Reading position against active compound configuration.

You cannot, today, translate geopolitical compound signals into your specific portfolio implications as a systematic capability. Nobody can. The capability does not exist in the enterprise AI stack, it does not exist inside the data and analytics providers your team uses every morning, it does not exist inside the advisory houses your board reads on plane rides, and it does not exist inside the vertical AI wrappers arriving from every direction. It exists only in your head, informally, ad hoc, in the moments between the compound configuration surfacing on your desk and the executive committee meeting arriving on your calendar, and it exists there because you have not been able to source it anywhere else.

72%Of executives cite geopolitical instability as their top economic risk · McKinsey · March 2026

The vocabulary of the gap is nonetheless already in your boardroom's language, and McKinsey and BCG put it there. McKinsey's March 2026 global executive survey found that 72 percent of executives now cite geopolitical instability as their top economic risk, up 21 percentage points in three months, the steepest single-quarter shift in the survey's history since 2008. BCG named the same shift in December 2025, framing the overarching challenge for boards and CEOs as a global move toward multipolarity that, in their words, "demands that companies reassess everything from where they manufacture and market to how quickly they can adapt to change." Their prescription runs across the same territory in every partner deck since: build geopolitical muscle, develop preemptive risk management capability, move faster than the shift arrives.

21ppSteepest single-quarter shift in the survey's history since 2008

Two years of that vocabulary have landed in every boardroom on the continent, and the capability the vocabulary describes has not landed with it. Bloomberg carries the data your team looks at every morning, FactSet carries the analytics on top of it, and Kensho, RavenPack, and GeoQuant surface the signals as they arrive. None of those surfaces translates the signals into your portfolio's implications, because none is positioned in the stack where the translation would happen. McKinsey and BCG do not compose the translation either. Advisory output at their altitude names the problem and prescribes at category level, and it does not translate the compound configuration into your specific portfolio position at your specific altitude on your specific governance calendar.

The gap is what you carry into every executive committee meeting where the compound has arrived and no incumbent capability composed the translation for you. The gap is what your board is now asking you about explicitly, three questions at a time. What are the implications to your annual plan and your risk tolerance when geopolitical instability is the number one concern in your boardroom and you do not have a translation of how it actually reaches your portfolio? If you hold your current concentration through a reactivation window (whether that concentration sits in your client portfolio, your revenue segments, your account portfolio, your engagement pipeline, or your commitment allocations) what happens, and if you rotate, what happens, and how do the cascading geoeconomic signals translate across both paths in a way you can put in front of the executive committee? How do you actively manage the cascading geopolitical impacts against your specific portfolio, your annual plan, your governance calendar, applying the geopolitical muscle BCG says you need to your actual position rather than to a hypothetical one that lives in a partner deck?

The gap between the vocabulary of the problem and the capability to actually manage the problem is what BearingA has built to close.

The translation composes across five dimensions simultaneously, because that is how you actually experience the exposure inside your function.

The first dimension is configuration as empirical unit. The compound is the object, not the event. The read composes against a configuration that has been resolved to component signals with the mechanism between them made explicit, because it is the mechanism that carries the exposure through to your portfolio, and the mechanism is what determines which specific portfolio positions are actually reached.

The second dimension is portfolio-position altitude. The read composes against your specific portfolio, your specific segments, your specific engagement pipeline, your specific account concentration, and never against a sector aggregate or a peer average. The translation lives inside your own model of your own position or it does not answer the question your executive committee asked, and it does not survive an audit dialogue where the reviewer is going to test what you specifically did with the specific exposure you specifically held.

The third dimension is what the paper's title names as five dimensions of position multiplied. Your position is not a single variable. Your revenue exposure interacts with your cost structure, which interacts with your customer-side or client-side behavioural cadence, which interacts with your governance calendar, which interacts with your annual-planning cycle. The compound configuration reaches each of those dimensions differently, at a different speed, and with different consequence, and the translation must carry all five to be actionable inside the risk-tolerance framework your board has signed off.

The fourth dimension is institutional function altitude. The same compound reads differently to you than it does to your CFO, than it does to your CMO, than it does to your COO, than it does to the auditor sitting on your annual submission, than it does to the executive committee chair you are briefing next Thursday. Same configuration, different altitude, different mechanism reaching each function's decision surface. The substrate composes each read at each altitude specifically, and each read carries the disconfirming signals the reader at that altitude would use to falsify it.

The fifth dimension is cannot-be-wrong as the epistemic floor. Every load-bearing claim traces to primary source at the drill-down, and every claim carries its interpretive geometry as part of the claim itself, naming what the substrate supports definitely, what it supports conditional on what, and what the read explicitly does not warrant. The read cannot manufacture confidence the sources do not support without breaking the discipline your governance function is going to sign off against.

The customer cascade as the piece nobody was translating

The most distinctive move in the translation is not the one your team expects to see, because most sector analysis traces signals through the production cascade, where input costs move first, feedstock repricing lands next, and margins compress at the terminus. The translation your board actually needs traces through the customer cascade instead, where the signal reaches your customers, or your clients, or your portfolio companies' customers, before it reaches costs, and their behavioural response lands on revenue before the input-cost story matters at all.

Three empirical cases show where the customer cascade decided the translation that a production-cascade view would have missed.

In Q4 2022, as Euribor crossed 2 percent, European industrial-automation buyers deferred committed capex within the quarter, and the demand-curve shift landed on their suppliers' order books before feedstock and energy input-cost compression reached those suppliers' COGS. Suppliers tracking only their production cascade read the margin pressure a quarter late, and the customer-side deferral had already repriced the revenue line, so any portfolio holding exposure to those suppliers (as a lender, an equity holder, an account portfolio at an agency serving them, or an engagement pipeline advising them on cost transformation) carried the mispricing until the customer-side signal reached the performance-monitoring surface a quarter after the demand reset had landed.

In 2012 and 2013, corporate travel desks cut booking classes and renegotiated volume agreements as austerity reached expense policy across the region, and the customer-side demand reset landed for European flag carriers before fuel-cost passthrough mattered to carrier unit economics. Airlines modelling only the fuel cascade missed the corporate-account behavioural change that was already moving revenue, and any portfolio holding aviation exposure carried the same delay.

In 2021 and 2022, as TTF gas elevation cascaded through European chemicals, embedded-feedstock cost passed into intermediate contracts before downstream customers could absorb or reprice. Producers reading only their own gas-input cascade missed the point at which their customers' customers deferred, substituted, or reformulated, and the demand-side compression reached volume before contract repricing closed, and any portfolio holding chemicals-sector exposure carried the customer-cascade transmission risk without pricing it, because the transmission was not visible in the input-cost feed.

The structural claim these cases together support, and that every subsequent deployment across the commercial registers has validated, is that for most institutions in most compound configurations the customer-side cascade reaches your portfolio's revenue or cost line before or alongside the production cascade rather than after it. Incumbents at production altitude read one hop of that cascade and incumbents at aggregate-macro altitude read zero hops, while a compound read composed at position altitude reads through as many hops as the configuration transmits through, naming the specific behavioural mechanism at each hop and the disconfirming signals at each link. Your portfolio's revenue and cost consequence sits at the terminus of that cascade, and that is where the translation has to compose to if it is going to answer the question your executive committee asked.

Bloomberg has the data your team looks at every morning, FactSet has the analytics on top of it, Kensho, RavenPack, and GeoQuant produce the signals your dashboards already pull, McKinsey has the discipline your board reads on plane rides, and BCG has the prescription your CEO quotes in the strategy off-site. The question you are left with after mapping that stack is where in it the translation of compound configuration against your portfolio position actually happens.

The answer empirically is that it does not happen anywhere in that stack until the substrate is composed for you at your specific position. The substrate slot for that composition sits at a specific altitude in the enterprise AI stack: below the model layer, above the data-provider layer, distinct from the analytics-on-data layer, orthogonal to the vertical-domain-wrapper layer, and compositional with the consulting-altitude houses that named the problem in the first place. It is a slot that has not been filled at institutional depth because no incumbent is positioned to fill it without leaving the layer their business model actually operates from.

On Claude, MCP, and the enterprise AI stack

Claude sits at the model layer of that stack, and MCP composes the vendor-neutral orchestration interface by which enterprise Claude deployments compose against internal data, and both are essential to how the substrate reaches you in your working environment. Neither of them is the substrate itself.

If you wire an agent to all your internal data (your positions, your annual-planning documents, your board and audit-committee correspondence, your account and client reviews, your engagement records, the whole enterprise metabolism you operate on) you still need the substrate composed against your portfolio because the methodology you would measure against is not in your data. Configuration-as-empirical-unit lives in the corpus of resolved historical precedent your current position is measured against, which is the seventeen prior compound-configuration cycles the substrate has composed against, the four-tier defensibility tagging that qualifies every source at ingestion, and the cross-applicability mapping absorbed on continuous calendar time. MCP is what makes the substrate plumbable into your Claude deployment, and MCP does not produce the substrate.

The low-commitment entry for you at this stage of your evaluation is to invoke read_position via MCP against your own portfolio, revenue segments, engagement pipeline, or client mandate, without commissioning a bespoke composition. The output is a translation at your position at your altitude on your calendar, delivered inside the environment your function already operates in.

The substrate holds a structural position rather than an absolute one, and the paper carries five explicit bounds as intellectual honesty rather than as defensive positioning. Your evaluation criteria are best served by knowing where the substrate does not claim.

The moat is structural rather than permanent. A competitor with capital can approximate the methodology and build a corpus at scale, and what cannot be matched inside an 18-month window is the cycle-compounded substrate that seventeen cycles produced, the filter calibrated against 25 drift-modes, and the Signal Watch ledger that cannot be backdated. The horizon over which the compound advantage holds is bounded at approximately three years rather than being absolute.

The methodology is not patented. It is canonical and openly articulated, and verify.py operates in the public repository, and the moat is at the architectural composition layer rather than at the legal layer. For your evaluation, this means you can inspect the methodology directly rather than taking a vendor claim on trust.

The primary sources are not proprietary. Most are public, including Federal Reserve Bank of St. Louis, MGI, BIS, the Caldara-Iacoviello Geopolitical Risk Index, central-bank lending surveys, IMF, and NGFS, and the moat is what BearingA selected from those sources, tagged at four-tier defensibility, and mapped for cross-applicability. The moat is in the curation rather than in the access, which is a distinction your data-governance function will need to understand before signing off.

The technology stack is not proprietary. HMM regime characterisation is standard practice, Kalman filtering is established practice, and MCP is operable by every enterprise AI product in the stack, so the technology is infrastructure that the whole industry works from. The substrate and the practice are the asset, and the technology is what the asset runs on.

Individual read accuracy is not the commitment the paper claims, and this is the bound most directly relevant to your evaluation. Some reads land as anticipated and some do not, and every read is registered before the outcome window closes and revisited against outcome, publicly, at bearinga.com/signal-watch. The commitment is transparency across the full record rather than accuracy on any individual read, and that distinction is what the pre-registration ledger operationalises, and it is what your executive committee should evaluate the substrate against rather than a hit-rate on selected past reads.