere). Durability is this behaviour's strongest closing beat.
Two free sessions. No credit card.