This is the second post in a series exploring the idea of skills you earn vs. skills you learn. The previous posts can be found in my archives.
In 2013, something interesting was happening inside large enterprises. For the first time, the day-to-day work of knowledge workers was leaving a digital trail.
Not in the transactional systems — the ERPs, the CRMs, the HR platforms — that had been accumulating structured records for decades. Those systems were good at capturing what had happened: a deal closed, a course completed, a performance rating entered. What they couldn’t capture was how it happened. Who collaborated with whom. Who was at the center of the knowledge flows that made a project succeed. Whose judgment people actually trusted when a decision needed to be made, regardless of what the org chart said.
Enterprise social platforms — the early versions of what would become Slack, Teams, and their predecessors — were generating a different class of data entirely. Interaction data. The unstructured residue of people actually working: messages sent, documents shared, questions asked and answered, ideas sparked and built on. For those of us working in analytics at the time, this felt significant. Not because the data was clean or easy to work with — it wasn’t — but because it was, for the first time, a signal that pointed at something real.
I was part of a team at IBM Research exploring what you could learn from this signal. We called it Project Breadcrumb — the idea being that every interaction left a breadcrumb, and that if you could follow enough of them you might start to understand what was actually happening across an enterprise, rather than what the formal systems said was happening. It was noisy, technically hard, and raised uncomfortable questions almost immediately. But it was pointing at something that the structured data couldn’t see.
The problem with how we measure people
The more we worked with this data, the harder it became to ignore a particular pattern. The people who showed up as most valuable in the interaction graph — the ones at the centre of knowledge flows, the connectors, the quiet contributors whose removal would have caused the most disruption — were often not the people who showed up as high performers in the formal HR systems.
This wasn’t surprising, exactly. Performance management systems measure what they can measure: individual output, goal completion, manager ratings. They struggle with collective contribution, with the value of the person who makes everyone around them more effective, with the expertise that gets shared freely rather than hoarded. It’s the point guard who sets up the play so that the team can score. Traditional performance management systems tend to reward the people who are good at being visible — at presenting their achievements upward, at occupying space in meetings, at self-promotion as a professional skill.
At the time I was also interested in what these signals can tell us about a variety of individual properties, gender being one, and this issue had a distinct gender dimension. Research consistently shows that women are more likely to attribute success to collective effort, less likely to self-promote, and more likely to be penalized for the same self-promotional behaviours that are rewarded in men when they do deploy them. But this isn’t only a gender issue. It’s a systematic bias in how we measure professional contribution — one that disadvantages introverts, people from cultures where individual self-promotion is not the norm, and anyone whose most valuable work happens in collaboration rather than in the spotlight.
The enterprise graph made some of this visible. Not perfectly, not without its own methodological problems, but enough to ask the question seriously: what if we measured people by what they actually do, rather than what they say they do?
The Career Dashboard idea
What would it look like to give someone a dashboard of their own behavioral evidence? Not a performance review assembled by a manager, but an analytics-driven picture of their actual contribution: who they’d connected, what knowledge they’d shared, how their networks had grown, what outcomes they’d been part of. A GitHub widget that showed not just that you’d written code, but something about the quality, complexity, and reuse rate of that code. A picture of professional contribution that a CV could never capture.
The idea was that this dashboard would belong to the individual. Private by default. Their view of their own data, before anyone else’s. The question of what to share, with whom, and for what purpose would be theirs to answer — not their employer’s, not a platform’s.
Looking back at this from 2026, the design principles are recognizable. Individual in control. Data serves the person before it serves the institution. Selective disclosure. The employer sees what the employee chooses to share. These are the same principles that underpin self-sovereign identity and the verifiable credential model — arrived at through a completely different route, from a completely different starting point, but converging on the same architecture of trust.
What the graph couldn’t do
The enterprise social graph was powerful within its context. But it had a hard boundary: the enterprise itself. The interaction data lived in systems the employer controlled. The analytics ran on infrastructure the employer provided. And when the employee left — took a new job, changed sector, started something of their own — none of it came with them. It evaporated at the boundary. And even within the enterprise, it wasn’t particularly portable. I couldn’t easily use it to prove my contribution during a performance review or to demonstrate my proven skills to a new client.
This wasn’t an oversight. It was structural. There was no standard mechanism, no portable format, no cryptographic proof that could carry a behavioral signal in a way that another party could verify and trust.
So the Career Dashboard stayed a concept. The insight — that the most valuable evidence of professional capability is behavioral, not declarative, and that it ought to belong to the individual who generated it — remained true. But the infrastructure to make it real didn’t exist.
That gap turned out to matter more than we realized at the time. And closing it required a decade of work in a completely different part of the technology landscape.
More on that in the posts ahead. But first, the question that started to dominate the conversation almost as soon as we began: who does this data actually serve — and what are our obligations to the people we’re analyzing?
Marie Wallace leads the Digital Identity Innovation practice at Accenture. She has been writing about the human side of data at allthingsanalytics.com since 2011. All opinions expressed are her own.
Marie's Ramblings & Ruminations 