Your careers site looks convincing. Why do suitable candidates still leave?

Recruitment journeys are rarely one continuous session. Someone can view a vacancy on a phone while travelling, revisit it on a computer, discuss it with a partner, contact a recruiter directly or apply through an external platform. A candidate can also save the address for later. Standard session metrics cannot reliably identify these paths without intrusive tracking, and intrusive tracking is neither automatically lawful nor necessary for answering the editorial question.

Separate at least four observable states: a visit with no application start; an application that begins but is not submitted; an informed decision that the role does not fit; and a later or external route that the website cannot observe. These should not be collapsed into the label 'lost applicants'. A lower start rate might reflect clearer job information and better self-selection. A lower completion rate caused by broken forms is different. Supplement aggregate observations with voluntary interviews or task tests, then check how common a suspected mechanism appears in an appropriate sample. Distinguish measured behavior from the researcher's explanation of that behavior.

This comparison is an investigation guide rather than a classification of people. Several mechanisms can operate simultaneously, and the same person can react differently in another context.

The goal cannot be to push every visitor through the same application button. Better information should help suitably interested people proceed, while allowing others to rule a job out early without manipulation or unnecessary effort.

Jobseekers evaluate instrumental attributes such as salary, hours, contract arrangements, travel and training. They also infer symbolic characteristics: dependable, supportive, prestigious, closed, informal or excessively demanding. Lievens and Highhouse report that symbolic trait inferences add explanatory value to job and organisational attributes in their studied populations [S07]. That helps explain why a perfectly accurate description can still feel unsettling if the visual and verbal context suggests a different kind of organisation. It does not prove that a certain colour palette or photograph always has the same effect.

Prior reputation matters. Reviews, employee contacts, advertising and news can establish a starting image before someone opens the site. Research on employer reputation and website characteristics suggests these factors can interact in predicting attraction [S05] [S09]. A polished site can reinforce an existing reputation, contradict it or make an unknown employer seem less legible. Therefore record what the candidate thought before exposure and what changed afterward. Otherwise you may credit the website for an association acquired elsewhere, or blame it for a concern that was already present.

'You will have freedom to make a difference' is a pleasant promise. It is not yet a description of how decisions are made, which tasks people own, what support exists or whether evening calls are mandatory. A candidate might read freedom as autonomy. Another might read it as being left to cope alone. Both interpretations are plausible until research identifies what the message actually evokes and whether the advertised work supports it.

Employer branding is therefore connected to job design and management practice, not merely to the style of a careers page. Reviews of employer image and employer branding describe a broader system of signals and experiences than advertising alone [S08]. Show what can be verified: a typical working day, a real training pathway, realistic supervision and material conditions. Where those details cannot be verified, solve the organisational information problem before polishing the copy. Better words cannot fix a workplace promise that is substantively untrue.

Werkhaven and this entire case are fictional. No real Neurofactor research, employer data, applicants, employee interviews or improvement rates are represented. The invented company maintains technical installations at business locations and wants to recruit field-service technicians. It uses an attractive careers site with the headline 'Make an impact every day', cheerful team imagery and a prominent Apply button. Its vacancy mentions flexibility, independence and varied days, but gives little detail about on-call shifts, escalation support, paid training or travel requirements.

Imagine three possible visitors, none of whom is a research participant. One is trying to work out whether the on-call pattern can be planned around family commitments. One needs reassurance that technicians can call for support on difficult assignments. One is interested in whether technical expertise can develop over time. All three might close the page. The identical behaviour does not imply identical motivation, nor are these 'candidate types' suitable for automatic profiling. The case creates questions to test with real participants in a properly designed and authorised study; it contains no claimed results.

For the imaginary Werkhaven site, the following matrix shows possible meanings a candidate might attach to elements. These are hypotheses, not verified reactions.

Do not respond by deleting every strong word. The real job may involve unpredictable events. If so, explain that honestly and show the support and limits, rather than creating a misleadingly reassuring picture.

A visitor arriving from a specific vacancy advertisement may be searching for salary range, base location, training requirements or exact shift patterns. Someone who browses directly from the homepage may be exploring culture, purpose or career opportunities. Both use the same website, yet they are answering different questions. If the research design ignores entry point and intent, it can wrongly attribute a mismatch between information and need to a negative employer association.

Begin with a short expectation interview before showing the page. Ask what participants believe they already know, which uncertainty they want to resolve and what would make the next step possible. Observe a defined information task without forcing an application. After exposure ask what they learned and what remains unclear. Compare the two moments without suggesting particular answers. A candidate saying 'I do not like this' provides useful context, but the investigator must still find out whether the reaction came from missing facts, a relevant incompatibility or a symbolic interpretation of the employer.

A form that loses entered information is first and foremost a usability failure. It may also encourage a visitor to infer that the employer is disorganised or unconcerned about applicants. That secondary interpretation is worth investigating but cannot be read from the error alone. Studies examining recruitment sites and online application systems link perceived usability and efficiency to applicant reactions [S03] [S04] [S06]. The field study by Sylva and Mol is especially useful because it concerns people actually interacting with an online application process.

These findings still require restraint. Different studies involve students, applicants at a particular organisation, or self-reported attraction rather than recorded hiring. They do not entitle a marketer to promise that removing a field will cause a precise increase in qualified hires. Start by repairing verified technical problems and ensuring accessibility. Only then investigate whether the experience has changed perceived professionalism or trust. Technical repair and persuasive communication are different interventions and should not be evaluated using a single blended explanation.

A recruitment funnel is easier to understand when the indicators are treated as distinct observations rather than alternative names for success.

A high conversion rate could be achieved by withholding inconvenient facts, but that would be poor recruitment. A lower click-to-apply ratio combined with better understanding and fewer technical barriers may be more valuable, depending on the defined objective.

Good information includes the awkward parts of a role: travel, on-call periods, shift patterns, physical requirements, onboarding, decision rights and supervision. Realistic job preview research considers how giving more accurate information can affect applicant expectations and different recruitment or employment outcomes, with effects that depend on timing and the type of preview [S12]. It should not be simplified into the proposition that more disclosure always generates more applications.

Werkhaven could show a sample maintenance call from alert to resolution, including who handles escalation and what happens when a junior colleague needs help. That would make 'you are supported' more specific. If nights or unpredictable incidents genuinely form part of the job, they should remain visible. An unsuitable applicant stepping away after seeing those facts is not necessarily a failure. If the actual workload is unsafe, however, transparency does not excuse it. The employer must first address the underlying work practices, then describe the corrected reality truthfully.

These are hypothetical content alternatives for the invented company. They must be approved against the actual job before any real employer uses similar wording.

Specific evidence helps a candidate decide what the promise actually means. It also creates accountability: the organisation must be able to produce the policy, process or actual example on which each claim relies.

A role requiring permanent night availability will not suit every qualified technician. Making the requirement clear is fairer than hiding it behind 'flexibility'. The website has done useful work when somebody discovers the mismatch early. A different problem arises if the same qualified person cannot locate the salary or is discouraged by a form that fails on a screen reader. Both produce departures, but only the second is a preventable recruitment barrier attributable to the process.

The same distinction applies later in selection. The quality and perceived fairness of hiring procedures relate to applicant reactions in meta-analytic research [S13]. That evidence does not demonstrate that a specific careers-page element caused a first-visit drop-off. Keep the stages separate and collect only the evidence needed for the decision. Avoid recommending coercive nudges that increase submissions by making candidates overlook constraints that would matter in their working lives.

Imagine that Werkhaven tests version A, containing its original general recruiting language, against version B, containing the same actual job conditions but making the rota, support, hiring stages and privacy explanation easier to locate. Both versions use an equivalent, accessible layout. Before the test, the investigators identify the primary question: for example, whether candidates correctly understand the expected on-call schedule. They then choose measures that distinguish comprehension from attraction and technical completion.

In a controlled task a participant receives a consistent job description and may decide whether to continue to a simulated next step. A debrief asks what the participant believed about the role and what remained uncertain. No fictional result is reported: this is a proposed design, not a completed Neurofactor project. Better clarity might reduce inappropriate applications while helping suitable candidates proceed. The analysis must not label the first outcome a failure simply because the start rate is lower.

Each research method has a distinct job. A more technical instrument is not automatically a better instrument, and unnecessary personal tracking can create new problems.

Combine qualitative interpretation with relevant behaviour. Do not add EEG, biometric data or extensive analytics unless the research decision genuinely warrants them and appropriate protections exist.

Step 1 - Define the recruitment decision. Select one job family, entry context and stage. Step 2 - Check technical integrity. Test application links, mobile load, form errors, accessibility and measurement definitions before diagnosing psychology. Step 3 - Verify the job reality. Have managers and employees validate shifts, responsibilities, support and hiring promises. Step 4 - Form competing hypotheses. For each observed exit, record a meaning-related explanation and a practical alternative.

Step 5 - Conduct voluntary qualitative research. Observe what relevant participants seek and ask what each page element communicates. Step 6 - Choose fit-for-purpose instruments. Use comprehension questions, user tasks, open association work and, where justified, a validated indirect measure. Step 7 - Register the evaluation plan. Specify the main outcome, sampling logic and planned subgroup analysis in advance. Step 8 - Compare the versions fairly. Use a suitable randomised or otherwise defensible comparison and do not change the job facts between versions. Step 9 - Review downstream outcomes and fairness. Check appropriate applications, task success, candidate experience and unintended exclusion. Document uncertainty and corrective actions, not just the most favourable graph.

Recruitment research distinguishes early awareness, pursuit intentions, application actions and later selection experiences. Breaugh's research review identifies questions that require careful attention to context, recruitment method and applicant characteristics [S14]. Results from one stage cannot simply be transferred to another. A site can generate attraction while a delayed recruiter response destroys trust; a form can be accessible while the actual job proves inconsistent with its description.

Trust itself is not a single magic score. A multidimensional model of trust developed for online commerce distinguishes beliefs, expectations and intentions [S15]. It provides an indirect conceptual comparison, not evidence of a recruitment-site treatment effect. You might investigate whether an identifiable contact person and transparent data policy change perceived trust. You would still need a recruitment-specific study to determine whether that change affects informed application behaviour. Use scientific theories to define better questions, not to invent universal conversion multipliers.

The appropriate intervention depends on the established mechanism. Changing an unrelated page can make reports look busy while leaving candidates no better informed.

If a recruiter continues to ignore applicants, a friendlier visual design cannot solve that organisational failure. The promised and actual candidate experience must match.

For the fictional technician role, 'You never face a problem alone' should only be used if the organisation can say who is available for support, during which shifts and with what response expectations. Likewise, 'rapid and personal recruitment' needs an agreed contact process. A claim such as 'excellent progression' should map to real training options rather than employee testimonials chosen only for promotional effect.

Create a simple claim ledger: wording, source, responsible owner, exceptions and review date. Have the operations team verify job realities, recruitment verify the application stages, and communications translate those facts into accessible copy. When there is no reliable evidence, say what is unknown rather than inventing certainty. This may make the site less superficially polished, but the purpose is not visual perfection. It is a credible, understandable decision environment in which interested candidates can decide whether the work and the employer suit them.

Choose the primary measure based on the uncertainty you are trying to address. If the hypothesis concerns shifts, assess accurate understanding of shifts. If the hypothesis concerns process friction, assess whether candidates can complete a representative task without avoidable errors. If the hypothesis concerns employer meaning, examine coded associations and a relevant voluntary next step. Do not select a favourable click metric only after inspecting the data.

Record secondary measures such as accessibility, perceived fairness, appropriate self-selection and later verified application stages. Consider sample size, seasonality, vacancy differences and channel changes before interpreting trends. A mobile campaign in a difficult labour market is not comparable with an earlier desktop campaign merely because both share a job title. Report confidence intervals and uncertainties where appropriate. Avoid deriving a company-wide 'drop-off reduction' from a handful of qualitative interviews or a test whose randomisation was compromised.

A person viewing a vacancy after a late shift may have less time and a different immediate goal than someone comparing employers at home. Some visitors already know employees in the company; others have never heard of it. Some arrive directly from an advert that already included salary or location. An unexplained exit might reflect any of these factors. Recruitment-site research that ignores the wider information journey risks inventing causes for patterns that are only coincidental.

You can capture relevant context with minimal, voluntary questions. Do not silently infer psychological vulnerability from browsing histories, disability status or unrelated personal attributes. If analysis requires group comparisons, specify the legitimate purpose, assess whether the sample supports them and avoid using protected characteristics to engineer exclusion. What matters is the concrete information task and its outcome. Someone can be a good applicant and still dislike a particular advertisement; that is not a stable 'resistant candidate' personality.

Association research can investigate whether an employer is quickly linked with concepts such as 'supportive', 'unpredictable', 'fair' or 'closed'. Such a finding could help define which content should be clarified. Indirect tasks only deserve a role where their measurement properties, stimuli and interpretation are suitable for the actual research question. They do not reveal an individual's mind or predict who will accept an offer. Eyetracking can add information on attention; EEG, under a sound protocol, can capture specific electrical signals. Neither reads motives from a website visit.

For more on associative target groups, employer branding through neuroscience and identity congruence, see the verified Neurofactor knowledge-base articles. Use them as conceptual background while checking the primary scientific methods for any consequential claim. The correct method is the least intrusive one capable of informing the decision, not the method with the most impressive equipment.

First, equating session exits with rejected job offers. Second, treating all candidates as if they had identical information needs. Third, diagnosing a 'negative culture association' from a click map without asking what the person understood. Fourth, improving the Apply button while ignoring inaccessible forms or inaccurate job conditions. Fifth, reporting an attractive case study without explaining who was studied and which outcomes were actually measured.

A sixth mistake is optimising exclusively for application volume. It can reward misleadingly broad promises and place unnecessary effort on candidates who later discover important constraints. A stronger process reports whether people understood the role, whether suitable participants could act, and whether barriers were removed fairly. Not every person needs to apply for the study to count as helpful. At the same time, a drop in starts caused by a technical defect is not successful self-selection. Investigate the mechanism before evaluating the headline rate.

HR and operational managers own the accuracy of job descriptions, training arrangements, shift schedules and day-to-day conditions. Recruitment owns process expectations, candidate communication and the ATS handoff. UX and communications translate verified facts into accessible content, test navigation and avoid unsupported symbolic promises. Analytics and privacy owners define legitimate measurements and safeguards. Neurofactor can support a disciplined association diagnosis when it is connected to actual job evidence and candidate behaviour.

Relevant existing articles explore associations in recruitment, recruitment clicks without applications, job titles that discourage candidates and application anxiety. These are established published CMS records; this new article remains an unpublished draft. The related links provide context, not evidence that any fictional intervention has been implemented.

Pick a job that the organisation genuinely needs to fill, verify its conditions and identify the specific decision candidates struggle to make. Is the issue understanding shift work, judging whether support exists, finding a salary range or navigating the application form? Write down the proposed association and at least one alternative explanation. Choose a measure that can separate them. Create a content variant only after that diagnosis and compare it fairly. A new design should follow evidence, not substitute for it.

Neurofactor's role can be to connect employer associations with observable decisions during the candidate journey. The purpose is not to use neuroscience to make hesitant people submit an application against their better judgement. It is to remove avoidable uncertainty while retaining the information candidates need to decide freely. A strong careers site is one where an appropriately interested person understands the opportunity, sees the limits and can take the next step without artificial obstacles.

Candidates leave careers sites for different reasons. Some learn that the job does not fit, some want to think, some are blocked by a process, and some form avoidably negative assumptions because information is absent or ambiguous. A rigorous response combines association research, realistic job evidence, usability and appropriate behaviour measures. It does not confuse attraction with application, or an application with a successful employment decision. The best improvement is not a page that convinces everyone. It is a truthful and accessible journey that helps suitable candidates proceed and everyone else understand the decision they are making.