Fictional example. The organisation, its figures and its people are invented to demonstrate reasoning. All numerical values are illustrative only and are not findings from real participants, organisations or datasets.

Case brief:A 420 person logistics business is planning to introduce automated stock picking at its main DC in 18 months time. The exec team have asked the people function to approve 60 further hires to “support the automation programme”. The people function are of the view that the bottleneck is not necessarily headcount. The existing records suggest that the experienced warehouse people have a huge amount of process knowledge, but very few of them have the basic data, fault diagnosis and maintenance skills to go with their operational expertise. The annual turnover in the warehouse is 22% and the two existing maintenance staff are both within 5 years of retirement.

Question. To what extent could an evidence-based, skills-based approach to workforce planning be employed in order to assess the executive team’s request for 60 additional hires? What would be a defensible workforce plan for the next 18 months? What is the role and the limitations of people analytics in this context and what are the associated ethical and equality risks?

What the examiner is looking for is a strong answer that does four things: treats the request as a hypothesis to be tested rather than an instruction to be carried out; applies a recognised workforce planning framework to the scenario rather than just describing the framework in general terms; weighs the evidence (including the quality of the evidence and the gaps in it) before making a recommendation; and demonstrates professional judgement in relation to fairness, data protection and the limits of what workforce data can tell you.

Model answer

Introduction

Requests for additional headcount are by far the most common decision that gets to people function and one of the easiest to get wrong. A request for 60 people is a proposed solution, it’s not yet a statement of the problem. The executive team’s real concern is whether the business will be able to run an automated picking operation safely and effectively in 18 months. Headcount is one way of addressing that concern, it’s not the only way and it may not be the best way.

This answer is that the request should be tested via a skills based workforce planning approach rooted in evidence based practice. The relevant frameworks are set out first: evidence based management, the CIPD’s approach to workforce planning and the human resource architecture approach to how an organisation should source its various capabilities.

The application of the frameworks to the scenario is then set out in five stages: clarifying the strategic change, translating that into critical capabilities, mapping the current supply, analysing the gap and then comparing the various sourcing options. A recommended plan is then set out of 18 hires, 24 development places, 10 roles absorbed via job redesign and 8 roles covered temporarily by contractor. The plan is to be reviewed every 6 months. The whole exercise looks at what people analytics can and can’t contribute and the equality and data protection safeguards that a defensible plan will require.

Why the request should be treated as a hypothesis

Evidence based management asks decision makers to draw on four sources of evidence: the best available scientific research, organisational data, the professional judgement of practitioners and the values and concerns of the people affected (Barends, Rousseau and Briner, 2014). The CIPD has taken this as the basis of its guidance on evidence based practice and emphasises that the starting point should be a clear statement of the problem rather than a preferred solution (CIPD, n.d.-a).

Applied in this context it means re-framing the executive request as a testable proposition: the organisation needs 60 further people in order to operate an automated pick system. That proposition is of course predicated on a number of assumptions which may or may not be valid.

  1. It assumes the constraint is one of volume of labour rather than capability of labour.
  2. It assumes that we can source the right people externally in the time available.
  3. It assumes that we can’t develop or re-train the existing workforce.
  4. And it assumes that the job design of the future operation will be essentially the same as it is today, just with more people in it. Each of those assumptions can be tested against the evidence.

Treat the request as a hypothesis in no way amounts to obstruction. It’s a way of protecting the organisation from the cost of making a wrong decision in either direction. Over-hiring leads to fixed costs and the risk of later redundancies once the system has stabilised. Under-investing in capability is the more serious risk as it leads to the situation where you end up with expensive automation that is either sitting idle or running unsafely because no-one can actually troubleshoot or manage the exceptions.

The research into automation bears this out: automation will tend to substitute for the routine and will actually increase the value of the non-routine problem solving, judgement and adaptability that it complements (Autor, 2015). The capability that is most important after automation is therefore likely to be of a different kind rather than simply greater in volume.

The frameworks that structure the analysis

Workforce planning as a staged process. The CIPD describe workforce planning as a fundamental business process whereby the changing needs of the organisation are matched to people strategy (CIPD, n.d.-c). Their specialist knowledge standard defines a six stage process from baseline to delivery, applicable across the tactical, operational and strategic horizons and identifies a range of ‘workforce levers’ to close the gap (build, buy, borrow, bind and bot) (CIPD, n.d.-d).

Gibson (2021) has a very similar six stage model and suggests that workforce planning is most effective as an iterative cycle rather than as an annual exercise. The time horizon is important here in that an 18 month automation programme is right on the edge of the operational and the strategic so the plan has to be designed to deliver the necessary near term readiness whilst allowing for the plan to be adjusted as the technology comes on stream.

Skills rather than jobs as the unit of analysis. Traditional workforce planning looks at posts. A skills based plan looks at capabilities. It acknowledges that a job may require a number of capabilities and that an individual may possess capabilities that are not reflected in their current job title. The case for moving to a skills based approach is further underpinned by employer expectations of ongoing skills disruption.

The most recent WEF survey found that employers expect 39% of their workers’ core skills to change by 2030 and that 85% of them plan to make upskilling a priority (WEF 2025). These are employer expectations of the future rather than actual measured outcomes and should not be interpreted as a prediction for this organisation. They do however provide a good reason why the skills approach has become the norm.

Matching sourcing to the strategic value and uniqueness of capability. Lepak and Snell (1999) in their HR architecture work argue that you shouldn’t source all capability the same way. Capabilities that are both strategically valuable and unique to the firm are best developed in-house as they are hard to buy and provide a sustainable advantage.

Capabilities that are valuable but generic can be sourced from the labour market. Capabilities of lower strategic value or ones that are only needed for a short period of time are the ones that are good candidates for contracting or alliances. That gives the build, buy, borrow approach a sound basis. Cappelli (2008) throws in a practical caveat that long term forecasts of talent demand are often wrong so you should go for options that can be flexed rather than commit everything to one forecast. Shorter development pipelines and a mix of in-house and external sourcing are the kind of options that make sense.

Decision science rather than reporting. Boudreau and Ramstad (2007) contend that decisions about people should be focused on the ‘pivotal’ roles and capabilities, i.e. those where a change in quality or quantity will have the biggest impact on strategic outcomes. In an automated warehouse the pivotal roles are not going to be the general picking roles which automation will have partially replaced. They are more likely to be the roles that keep the system going and the roles that bring it back when it breaks.

Stage 1: Clarify the strategic change

The first thing to do is to set out what the change to the operation will be rather than simply state that automation is coming. In the example scenario the workshops with the ops leads and the technology supplier have established four facts.

  • Firstly that automated picking will take around 70% of the order lines once stable with manual picking remaining for the large and fragile.
  • Secondly that the system will require ongoing monitoring, exception handling (e.g. mis-picks, blocked totes) and first line fault diagnosis on every shift.
  • Thirdly that the supplier will provide second line maintenance for the first year only.
  • And fourthly that there will be a commissioning period of about four months during which the old and new systems will be running in parallel and therefore there will be a temporary increase in labour requirements.

This last point is important. Some of the 60 person request from the exec team is likely to be a genuine but temporary spike in demand during the parallel running period. Permanent hiring is a very poor way to respond to a temporary spike.

Stage 2: Translate the change into critical capabilities

The next stage is to move from the operational change to the capabilities that it will require. Working with the operations managers we identify six capability areas and define each one in observable terms so that they can be assessed rather than assumed:

• System operation and monitoring: running the system, looking at the dashboards and dealing with alerts within the prescribed tolerances.

• Exception handling: dealing with mis-picks, totes that are blocked and inventory variances without any interruption to the line.

• First line fault diagnosis and preventative maintenance: diagnosing the common faults, carrying out the routine maintenance and correctly escalating.

• Data literacy: the ability to interpret the throughput and error data and use that to improve shift performance.

• Process knowledge: an understanding of product handling, customer requirements and safety practice as per the existing operation.

• Leading teams through change: managing mixed human/automation teams, maintaining safety and supporting our colleagues whose jobs are being changed.

Defining them in that way prevents the very common mistake of treating ‘digital skills’ as a single entity. Data literacy and fault diagnosis are two very different capabilities with very different development paths and lead times.

Stage 3: Map current supply

The org then has to establish what capability it actually has. That’s where people analytics comes in and where its limitations are first apparent. The records of training, certification and role history can provide a partial view. In the fictional data we have 40 operatives with forklift and safety certs, 12 with basic data handling courses and two technicians with electrical maintenance certs. Process knowledge is very extensive in fact with the median length of service of warehouse staff being seven years.

However the records will only show what people have been trained in, not what they can actually do. All sorts of skills that are relevant to automation (e.g. informal troubleshooting, experience of other technologies outside of work) are simply not visible in the HR records. A credible skills map therefore requires a short, structured skills assessment which combines the individual’s self assessment with manager validation and, where appropriate, a practical test for technical skills.

In practical terms, the assessment could be designed as a three-stage process: first, a short survey for all warehouse staff asking them to rate their confidence and recent experience across the six capability areas defined earlier; second, a structured manager review to validate or moderate self-assessments based on observed behaviours and informal feedback; and third, a simple hands-on practical test in areas such as first line fault diagnosis to identify actual competence.

For example, staff could be asked to resolve simulated faults on a training system under observation. The results of all three components would then be combined into an individual capability profile. As this will be the collection of new personal data it will need a clear purpose, transparency around how the results will be used and a DPIA where the processing is likely to result in a high risk to the individual (UK GDPR Art 35). People need to know whether the assessment will be used as part of the selection process for a new role as that has a direct bearing on both fairness and the honesty of the responses.

31 operatives are judged to have a high level of aptitude for system operation and exception handling and 9 are judged to have aptitude for fault diagnosis with further training. The supervisors’ strength is in process knowledge and safety with less confidence in data use.

Stage 4: Analyse the gap by capability and time horizon

Comparing the required capability against the current capability produces a gap profile rather than a headcount figure. Table 1 below is the (fictional) result of that.

Capability areaCurrent strength (1–5)Required strength (1–5)UrgencyMain gap
System operation and monitoring24HighVolume and skill
Exception handling24HighSkill
Fault diagnosis and maintenance14CriticalSkill and succession
Data literacy23MediumSkill
Process knowledge44LowRetention risk
Leading teams through change24HighSkill

Table 1. Illustrative capability gap profile. The scores are the planning team’s structured judgements, not psychometric scores.

Three conclusions can be drawn. Firstly, the most significant gap is in the area of maintenance and fault diagnosis. The organisation is suffering from a skills shortage in this area and is also at risk of succession as the two technicians currently in post approach retirement.

This is the key capability in Boudreau and Ramstad’s (2007) sense of the word: if it fails the whole automated system comes to a halt. Secondly, there are genuine gaps in the areas of system operation and exception handling but these are largely addressable through the development of the existing workforce as the assessment found the current operatives to be very capable. Thirdly, process knowledge is a strength but is at risk. A plan which involved the recruitment of large numbers of external staff whilst leaving the existing workforce in some degree of uncertainty about their future could well lead to the very type of turnover that would destroy the process knowledge.

The time horizon really tightens this down. System operator development can be done in 3-6 months; maintenance technician development takes 12-18 months which is the entire window. That difference in lead time is probably the most compelling argument for a mixed sourcing approach.

Stage 5: Compare sourcing options

The planning team then go on to compare the options for each gap. Table 2 sets out the fictional appraisal.

OptionBest suited toSpeedCost profileMain risks
Internal development (build)System operation, exception handling, data literacy, supervisor capabilityMediumModerate, front-loadedTraining capacity; backfill during training
Targeted recruitment (buy)Experienced maintenance technicians; one automation leadSlow for scarce skillsHigh per hire; ongoingTight labour market; integration
Contractors (borrow)Parallel-running peak; interim maintenance coverFastHigh but temporaryKnowledge leaving with contractors
Job redesignRoles where automation removes routine picking tasksMediumLow direct costConsultation; perceived downgrading

Table 2. Illustrative sourcing appraisal.

The appraisal is based on Lepak and Snell (1999). Exception handling and system operation in this instance are valuable and fairly site specific so it makes sense to build them in-house and it also protects the process knowledge. The experienced maintenance staff are valuable but fairly generic so some can be hired in. The parallel running peak is a one-off so it should be borrowed rather than purchased. Job redesign deals with the fact that automation takes the routine picking out of some jobs; those jobs can be redesigned to take on the monitoring and exception handling rather than just leave the jobs as they are and add on new posts.

The recommended plan

On the fictional evidence the 60 requested posts are as follows:

ElementNumberRationale
External hires18Four experienced maintenance technicians, one automation lead, and 13 operatives to replace expected turnover in roles not covered by redesign
Internal development places24Fifteen operatives into system operation and exception handling; six into a maintenance apprenticeship route; three supervisors into a change-leadership programme
Roles absorbed through job redesign10Picking roles reshaped to include monitoring and exception tasks
Roles covered by contractors8Parallel-running peak and interim maintenance cover for months 1–9
Total60 

Table 3. Illustrative reallocation of the original request.

The bridge between the executive request and the recommendation is critical for credibility in that all 60 of the requested posts are there. The plan doesn’t say the executive team were wrong to have identified a need at all; it simply says that the need is better met by a different mix of levers.

On the evidence the 60 requested posts are as follows:

The plan is supported by a six month capability review. At each of the reviews the planning team will be checking that development is on track, that contractor reliance is indeed declining as planned and that the assumptions of Stage 1 still stand.

Clear triggers for review are set in advance: for example, if fewer than 80% of staff in development pathways reach the required capability milestones on time, if turnover in critical roles rises significantly above historic levels, if operational indicators such as system uptime fall short of targets for two consecutive months, or if the business context or technology implementation timeline changes. These indicators ensure that the plan remains adaptive and that any emerging risks or deviations are addressed promptly. Gibson’s (2021) iterative model and Cappelli’s (2008) argument for adjustable commitments both underpin this: the plan is a set of decisions which can be amended, it is not a fixed forecast.

Measuring whether the plan is working

A common failing with workforce plans is that they measure activity rather than outcomes. A dashboard which simply shows training completions would tell you very little about whether the organisation can actually run the new system. Rasmussen and Ulrich (2015) go on to say that HR analytics only avoids becoming yet another management fad when it starts with business questions and links the people metrics to the actual business outcomes.

The plan uses four groups of indicators.

Capability indicators are the number of staff certified as competent in each capability area (certification is by practical demonstration not by course attendance). Deployment indicators are the number of shifts which have the requisite mix of certified operators and fault diagnosis cover.

Operational indicators are system uptime, mean time to recover from faults and pick error rates.

Retention indicators are the rates of turnover in the critical roles and in the experienced staff. The link between the people and the operational measures has to be interpreted with care as improved uptime for example could be due to a supplier software update rather than anything to do with the workforce capability so alternative explanations must be considered before any conclusion is drawn as to the effect of the plan.

The role and limits of people analytics

People analytics makes three real contributions to the decision. It provides the factual baseline (headcount, turnover, tenure, certification) that underpins the discussion in a data rather than an impression driven way. It enables scenario modelling (e.g. what would the number of hires be if turnover was at X, Y or Z). It also enables the plan to be monitored against the indicators.

Its limitations are just as important. Marler and Boudreau (2017) review found the evidence base on HR analytics to be very limited and that organisations often simply don’t have the data quality or analytical capability to use it effectively. Data quality, in particular, can be improved through regular data audits, robust validation processes, and training staff on accurate data entry and maintenance. These steps help ensure that workforce data is both comprehensive and reliable for decision making. Three of those limitations are directly relevant here.

Firstly the data are a record of the past not a predictor of the future. A 22% turnover in a manual job doesn’t necessarily tell you anything about future turnover once the jobs have been redesigned. The redesign might improve retention by making the work more interesting or it might have the opposite effect if the redesign is perceived by staff as a loss of control. The model therefore needs to look at a range of turnover scenarios rather than simply project one figure forward.

Secondly the records are incomplete and may reflect historic bias. The CIPD point out that if the data is incomplete or biased then the analysis will simply reproduce that bias (CIPD, n.d.-b). In the scenario described, access to training may have been at the discretion of the manager or may have been contingent on shift pattern.

If part-time staff (who are disproportionately likely to have caring responsibilities) were less likely to be offered training then a supply map based on training records would understate their capability and could in effect be directing development places to full-time staff. A plan which simply accepted such data at face value could be amounting to indirect discrimination as defined by the Equality Act 2010 (s.19) (i.e. the application of a provision, criterion or practice which puts people who share a protected characteristic at a particular disadvantage without any objective justification).

Thirdly, numbers can’t answer questions of value. You can use analytics to work out the cost of different options, but you can’t use them to work out how much importance to attach to things like job security for long serving staff or to the organisation’s reputation as an employer. Those are judgement calls for the leader and are precisely the type of judgement that evidence based practice recognises as one of its four sources of evidence (Barends, Rousseau and Briner, 2014).

Ethical and equality safeguards

A defensible design incorporates the safeguards into the design itself rather than adding them on at the end.

Equality impact assessment. Prior to the allocation of posts and the creation of new roles, the organisation should carry out an equality impact assessment to ensure that the selection criteria don’t have a disproportionate impact on any particular group. Practical tests should be available and reasonable adjustments should be available to disabled candidates. The development should be scheduled to allow for part-time and shift workers to take part.

Transparent selection for development. The criteria for selecting the 24 development places should be published in advance (based on the capability definitions from Stage 2) and be open to application rather than simply allocated via manager nomination. That in turn reduces the risk that it’s the result of informal networks.

For example, an open call for development opportunities could be advertised to all eligible staff via email and noticeboards, with a published rubric outlining how applications will be assessed against specific capability definitions. Applicants would be invited to submit short statements describing their experience and motivation, and these would then be scored transparently by a panel using the published criteria. Applicants would receive feedback on their applications, and all selection outcomes would be communicated openly to ensure the process is perceived as fair and transparent.

Consultation on job redesign. Redesigning jobs will change people’s work and may change their terms. Good practice is to consult early and genuinely and where there are changes to contractual terms this is a legal requirement. Consultation also improves the design as the people who have the experience of the job will often be in a position to tell you which of the tasks the technology is going to struggle with.

Data protection. The skills assessment should collect only the data necessary for the purpose for which it’s being used, store it securely and not be used for any other purpose (e.g. performance management) without the employee being made aware of that.

Honesty about uncertainty. The staff need to be told what the plan is, what it is based on and when it will be reviewed. If you overstate the level of certainty about the future of the role then you’re going to damage the trust you have with the team if the plan subsequently changes.

Counter-arguments and their limits

A reasonable argument can be made for the original proposal (i.e. the 60 person hire). The executive team could well point to the fact that it’s a much simpler proposition, easier to communicate and less reliant on the organisation’s in-house training capacity. There’s some merit in that argument, particularly if the organisation has no prior history of delivering such a large-scale development programme. The entire basis of the plan is its reliance on internal development and that is a real risk. The six month review should therefore contain a trigger: if development is not progressing as expected then further recruitment can be released.

It could equally be argued that the capability scores in Table 1 are entirely subjective. They are. They are essentially structured judgements and the text should make that clear. In practice, structured judgements involve a panel of assessors who each score the capabilities independently based on defined criteria and supporting evidence.

The panel then meets to review and discuss any significant differences in scores, aiming to reach consensus where possible or to document the rationale for any remaining disagreements. This process ensures that the scores are not simply a matter of individual opinion but result from a transparent, systematic evaluation. Their value lies in the transparency of the assumptions and process, not in any degree of precision. A planning team that records who scored which capability, on what evidence and where there was disagreement is far more credible than one that just produces a single confident number.

Conclusion

The exec’s request for 60 hires was clearly identifying a genuine need, just that they’d jumped to a solution before the problem was actually defined. A skills based, evidence based approach to the issue turns the question on its head from “how many people do we need?” to “what capability do we need, by when and how do we go about sourcing each element of that?” In the context of the example it makes very clear that the real constraint is on maintenance and fault diagnosis capability.

Operational capability is very much something that can be built in house. Part of the requirement is clearly a temporary peak during the parallel running phase. The proposed plan of 18 hires, 24 development posts, 10 redesigned roles and 8 contractor roles takes account of every single post requested and does so at a lower fixed cost and importantly protects the process knowledge on which the new operation is predicated.

People analytics underpins this decision with data and enables scenarios to be tested, but it can’t eliminate uncertainty and if the data it’s based on is not properly interrogated it can simply reiterate the historic bias of the organisation. The plan is therefore presented as a series of reviewable decisions with clear assumptions, equality safeguards and a review cycle rather than as a forecast. That is at the heart of professional judgement in workforce planning: being clear about what the evidence does and doesn’t support and what the organisation will do if its assumptions turn out to be wrong.

Common weaknesses

• Just describing the frameworks is of little value. You don’t earn any marks for listing the six stages of workforce planning unless each of those stages actually changes your analysis of the situation.

• Accepting or rejecting the request outright. The best answers reframe the request and show how the evidence leads to a different set of levers.

• Treat analytics as neutral. The answers should explain how the supply map could be skewed by incomplete or biased records, with a reference to equality law.

• Measuring activity rather than capability. Training completion is an input not evidence of capability.

• Ignoring uncertainty. A plan without review points or triggers for change is less defensible than one that admits what it does not know.

References

Autor, D.H. (2015) ‘Why are there still so many jobs? The history and future of workplace automation’, Journal of Economic Perspectives, 29(3), pp. 3–30.

Barends, E., Rousseau, D.M. and Briner, R.B. (2014) Evidence-based management: the basic principles. Amsterdam: Center for Evidence-Based Management.

Boudreau, J.W. and Ramstad, P.M. (2007) Beyond HR: the new science of human capital. Boston, MA: Harvard Business School Press.

Cappelli, P. (2008) Talent on demand: managing talent in an age of uncertainty. Boston, MA: Harvard Business Press.

CIPD (n.d.-a) ‘Evidence-based practice for effective decision-making’. Available at: https://www.cipd.org/en/knowledge/factsheets/evidence-based-practice-factsheet/ (Accessed: 13 September 2026).

CIPD (n.d.-b) ‘People analytics’. Available at: https://www.cipd.org/en/knowledge/factsheets/analytics-factsheet/ (Accessed: 13 September 2026).

CIPD (n.d.-c) ‘Workforce planning’. Available at: https://www.cipd.org/en/knowledge/factsheets/workforce-planning-factsheet/ (Accessed: 18 November 2026).

CIPD (n.d.-d) ‘Workforce planning: specialist knowledge’. CIPD Profession Map. Available at: https://www.cipd.org/uk/the-people-profession/the-profession-map/explore-the-profession-map/specialist-knowledge/workforce-planning/ (Accessed: 18 November 2026).

Equality Act 2010, c. 15. London: The Stationery Office.

Gibson, A. (2021) Agile workforce planning: how to align people with organizational strategy for improved performance. London: Kogan Page.

Lepak, D.P. and Snell, S.A. (1999) ‘The human resource architecture: toward a theory of human capital allocation and development’, Academy of Management Review, 24(1), pp. 31–48.

Marler, J.H. and Boudreau, J.W. (2017) ‘An evidence-based review of HR analytics’, International Journal of Human Resource Management, 28(1), pp. 3–26.

Rasmussen, T. and Ulrich, D. (2015) ‘Learning from practice: how HR analytics avoids being a management fad’, Organizational Dynamics, 44(3), pp. 236–242.

UK General Data Protection Regulation (UK GDPR), Article 35. Retained Regulation (EU) 2016/679, as amended by the Data Protection, Privacy and Electronic Communications (Amendments etc) (EU Exit) Regulations 2019.

World Economic Forum (2025) The future of jobs report 2025. Geneva: World Economic Forum. Available at: https://www.weforum.org/publications/the-future-of-jobs-report-2025/ (Accessed: 18 November 2026).