Metric Ambidexterity Is Why Some Analytics Teams Deliver Value and Others Do Not

Two organizations buy the same analytics platform. Same vendor, same data warehouse, same consulting engagement, same number of licenses. One starts making decisions differently within six months. The other gets dashboards nobody reads. I have seen this pattern enough times that the technology explanation stopped making sense years ago. The same technology, the same data, and wildly different outcomes. The productivity paradox has been trying to explain this since Brynjolfsson (1993) named it. Why does IT spending not translate into measurable performance gains? Four explanations: mismeasurement, time lags, redistribution, mismanagement. The mismanagement explanation keeps recycling because organizations keep buying technology without building the organizational conditions that convert spending into value. But "mismanagement" tells you what went wrong, not what the ones who got it right actually did differently.

I think the missing piece is about measurement itself. Vijayasarathy and Jetley (2025) introduced a concept they call metric ambidexterity: the ability of an analytics team to simultaneously exploit established metrics and explore new ones. Their paper argues that analytics competence creates IT business value through this dual capacity, and that organizations whose analytics teams can do both, refine the measures they already track while also developing novel measures that capture emerging phenomena, produce more business value than teams that can only do one or the other. I want to be transparent that I do not have this paper in my local study materials, so I am working from its DOI and abstract rather than a full reading. But the theoretical move it makes is specific enough that I can evaluate it against what I know, and it connects to a stream of IS theory that I have been working through all semester.

March (1991) separated organizational learning into two modes. Exploitation refines existing capabilities. It is about efficiency, selection, reliability, and the steady extraction of value from what the organization already knows how to do. Exploration experiments with new possibilities. It is about search, variation, risk-taking, and the uncertain pursuit of what might become valuable. The tension between them is structural. The returns to exploitation are more certain and more immediate. The returns to exploration are distant and uncertain. When organizations have to allocate scarce resources between the two, exploitation wins. I wrote about this in the context of AI adoption, where most organizations are doing pure exploitation, optimizing existing workflows with AI while calling it transformation, and very few are exploring what genuinely new capabilities AI makes possible.

Metric ambidexterity takes the exploration-exploitation tension and applies it to something very specific: the measurement function inside analytics teams. An analytics team that can only exploit metrics is a team that runs the same reports, tracks the same KPIs, and measures the same outcomes quarter after quarter. The organization gets visibility into what it already knows how to see. An analytics team that can only explore metrics is a team constantly prototyping new measures, testing new data sources, and chasing novel signals. The organization gets visibility into emerging phenomena but never stabilizes any measure long enough to act on it systematically. The team that can do both is the one that institutionalizes proven measures while also developing new ones for changing conditions. That is metric ambidexterity.

This is a narrower and more operational construct than the general organizational ambidexterity that the management literature talks about. Tushman and O'Reilly (1996) argued for separating exploration and exploitation into structurally independent units with different cultures and different metrics. At the organizational level, that advice is sound but hard to implement, as I wrote about before. Most organizations never build the structural separation. They run both modes through the same team with the same budget and the same performance reviews. Metric ambidexterity narrows the scope to something an analytics leader can actually work with. The unit of analysis is not the whole organization. It is the measurement function. Can your analytics team stabilize a revenue-per-customer metric that the CFO trusts enough to use in quarterly reporting, while also prototyping a customer-lifetime-value metric that might replace it in two years? That is the ambidexterity question at the level where it can be answered.

The connection to dynamic capabilities is direct. Teece et al. (1997) defined dynamic capabilities as the ability to sense, seize, and transform. Sensing is detecting new opportunities and threats. Seizing is mobilizing resources to capture value. Transforming is reconfiguring the organization around new realities. Torres, Sidorova, and Jones (2018) mapped business intelligence and analytics directly onto this framework. BI&A serves the sensing function: it tells the organization something changed. But sensing without seizing produces nothing. Their empirical work showed that BI&A reaches firm performance through business process change capabilities, not through analytical output by itself. The dashboard that cannot trigger process change is decoration. I wrote about this in the context of the productivity paradox: the dashboard that sits untouched is the paradox in miniature.

Metric ambidexterity sits right at the junction of sensing and seizing. An analytics team that can only exploit established metrics is doing sensing with a fixed aperture. It sees what it has always been configured to see. An analytics team that can explore new metrics is widening the aperture. It is sensing what the organization does not yet know it needs to see. But exploration without exploitation means the new signal never becomes operationalized. It stays in prototype. The metric that might have changed how the organization makes decisions never becomes stable enough to be trusted. Metric ambidexterity is the capacity to sense new signals and seize them by institutionalizing the measures that prove useful, while retiring or deprioritizing the ones that no longer serve.

The reason this matters for the productivity paradox is that it gives a mechanism for the heterogeneity that Brynjolfsson's four explanations leave unexplained. Two organizations with similar IT investments, similar complementary resources, and similar management quality can produce different levels of business value. The productivity paradox asks why IT spending does not always produce value. Metric ambidexterity asks a more specific question: what is the organizational capacity that determines whether an analytics investment translates into better decisions? The testable answer is the capacity to simultaneously exploit and explore measurement.

This also reframes the IT-business alignment problem that Henderson and Venkatraman (1993) named. Alignment is usually framed as whether IT strategy tracks business strategy. But if metric ambidexterity matters, then part of the alignment problem is whether the analytics function can develop measures that match where the business is moving, not just where it is. An analytics team stuck in pure exploitation mode is perfectly aligned with the past. An analytics team in pure exploration mode is constantly prototyping measures that never stabilize enough for the business to use.

Eisenhardt and Martin (2000) made the point that dynamic capabilities are not vague organizational traits. They are specific, identifiable routines. Metric ambidexterity, if the Vijayasarathy and Jetley framework holds up under empirical scrutiny, is exactly that kind of identifiable routine. Does the analytics team have a process for developing new metrics? A process for retiring old ones? A governance mechanism that decides when a prototype metric becomes a production metric? Protected exploration resources that are not consumed by the pressure to deliver next quarter's reports? Those are operational questions, not abstract strategy questions.

The 16 percent digital transformation success rate is partly a measurement success rate. Organizations that fail at transformation are often measuring the wrong things, or measuring the right things in a way that cannot adapt when conditions change. Metric ambidexterity does not solve the transformation problem. But it does explain one of the mechanisms through which transformation succeeds or fails. The organizations that get value from analytics are not the ones with the most expensive platform. They are the ones whose measurement function can operate in both modes at once.