Brinson attribution is most informative when the portfolio and its benchmark share a similar structure. That assumption generally holds for diversified, benchmark-oriented portfolios. It starts to weaken once portfolios become thematic, sector-focused, or otherwise constrained.
In earlier pieces, we covered traditional Brinson attribution and then expanded it with more granular, bottom-up views. This post extends that line of thinking by changing the unit of analysis itself. Instead of working at the sector or regional level, it looks at attribution through GICS industries, and why that shift can materially change how performance is explained.
In essence, this is not a new attribution methodology, the mechanics remain the same. What changes is the level at which active decisions are evaluated.
Sector-level Brinson assumes that sector weights represent meaningful active choices. In many portfolios, that assumption does not hold.
Thematic strategies, sector sleeves, or portfolios built around a narrow investment idea often sit almost entirely inside one or two sectors. In those cases, sector allocation effects are mechanically limited. Most of the attribution output is pushed into selection, even when the underlying decisions were not primarily stock-specific.
From a technical standpoint, the attribution is behaving as designed. From an analytical standpoint, it stops being particularly helpful. It becomes difficult to separate structural positioning from security-level outcomes.
Industry-level Brinson addresses this by moving the control variable down one level, from sectors to GICS industries. Allocation and selection effects are computed using the same framework. The difference is simply the granularity at which active weights are defined.
This matters because many portfolio decisions are expressed at the industry level, even if reporting is not. Managers may rotate between software and IT services, or between medical devices and pharmaceuticals, without changing their sector exposure in any meaningful way. At the sector level, these decisions are largely invisible. At the industry level, they are not.
Two portfolios can share the same sector exposure and still reflect very different views once industry composition is taken into account. Industry-level attribution makes those differences explicit without requiring custom benchmarks or ad hoc adjustments.
Once attribution is segmented at the industry level, allocation effects often reappear in portfolios where they previously looked muted or nonexistent. This is especially true for concentrated strategies that are fully invested within a small number of sectors.
Selection effects also tend to shift. What previously showed up as stock selection frequently resolves into industry allocation combined with more moderate within-industry selection. This does not change the economic outcome, but it does change how performance is interpreted and discussed.
For investment teams, this provides a cleaner link between portfolio construction and attribution results. For risk and oversight functions, it offers a more accurate view of where active risk is actually being taken.
Industry-level Brinson is not intended to replace sector- or region-based attribution across the board. It is better understood as an additional view that becomes useful under specific portfolio structures.
Broad, diversified portfolios may still be best analyzed at the sector or regional level. Thematic or constrained strategies often benefit from industry-level segmentation. The important point is that attribution should be adaptable to the strategy rather than fixed to a single reporting hierarchy.
A modular attribution framework allows these views to coexist and be applied where they are most informative, without forcing a one-size-fits-all approach.
Increasing attribution granularity has traditionally meant higher operational complexity. More dimensions often implied more fragile pipelines, more manual intervention, and slower reporting.
When we implemented industry-level Brinson at Kiski, a key requirement was that it could run alongside standard attribution without introducing additional operational overhead. By keeping the system modular and integrating it directly into dashboards, higher-resolution attribution could be supported with the same reweighting logic and data processes already in place.
This makes higher-resolution attribution practical rather than experimental.
Industry-level attribution does introduce more variability, particularly over short horizons or in smaller industries. Results need to be interpreted in context and often benefit from aggregation or rolling views.
As with any attribution framework, the output is a diagnostic tool. It helps structure discussion and analysis, but it does not replace judgment about portfolio intent or market conditions.
Brinson attribution is a flexible framework, and its usefulness depends heavily on how closely its structure aligns with the way portfolios are actually built.
Industry-level segmentation is one example of how the same underlying methodology can be adapted to different portfolio designs. In our experience, this kind of customization is often less about adding complexity and more about choosing a level of detail that matches the investment process.
For attribution to remain useful, it needs to evolve with portfolio construction. That evolution does not require new theory, just a willingness to adjust the lens.
