Use Order-of-Magnitude Estimation to Address Incomplete Cost and Complexity Data in APM

Overview

APM depends on cost and complexity data to prioritize rationalization, defend investment decisions, and report portfolio health to leadership. In practice, that data is rarely complete. Precise, accounting-grade cost figures do not exist for every application, and true complexity — how intricate an application actually is to change, retire, or migrate — resists simple measurement even when cost data is available. Organizations that wait for complete, precise figures before acting on cost or complexity produce no portfolio analysis at all, because the wait never fully ends. The absence of clear, explicit figures is not a temporary gap to be closed before APM can function — it is a permanent structural feature of the discipline that a mature APM program must be built to work around.

Best Practice

Adopt Order-of-Magnitude (OoM) estimation as the standard response to incomplete cost and complexity data, built from the quantity, type, and size of an application’s related assets rather than waiting for precise figures that may never arrive. Applications with more Integrations, more diverse underlying technologies (e.g., multiple database types in a persistence polyglot), more Software Licenses and Subscriptions, or more complex Lease and hardware dependencies carry more cost and complexity than applications with fewer, simpler related assets — and this relationship holds directionally even without precise dollar figures behind it. Score applications into OoM tiers (e.g., small, medium, large, extra-large) using these related-asset signals, document the reasoning and confidence level behind each tier explicitly, and refine the tiers as better data becomes available. This chapter establishes the general principle; the specific methodology for using Integration and technology-diversity signals to produce OoM scores is addressed in a dedicated chapter on that topic. Where genuinely precise financial figures are available for a specific application, follow the guidance in “Use precise financial figures where available - use orders of magnitude where not” rather than defaulting to related-asset-signal estimation; OoM estimation from related assets is the fallback for applications and decisions where even a financial-figures-based estimate is not available, not a replacement for precision where it exists.

Modern AI tools make this practical at portfolio scale. Rather than manually tallying an application’s Integrations, Databases, Software Licenses, and other related assets one inventory at a time, AI can scour the Enterprise Model — the governed graph of connected inventories — for every relationship a given application participates in, and derive an OoM complexity or cost score automatically. This turns what would otherwise be a labor-intensive manual exercise into a repeatable, portfolio-wide scoring pass that can be refreshed as inventory data changes.

Wherever an Order-of-Magnitude estimate is recorded, capture its confidence level and source alongside the estimate itself — whether it came from a documented related-asset calculation, an AI-generated inference pending validation, or an experienced practitioner’s informed judgment. This provenance record matters as much as the estimate: a low-confidence, unvalidated estimate should carry different weight in a rationalization decision than a well-documented calculation, even when both produce the same OoM tier.

A related discipline is avoiding false precision: presenting an Order-of-Magnitude estimate with more apparent exactness than the underlying data supports undermines the very discipline this chapter establishes. A tier labeled “Large” should be presented as “Large” or as a defensible range (for example, $500K-$1M), not as a single figure like “$742,318” that implies a level of precision the estimation method cannot actually support. False precision creates a false sense of confidence that can lead decision-makers to treat an estimate as more reliable than it is.

Benefit(s)

Order-of-Magnitude estimation gives APM programs a defensible way to act on cost and complexity questions immediately, rather than deferring every prioritization or investment decision until perfect financial data exists — data that, for many applications, will never exist. Related-asset signals are already sitting in governed inventories most APM programs already maintain, so OoM scoring requires no new data collection effort, only a disciplined method for interpreting data already on hand. Using AI to scour the Enterprise Model for these signals makes OoM scoring repeatable and scalable across the full portfolio rather than a one-off manual exercise limited to a handful of applications. And because OoM tiers are explicitly labeled as estimates with a stated confidence level, the organization retains the ability to refine specific applications’ scores as better data becomes available, without having delayed any decision-making in the meantime.