Investment Process Re-engineering: The Motivation

In the preceding article, we presented a conceptual description of investment process re-engineering as a framework that finds its origins and inspiration in the data sciences, bringing tools, processes and ideas from concrete operational applications in the latter field into investment management. Here, we take a step back to try and establish (or rather reiterate) why such re-engineering would be necessary in the first place. To this end, we will attempt to present some of the numerous advantages of this paradigm, particularly relative to the traditional fundamental/discretionary investment management model and allow them to argue for themselves.

?

An operational framework

The key to appreciating the benefits of IPROps is to understand what it seeks to achieve. By creating a standardized framework upon which analysts can give expression to their views through models that emit outputs of a similar nature, a unified interface that facilitates cross-communication and learning across different functional teams within an investment management organization is enabled. Thus, as each team translates its views (encoded as suitable models, which are in turn driven by the relevant input factors) into outcome-distributions over variables of common interest, these outputs can then be incorporated and harmonized, allowing for a unified expression of views on key drivers across the different business disciplines.

On an operational level, this allows managers to streamline and integrate business processes across diverse functions (e.g., investment research, risk, operations, etc.) relying on a common interface. Another application would be the capacity to integrate quantitative and discretionary investment disciplines under a common framework, allowing a quantitative manager to incorporate discretionary views and vice versa.

As highlighted in our earlier articles on the subject, we believe that one area where investment practice will see substantial change is in the way research is conducted, which will be increasingly based on new scientific paradigms that de-emphasize reductive analysis in favor of a synthesis-driven methodology that relies on a framework of adaptive systems, networks, and complex, non-linear relationships, all of which are facilitated by data. A data-driven approach will derive efficiency from focusing research efforts more on verifiable data-based primitives than elaborate pre-conceived financial theories, where data-validated models will then serve to concretize research insights. The centrality of models in IPROps sets it up as a viable tool in this emerging scenario. Indeed, while it may have seemed incredible a few short years ago when the earlier articles were written, today one can certainly see a situation where even a typical analyst will focus on data-mining insights while employing LLMs for the actual production of research reports!

In this regard, one of the key benefits of IPROps is that it enables the easy integration of the investment research and decision-making process into a data-driven, technology-oriented framework that seamlessly fits into the manager's overall digital transformation efforts. In addition, leveraging automation using scheduling and pipelining tools, etc., which are readily available in MLOps, helps facilitate the seamless coordination of various components of the investment decision-making process. MLOps tools also permit model tagging and versioning, not only over multiple generations but also across use cases. Thus, the models themselves can become factors in the decision-making process, which can be responsively depolyed in a rules-based fashion that integrates feedback from the environment, enabling much richer, more complex, and more dynamic execution.

?

The "Fourth Revolution"

From a strategic perspective, this also emerges as an important benefit. As highlighted earlier, IPROps helps managers align with important emerging patterns in business operations, especially with regard to data management. Its adoption provides the platform for managers to incorporate the extensive tooling and methodologies being developed in adjacent disciplines like ML and software engineering, which will certainly become critical in the near future, especially given the rapid evolution of these tools and frameworks over the past few years. Thus, adopting the process re-engineering framework sets the stage for rapid and agile adoption of a set of tools and practices that will doubtless become significant differentiators for organizations going forward.

Furthermore, because the adoption of data-driven business processes can be a fairly long and complicated process, the flexibility of IPROps is an important consideration in its favor in this regard. Since IPROps accommodates a wide variety of modeling approaches, this helps create a ramp for the adoption of increasingly sophisticated modeling practices in the transition to a more complex workflow, as circumstances may demand, all under the same basic framework. This also means that IPROps provides an agile framework for prototyping and experimentation, where a wide variety of old and new data sources and models can be tested on a given problem, and their results can be ranked and winners chosen for deployment, as is the case in MLOps. This is particularly important in the current milieu, where various forms of 'alternative' data sources and greater integration of machine learning have begun to dominate the search for alpha in the investment industry. In addition, the reliance of IPROps on view modeling and the extensive use of documentation and meta-data management that carry over from MLOps serves as a useful tool for accumulating and persisting knowledge within the collective, helping ameliorate the problems of knowledge losses that often accompany personnel turnover.

The immediate benefit of such an approach is that it permits the "industrialization" of investment management and research. The creation of standardized processes that could be coupled into a pipeline and deployed, managed, and monitored in a repeatable fashion is the essence of automation. Notably, models (as well as views) don't always have to be developed in-house but can come from an expanded range of internal and external sources, which significantly increases operational flexibility.

Thus, with properly verified models and well-configured pipelines in hand, the investment process becomes highly efficient as decision-making now flows seamlessly and near-instantaneously from data ingestion to investment execution. Furthermore, it becomes highly scalable as the same repeatable steps can be applied to any number of clients, subject to the limitations imposed by the market itself. It also becomes extensible because, outside of the required primary research, entry into new asset classes, strategies, or markets only involves swapping out models and, perhaps, reconfiguring a few segments of the pipeline, while the overall methodology remains the same. Contrast this to a traditional manager who must either resort to offering collective products and/or expanding analyst teams and portfolio managers to meet the needs of a growing clientele base, among other scalability constraints. Incidentally, the search for new approaches to counter pervasive scalability and cost-efficiency problems, particularly in less developed markets, was the primary motivation for IPROps.

?

Manager-client relationship: A new perspective

Perhaps the biggest benefit of IPROps, however, is that anchoring its processes on data and models makes their conclusions testable. This is one critical feature that is frequently missing in the traditional investment framework, where such critical evaluation, if it exists at all, is mostly done merely in a reporting context. For a traditional manager, the investment decision process is not entirely tangible, and any postmortem of past decisions is typically vague and subjective at best. For models with well-specified inputs, outputs, and assumptions, such evaluations can be concretized against actual outcomes. This allows for a self-correction mechanism via a feedback loop between the environment being modeled and the modeling workflow, allowing the system to make necessary readjustments. This is also an essential element of the systems-based approach highlighted earlier and brings obvious benefits to the quality of investment services by creating a pathway for continuous improvement across the various dimensions of the service platform.

There's a subtle but very significant additional benefit that comes from this. Pooling frameworks are, at heart, tools for scenario analysis. This means that IPROps has an in-built mechanism for scenario modeling, which is important for both risk analysis and strategic planning. An important corollary to this is that it gives managers the agility to proactively monitor key performance metrics under a wide range of model assumptions and possibly pre-configure solutions to be deployed responsively in line with environmental regime changes, as indicated by data. All of this is in addition to its capacity to statistically reflect measures of past performance in modulating decision-making for the uncertain future.

?

The return of active management

In previous articles, we surveyed trends in the investment management industry, tracing out the path to ascendancy of the passive investment management model over the active, along with more recent trends towards a middle ground, driven by new technology and perspectives on the market and technological advancements. We also posited that a data-driven approach to investment management (and financial services in general) would break down some of the primary drawbacks of active management by making cost-effective mass customization of investment management services feasible. IPROps offers one potential materialization of this possibility and also opens up new dimensions to the solution space.

The end-to-end nature of investment workflow execution as a CI/CD pipeline facilitates a new model for active management that we refer to as "inter-active investment management". Under IPROps, managers are not only able to produce a comprehensive model of views that capture their outlook on the market in all its richness, but they are also able—via distributions over variables of interest—to present this to the market in a tangible format. Moreover, with growing convergence in the age of data, APIs, and consumer-service platforms, the scope of possibilities continues to grow exponentially. Thus, in their interactions with clients, managers are able to model—to a reasonable degree—as many aspects of the client's needs and requirements as they have data access. This provides managers with the opportunities and tools to close the feedback loop between the end-user and the investment decision-making process, based on harmonized platforms where client needs and requirements can be injected into the workflow with an unprecedented amount of detail. With outcome distributions as a common interface, this client-side element of the workflow can also be subject to its own learning, modeling, and materialization as part of the decision-making pipeline.

Thus, not only can we now envision a comprehensive suite of utilities and capabilities for managing client needs (one that encompasses, for example, debt levels, spending habits, insurance needs, retirement plans, etc.) in an asset-allocation program optimized on a dynamic but consistent view of ever-changing market variables, IPROps also provides a framework for execution. Furthermore, clients now have the opportunity to interact with the manager's view distributions, e.g., by inputting upcoming decisions (e.g., plans to take a new mortgage or other major expenses) in near real-time and receiving feedback on both the optimality and possible impact on existing portfolio profiles, as well as potentially more beneficial alternatives. This indeed forms a substantial piece of the holy grail of capabilities that we predicted emerging technology would bring to the investment management industry in previous articles.

?

Bare necessities

We conclude with a word on practicalities. How would a manager need to structure a team to deliver IPROps as a viable operating model? At least three distinct skill sets form the foundation for an IPROps team, and these may span different roles or perhaps reside in a mix of multi-talented individuals.

First, of course, are the typical financial and economic analysts, whose quantitative orientation and abilities should match the manager's current capabilities and/or future plans. Their role, as usual, will be to generate theses and ideas about key economic themes and financial variables of interest, preferably going beyond mastering the theories to having considerable expertise in modeling them effectively. They will form the core of the team.

Data scientists on the team will bring the benefits of incorporating both traditional methodologies and the latest developments in data-driven approaches to modeling, where expertise in machine learning as well as classical statistical modeling will likely prove very useful. In particular, data scientists will probably find much ground for frequent collaboration with financial analysts in performing the all-important feature engineering of financial data, which will facilitate, in no small measure, the seamlessness of the CI/CD process as it applies to IPROps.

Data engineers will be required to build efficient data and model pipelines, helping the team keep abreast of the latest ongoing developments in tooling, processes, and best practices for both data and model management. They need not be fully embedded in the team, as decisive contributions from them will likely be somewhat infrequent, but the IPROps team will greatly benefit from regular interactions with the engineers' base team in IT, say.

All these roles will, of course, find much ground to work together to cross-germinate ideas and incorporate insights into the IPROps framework to harness its inherent orientation towards continuous improvement. Noteworthy is that while data-related skills are relatively scarce at the moment, the overall required skill burden for IPROps is fairly moderate and probably exists within the boundaries of many investment management organizations, such that it can probably be activated with relative ease and minor reorganization.

To wrap up, we hope that this presentation clarifies the need and role of process re-engineering in modern investment management and its enormous potential to transform both investment management operations and client experience.

?