Consider two people with the same job title at a large manufacturing company: Senior Systems Engineer. One spent the past two years building data pipeline infrastructure for a factory automation rollout. The other spent those same two years building localization tooling for a product line that ships to seven Asian markets. They hold the same title. They report into the same job family in the HRIS. Their compensation bands may be identical. But from an expertise standpoint, they are doing almost entirely different work, and the knowledge they carry is relevant to almost entirely different problems.
When an HR team or a project lead looks at the employee directory, they see two Senior Systems Engineers. If they need someone with data pipeline experience, they have a one in two chance of calling the right person. If they need someone with localization expertise, the same. But they have no way of knowing which is which without a separate conversation, and in a company with dozens of people holding the same title, the probability of a random contact yielding the right person drops quickly.
This is the job title problem: the titles organizations use for HR administration purposes are not designed to encode expertise, and they perform that function poorly. The gap is not a HR data quality problem. It is a structural problem with how job titles work.
Why titles exist and what they are actually for
Job titles serve three organizational functions. They establish compensation band anchors for HR administration. They signal seniority and authority in external-facing contexts (business cards, client emails, proposal documents). And they define approximate hiring criteria when a position opens up. None of these functions requires the title to accurately represent what a person actually knows or what they have actually built.
A title like "Senior Business Analyst" communicates: this person has been with us long enough and performed well enough to receive a senior designation, and their work involves some form of analysis in a business context. It does not communicate: this person spent four years working on supply chain demand modeling, developed deep familiarity with a specific ERP customization, and was one of three people in the company who understood how the legacy procurement system handled multi-currency reconciliation.
The specific knowledge is where all the operational value is. The title is a label that approximates a compensation grade and a functional category. Treating titles as expertise indicators is a category error that large organizations make constantly because titles are the data they have.
The HRIS as an incomplete picture of expertise
Most large organizations in Japan maintain some form of skills data in their HRIS. This typically includes a role history (the positions a person has held and for how long), a skills self-assessment conducted at hire or at the last HR review cycle, and training completion records. In more sophisticated organizations, there may be a competency framework that maps assessed proficiency levels to specific capabilities.
Each of these has a characteristic failure mode. Role history captures the administrative moves a person made, not what they learned in each role. A three-year stint as a project manager could have involved deep technical problem-solving or primarily scheduling and stakeholder communication. The history does not distinguish between these. Skills self-assessments are a snapshot that reflects how a person rates themselves at the time of the assessment, against a taxonomy that was designed for HR administration rather than for locating specific expertise. People rate their capabilities relative to peers, relative to their own expectations, and relative to what they think the organization is looking for. The variance in how people interpret rating scales across a large organization is significant. Training records tell you which courses a person completed, not what they retained or applied.
The result is that the HR system contains a lot of data about people and relatively little reliable data about what those people actually know. This is not a criticism of HR teams. It reflects a genuine problem with collecting expertise data at scale: the data that is easy to collect (titles, training completions, self-ratings) is not the data that best represents expertise.
What actually predicts expertise
The most reliable signals of what a person knows come from the work they have done, not from the descriptions of the work they are supposed to do. Contribution history, problem-solving traces, project outcomes, and peer engagement patterns all carry more specific and more current expertise signal than any formal HR data source.
Consider the difference between these two data points about the same person. The HRIS record says: "Project Manager, Digital Transformation Initiative, 2022-2024." The knowledge graph record says: this person authored twelve documents in the internal knowledge base relating to legacy system migration, was referenced in six post-project retrospectives as a source of guidance on data reconciliation, and has been tagged as a point of contact in seventeen internal messages about ERP integration edge cases in the past eight months. The second picture tells you exactly what expertise this person holds and how actively they are being called on to apply it.
This kind of expertise signal exists in most large organizations. It is distributed across project management tools, collaboration platforms, document repositories, and communication channels. The problem is that no one has assembled it into a searchable picture of who knows what. The data exists. The synthesis does not.
The implication for talent decisions
When HR teams make rotation decisions, succession plans, and project staffing recommendations based primarily on title and hierarchy data, they are working from a picture of the organization that misses most of what makes one person's expertise different from another's. This produces two kinds of error that compound over time.
The first is the expertise mismatch error: a person is assigned to a role because their title and seniority level match the requirement, but their actual expertise history does not match what the role actually needs. This tends to surface slowly, through under-performance or through the person needing more ramp time than expected, and the connection back to the initial expertise mismatch is rarely drawn explicitly.
The second is the invisible expert error: the person whose expertise best matches a specific need is not surfaced at all because their title category does not match the search criteria, or because their expertise was developed in a part of the organization that the person making the assignment does not have a relationship with. This person exists. They would have been the better choice. The organization simply had no way to find them.
We are not saying that titles are useless. They serve their intended purpose in HR administration. The point is narrower: organizations that rely on titles and formal role data as their primary expertise indicator are accepting a systematic error in their talent decisions that becomes more costly as organizations grow and as the distance between formal role descriptions and actual knowledge distribution increases. The fix is not to redesign job titles. It is to build a separate layer of expertise representation that captures what people actually know, and to use that layer alongside the formal HR record rather than treating the formal record as a complete picture of organizational knowledge.
