Most Irish civil service and local authority competitions shortlist on the application form alone. Nobody meets you at that stage. Someone reads what you wrote, scores it against published criteria, and that score decides whether you go any further. Candidates know this, and still most of them write the form as though a sympathetic colleague will read it. The person who reads it is not a sympathetic colleague. They are a board member with a large number of forms to get through, and there is a fair amount of research on how people in that position actually read and score.
This post sets out what that research says, and what it means for how you write. I spent years on the candidate side of these competitions, at the Department of Foreign Affairs and Trade, and I now spend a good part of my working week reading draft forms with clients. Understanding the reader changed how I do that work, and I think it will change how you approach your own form.
One caveat first. There is very little published research on Irish public sector shortlisting specifically, so what follows draws on adjacent fields: recruiter screening, essay and performance rating, and competency-based assessment in medical and vocational education. The mechanisms transfer, because a sifter scoring forms against indicators is a rater working through a stack, and rater behaviour is well studied. But these are mechanisms that shift the odds, not laws. A given board member may be fresh, careful and well moderated. The safe assumption is the ordinary case, not the best one.
What shortlisting is, and why it is built the way it is
Where a competition attracts more eligible candidates than a board can interview, the board reads the application forms and scores each answer against the competencies or capability dimensions in the candidate information booklet. Often it is a subset of the full framework: the AP 2026 competition, for example, shortlisted against a selection of the Capability Dimensions with a minimum score required under each. Pass marks are usually fixed before the sift begins, and board members moderate their scores against each other. The Public Appointments Service and the local authorities both run on this model, with variations in the detail.
It is worth knowing why the process is built this way, because the reason is also the most useful thing you can know about the reader. The strongest and most replicated finding in the selection literature is that combining information about a candidate by rule predicts later performance better than combining it by overall impression. Grove and colleagues analysed 136 studies in 2000 and found rule-based prediction equal or better than expert judgement across a wide range of settings, with expert judgement substantially better in only a small minority of studies. Kuncel and colleagues repeated the exercise for selection and admissions in 2013 and found a consistent and substantial loss of accuracy when assessors combined the evidence holistically, even experts who knew the job and the organisation well. Highhouse explained in 2008 why practitioners resist this: most of us believe our own insight improves with experience, and the research says it does not.
A scored sift, then, is an attempt to make human readers behave like a rule. Published indicators, fixed pass marks, moderation: all of it exists to stop the board scoring on impression. It mostly works. But the structure works only on what is written. Moderation can correct two board members who disagree about evidence they can both see. It cannot recover evidence you never put on the page, and the board is generally not permitted to infer context or fill in gaps from goodwill.
Who is reading your form?
A board member or trained sifter, scoring a large number of forms against fixed indicators, usually in one sitting. They are not reading for interest. They are looking for material they can credit against each indicator, and they can only mark what is on the page, at face value, with no knowledge of you beyond it.
Everything that follows is about how a competent, well-intentioned person reads under those conditions. None of it is a criticism of boards. It is a description of how human attention behaves when the task is large and the criteria are many, and it is the reason the process has the safeguards it has.
One reading, not two
Research on recruitment professionals found that their search for candidates is characterised by satisficing rather than maximising: the goal is a sufficient number of suitable candidates in the shortest possible time, not the best possible review of everyone available. That study looked at recruiters searching candidate databases rather than boards reading forms, so the transfer is my inference, but the conditions are the same: many applications, limited time, and a target number to reach.
For you, that means each answer gets one pass from a reader looking for reasons to score it. Anything that needs a second reading to be understood will often not get one. An example that only makes sense once the reader reaches your final paragraph may be scored on the first two-thirds, which contained the scene-setting and not the evidence.
The reader is holding a lot in their head
The literature on rater-based assessment in competency frameworks is consistent on one point: assessors face a high mental workload when scoring against multi-part criteria, and when working memory is overloaded they fall back on shortcuts. That is where unreliability enters. The load is highest when the assessor has to hold several competencies or indicators in mind at once, which is exactly what a sift asks of them.
So an answer that maps plainly onto the indicators lowers the reader's load and gets scored on its merits. An answer the reader has to mentally reorganise before they can score it gets scored on impression instead. Structure that mirrors the marking task is not cosmetic. It changes which route the reader's mind takes to a score. If the booklet lists four indicators under a competency and your answer shows those four things in that order, in plain words, the reader's job is easy and you may well benefit from it. If the same evidence is scattered across a long narrative, the reader has to do your work for you, and a satisficing reader often will not.
A weak answer costs more than its own marks
Raters tend to give more similar scores across separate criteria than the evidence justifies. A strong overall impression lifts the individual marks and a weak one drags them down. Thorndike named this the halo effect in 1920 and it is among the oldest replicated findings in rating research. The best explanation for why it happens is undersampling: when the rater has too little concrete information on a specific criterion, they fall back on their global impression of the candidate.
The consequence is that the overall read of your form partly produces the per-question marks. One conspicuously weak answer, or a generally padded and generic register across the whole form, may cost you on every other answer. The reverse is also true: a form that establishes credibility early can buy a marginal answer the benefit of the doubt. This is why the faults that look cosmetic, the stacked adjectives and the "I am a strong communicator with excellent interpersonal skills", cost more than their local damage. They shape the impression through which everything else is read.
Why was my form not shortlisted?
Usually because the evidence was not on the page in a form the board could score, and I will come to the specific faults below. But there is a second, less comfortable answer, and clients deserve to hear it: sometimes, partly, it is the stack.
Studies of essay marking in sequence have found position effects in both directions: some show a middling answer scoring lower after strong ones and higher after weak ones, others show markers drifting towards the scores they have just given. The best recent evidence comes from Radbruch and Schiprowski, who used data from high-stakes admission and hiring interviews where candidates were assigned to evaluators and time slots more or less at random. They found a contrast effect: a candidate's assessment fell as the quality of the other candidates seen by the same evaluator rose, the previous candidate mattered most, and the effect was strongest when the two candidates were similar. That was interviews rather than written forms, and the studies do not all agree on direction, but they agree on the point: a score depends partly on where the candidate sits in the sequence, and a marker's standard drifts over a sitting. Experienced, well-calibrated raters are less susceptible, not immune.
You will also see fatigue effects quoted, sometimes with dramatic figures. Hold those lightly. The most widely cited study, on parole decisions, has been seriously challenged on its methods, and the honest position is that sequential judgement can drift with tiredness, the true effect is probably modest, and nobody should put a number on it. I will not.
What you can do with this is limited but real. You cannot control your position in the stack, so the form has to survive a tired reader in an unlucky position. That means the evidence goes first, the answers score adequately even when skimmed, and nothing important depends on the reader's patience.
What persuades a doubtful reader
Across the screening research, concrete and verifiable material moves candidates forward faster. Specifics reduce the reader's uncertainty, and uncertainty is what a satisficing reader resolves by moving on. Three findings sharpen this.
First, verifiability matters on its own. Levashina and colleagues showed in 2012, in a high-stakes selection study of more than 16,000 applicants, that asking people to elaborate on their claims lowered scores overall, and that the effect was larger where a claim could be checked: being accountable for something verifiable makes people more careful about what they say. That study is about applicants rather than assessors, so the next step is my inference, but it is a short one. A board knows that every candidate presents well, and a claim nobody could check is easier to discount than one that names a number, a date or a system. "Reduced processing times" asks to be believed; "reduced the average processing time from eleven working days to six over the year" asks to be scored.
Second, assessors infer personality from thin cues whether or not they mean to. Cole and colleagues found in 2009, with 244 recruiters, that inferences drawn from CV screening were unreliable, often invalid, and still drove judgements of employability. Vague text is not neutral filler. It hands the reader an uncontrolled guess about who you are.
Third, and this is the finding I would most like every candidate to know about, there is some evidence that bias fades as the reader processes more relevant information. Sachs and colleagues ran an experiment in 2024 in which 402 people read a job application one snippet at a time. Impressions of the majority applicant were more favourable than of the minority applicant after three snippets; after five, the gap had closed. The authors are careful about what that shows, because it is one experiment with one hypothetical application and the closing of the gap may owe something to the content of the later snippets. But the pattern is consistent with a long-standing idea about discrimination: prejudice does most of its work when the reader has little individual information to go on. If that is right, dense, relevant evidence placed early in each answer is not only a scoring tactic. It is also part of the candidate's protection against being read as a type rather than as a person.
What this means for how you write
The research above maps onto a short list of faults I see in most first drafts. Each has a fix you can apply yourself.
- Do not describe yourself; describe what you did. "I am an effective team leader" could be pasted into a stranger's form without edit, so it distinguishes nobody and a board cannot mark it. Replace every assertion of a quality with the action that demonstrates it. My guide to the civil service capability framework explains what the indicators under each dimension actually ask for.
- Put the evidence first. The reader may not reach the end of a long answer with their attention intact. Open with what you did and what changed, then give as much context as the question needs and no more.
- Make yourself separable from the team. "We delivered", "it was decided" and the passive voice all hide the person being marked, and evidence that cannot be attributed to you personally may score as if it were absent. I have written about I-centred language for interviews; it applies to forms in exactly the same way.
- Give results the board could check. Tracked volumes, dates, before-and-after operational facts, a decision that held up under scrutiny. "Successfully delivered" with nothing behind it is not a result. Private-sector metrics that a public sector board cannot picture read as imported.
- Answer the indicator, not the label. Candidates asked to show strategic thinking often describe a busy delivery period and add the word "strategic". Read the behavioural indicators line by line and check that each sentence of your answer serves one of them. If the strategic words were deleted, would what remains still read as strategy?
- Cut anything built for you rather than for the reader. Elaborate scene-setting, stacked abstractions, three clauses where one verb would do. Machine-written padding falls into this category too, and boards are increasingly alert to it; the publicjobs.ie honesty agreement prohibits inappropriate use of AI in any case. Aim for a voice that is action-led, concrete and plain.
If the difficulty is not the writing but finding the examples in the first place, that is a different problem with its own research behind it. My post on finding examples for local authority application forms explains why cold recall fails and how to cue your own memory, and the method is the same for civil service forms. The application form checklist covers the final pass before you submit.
The form is the whole case
Put the research together and the reader looks like this: competent, structured, and loaded; marking many forms in a sitting; scoring what they can see against the indicators; satisficing under time pressure; and subject to halo, position and drift effects that no amount of professionalism removes entirely. Your form's job is to make crediting you the easiest thing that reader can do. Every fault in the list above makes it harder: generic text and padding invite the halo penalty and an invalid guess about your character, over-engineering and buried evidence raise the reader's load, the vanishing candidate and the missing result create the uncertainty a satisficing reader resolves against you, and a misread competency means the evidence is aimed at the wrong target entirely.
Underneath all of that sits the finding the whole process rests on. The sift is designed to score against anchors, so the anchors are the target. Write to them and the form does well when the process works as designed, and still reads best when the reader drifts. Nothing you know and did not write exists. The form is the entire case.
I do not write or rewrite forms for clients, and the publicjobs.ie honesty agreement is clear that nobody should. What I do is read a draft the way the research says a board will read it, and then work with you in a session to surface the evidence your own experience already contains. If that would help, my application form review gives you a written guidance note on the draft and a 45-minute session to work through it; there is an express version for imminent closing dates, and the same service exists for local authority competitions. If you are earlier in the process and the examples are the problem, start with a finding and building your examples session instead. Either way, email me at sarah@dedico.ie.
Frequently asked questions
How does shortlisting work in Irish civil service and local authority competitions?
Where a competition attracts more eligible candidates than can be interviewed, a board reads each application form and scores the answers against the competencies or capability dimensions published in the candidate information booklet, often a subset of the full framework. Pass marks are usually set before the sift begins and board members moderate their scores. Only candidates who reach the required standard are invited to the next stage. The board scores what is written on the form; it does not infer or import experience the candidate has not described.
Why was I not shortlisted when my experience matches the job?
Usually because the evidence was not on the page in a form the board could score. Common causes are answers that assert qualities rather than describe actions, examples where the candidate's own contribution cannot be separated from the team's, results that are not concrete, and answers where the evidence is buried at the end. Position in the marking order can also play a part, and no candidate controls that, which is why a form needs to score adequately even when read quickly by a tired reader.
Does the shortlisting board read the whole application form?
Assume each answer gets one careful reading, not two. Research on recruiters suggests that people screening many applications aim for a sufficient shortlist in the time available rather than the best possible reading of every one. Evidence that only becomes clear on a second reading, or that sits in the final lines of a long answer, may be missed. Put the strongest evidence first.
Can I ask for a review of a shortlisting decision?
Most competitions run under a code of practice that provides for a review of a selection decision, usually an informal review requested within a short period of notification and a formal review beyond that. A review examines whether the process was properly applied to your application; it does not re-score the form on its merits. The exact procedure and deadlines are in the candidate information booklet for your competition.
Sources
Cole, M.S., Feild, H.S., Giles, W.F. and Harris, S.G. (2009) 'Recruiters' inferences of applicant personality based on resume screening: do paper people have a personality?', Journal of Business and Psychology, 24(1), pp. 5-18.
Cooper, W.H. (1981) 'Ubiquitous halo', Psychological Bulletin, 90(2), pp. 218-244.
Grove, W.M., Zald, D.H., Lebow, B.S., Snitz, B.E. and Nelson, C. (2000) 'Clinical versus mechanical prediction: a meta-analysis', Psychological Assessment, 12(1), pp. 19-30.
Highhouse, S. (2008) 'Stubborn reliance on intuition and subjectivity in employee selection', Industrial and Organizational Psychology, 1(3), pp. 333-342.
Kuncel, N.R., Klieger, D.M., Connelly, B.S. and Ones, D.S. (2013) 'Mechanical versus clinical data combination in selection and admissions decisions: a meta-analysis', Journal of Applied Psychology, 98(6), pp. 1060-1072.
Levashina, J., Morgeson, F.P. and Campion, M.A. (2012) 'Tell me some more: exploring how verbal ability and item verifiability influence responses to biodata questions in a high-stakes selection context', Personnel Psychology, 65(2), pp. 359-383.
Radbruch, J. and Schiprowski, A. (2025) 'Interview sequences and the formation of subjective assessments', Review of Economic Studies, 92(2), pp. 1226-1256.
Russell-Rose, T. and Chamberlain, J. (2016) 'Searching for talent: the information retrieval challenges of recruitment professionals', Business Information Review, 33(1), pp. 40-48.
Sachs, T., Homan, A.C. and Lancee, B. (2024) 'Impression formation of majority and minority applicants during resume screening: does processing more information reduce prejudice?', Journal of Applied Social Psychology, 54(7), pp. 387-404.
Thorndike, E.L. (1920) 'A constant error in psychological ratings', Journal of Applied Psychology, 4(1), pp. 25-29.