the job description is the wrong unit of analysis
job titles average away the variation that matters, and the decisions enterprises now face need a smaller unit than the job.
Most enterprises have a career framework. It usually arrives as a slide with levels down one axis, job families across the other, and arrows pointing up. It tells people where they could end up. It says almost nothing about how to get there from where they stand.
That gap has a measurable cost. Gartner research published in 2022 found that only 25% of employees felt confident about their career at their current organization, and fewer than one in three knew how to progress over the next five years. Three in four employees looking for a new role looked outside the company first. The framework existed. The route did not.
LinkedIn's 2024 Workplace Learning Report shows the same shape from another angle. Only one in five employees had strong confidence in their ability to make an internal move, and just 33% of organizations had an internal mobility program at all.
what the evidence says about moving people sideways
The case for internal movement is not subtle. LinkedIn's 2022 Workplace Learning Report found that companies strong at internal mobility retained employees an average of 5.4 years, against 2.9 years at companies that struggled with it. The Josh Bersin Company's Internal Hiring Factbook, built on five years of hiring data, reports that promoted employees are 70% more likely to stay long term, that 62% of employees making lateral moves show high retention, and that external hiring can cost three to five times more than filling a role internally.
So the return is clear and the adoption is low. The usual explanation is culture. Managers hoard talent and recruiters prefer external candidates. Both are real. Both are incomplete. Managers hoard talent partly because nobody has shown them where else the person could go. Recruiters look outside partly because the internal option is invisible to them.
hierarchy is not adjacency
A framework organized by level answers a ranking question. It tells you that a senior analyst outranks an analyst. It does not tell you that the senior analyst is two skills away from a pricing role and eight skills away from a product role. That second fact is the only one that makes a decision possible.
Labour market research has been clear on this for years. In a 2018 study in Science Advances, Alabdulkareem and colleagues mapped workplace skills as a network and found that workers move between occupations drawing on the same skill set. Skills cluster into two poles, and workers rarely cross between them. Movement follows skill similarity rather than organizational proximity.
A 2021 paper in the Journal of the Royal Society Interface by del Rio-Chanona and colleagues built an occupational mobility network from observed worker transitions. It showed that an occupation's position in that network, not only its automation exposure, drives how long displaced workers stay unemployed. Occupations sitting in sparse regions of the network have few onward moves available, whatever the demand elsewhere in the economy.
Translate that into an enterprise and the conclusion is blunt. If your framework does not encode distance between roles, it cannot tell anyone where to go next. Worse, some of your roles are dead ends and nobody has noticed.
how to build an adjacency map in four weeks
This is a procedure a talent team can run with a spreadsheet and a series of interviews. No new system is required to produce the first version.
step one: pick the roles that carry the headcount
Take the 30 to 50 roles that account for most of your population. Do not start from the full job catalogue. A map of 900 roles is not a map, it is a second catalogue.
step two: describe each role in ten skills, not forty
For each role, list eight to twelve skills with a required proficiency level. The constraint matters more than it looks. A forty-skill profile describes a person's whole working life, and under that treatment every role looks distant from every other role. Ask the question that forces discipline: which skills would make a capable person fail in this role if they lacked them?
step three: measure distance, not seniority
For each pair of roles, count the skills the target role needs that the source role does not already require at that level. Then weight each missing skill by the time it takes to reach working proficiency. A tool takes weeks. A regulatory domain takes months. Judgment built from repeated decisions takes years. The weighted count is the distance between the two roles.
step four: set a walkable threshold
Decide what counts as reachable and write the rule down. A workable version: two roles are adjacent when the gap can be closed within one development cycle, meaning roughly two quarters of deliberate practice plus one stretch assignment. Draw an edge between every pair that meets the threshold. What you have now is a graph, not a ladder.
step five: read the map for dead ends
Check three properties. Reachability asks whether every role can reach a role at the next level within three hops. Dead ends are roles with no outbound edges at all. Bottlenecks are single roles sitting on the only path between two large populations.
Dead ends and bottlenecks are the findings that justify the whole exercise. They are usually specialist roles that the organization considers safe and the people inside them consider a trap. Both groups are right, which is why those roles produce quiet, expensive attrition.
the three tests a path has to pass
Before publishing any path, put it through three checks.
- Named next step. The path states one specific role, not a family and not a level.
- Countable gap. The path lists the skills to acquire, with a number attached. "Develop commercial acumen" is not a gap. "Reach working proficiency in contract structuring and margin analysis" is a gap.
- Existence proof. At least one person has made that move in the last three years. If nobody has, the path is a hypothesis and should be labelled as one.
Most published career paths fail the third test. That is the main reason employees stop believing them, and belief is the whole asset.
who owns the map
Two ownership decisions determine whether the map survives contact with the business. Line leaders own role profiles, because they are the only people who know what a role requires this quarter. A central team owns the distance rule and the threshold, because consistency is the point. If every function invents its own definition of adjacency, the map stops being comparable and quietly reverts to a ladder.
Refresh cadence matters more than the initial build. Role requirements drift, and profiles left untouched for two years imply distances that are fiction. People then walk paths that no longer lead where the map says they lead. Building the map once is a project. Keeping it accurate is an operating discipline, and that is where skill intelligence tooling starts to earn its cost, because inferring skill profiles from work evidence is faster than re-interviewing four hundred managers every year.
The deeper change is what the artifact represents. A ladder is a picture of the org chart. A map is a picture of the workforce. One tells people how they rank. The other tells them where they can go on Monday.