Decline Curve Analysis, Platform Productization, and the Next Layer of Production Forecasting
Published: August 24, 2026
In Spotfire 15.0, Arps decline-curve fitting moved from custom implementation to a native platform capability. Exponential, hyperbolic, and harmonic models can now be applied directly within visual analytics workflows. At the same time, large language models now make generating those same implementations trivial: a reservoir engineer can ask Claude or a comparable tool for Arps fitting code and receive a working script in seconds.
Commoditization from Two Directions
This is not the invention of new petroleum engineering mathematics. The Arps equations have been public for decades. Python implementations have been available on GitHub and in community exchanges for years. Engineers have long embedded decline-curve routines as data functions inside Spotfire and similar tools, and spreadsheet-based Arps fits have been routine for even longer.
The implementation layer of decline-curve analysis has been pushed closer to pure commodity status.
The progression is now clear:
- Known method: the Arps equations and their modified forms.
- Long-available and AI-assisted implementation: spreadsheet templates, GitHub scripts, community data functions, and code generated on demand with a large language model.
- Platform feature: native support that respects the data model, interactive selection, and visualization context of the host environment.
- Question 01 Does it systematically use information beyond the subject well's own production history? A pure historical fit, no matter how carefully executed or how elegantly integrated into a platform, remains limited to what that single well has already produced. Approaches that bring in multi-well patterns, completion design, or reservoir context are operating on a different information base.
- Question 02 How do I tell the difference, in practice, between a forecast that is merely well-fitted and one that is actually better-informed? A well-fitted curve can still be wrong in the ways Arps has always been wrong on edge cases. A better-informed forecast should show greater consistency across assets and analysts, better behavior on short-history wells, and fewer surprises when operating conditions or interference effects appear. The test is not the quality of the residual on the history; it is the quality of the prediction when the history is incomplete or non-representative.
- Question 03 Where does Arps still earn its keep, and where should I treat it only as a downstream continuation of something richer? Pure historical Arps remains useful when the well is firmly in boundary-dominated flow, the operating conditions are stable, and the decision tolerates the known limitations of the method. In early life, in complex interference settings, or when the forecast will drive capital allocation, it is more appropriately treated as a transparent mathematical continuation of a forecast that was already informed by richer information.
Each step removes friction. However, when implementation is no longer scarce, the evaluation standard must shift from “Can the method be run?” to “What information makes the resulting forecast trustworthy?”
Lower Friction Does Not Equal More Reliable Forecasts
The calculation was never the scarce part for competent engineers. What remains scarce is a forecast that is reliable for the decision at hand. Historical Arps has well-documented failure modes—early transient flow, changing operating conditions, well interference, multi-phase effects, short histories, and related edge cases. These are common, not rare. A lower-friction workflow can increase the number of forecasts without increasing their validity. The limitation lies in the information available to a single-well historical fit, not in the equations themselves.
A better-informed forecast is one that systematically uses information beyond the subject well’s own production history—multi-well patterns, completion design, reservoir context—and that shows stronger behavior on short-history wells, greater consistency across assets and analysts, and fewer surprises when operating conditions or interference appear. Historical Arps remains particularly well suited to mature conventional production and other wells exhibiting stable boundary-dominated decline. In those settings, however, the method is also increasingly standardized and therefore especially susceptible to platform commoditization.
Effort spent productizing known equations inside individual workflows has diminishing returns as platforms absorb more of that functionality and AI reduces the remaining implementation burden. Effort spent improving the information content of the forecast continues to compound.
A Practical Distinction
Three questions separate a forecast that is merely well-fitted from one that is better-informed:
These questions can be asked of any workflow, platform feature, vendor model, or script. They focus attention on the information difference rather than the appearance of the fitted curve.
Closing Observation
Native platform support and AI-assisted coding have further commoditized an implementation layer that was already well understood. That is healthy evolution. It does not, however, make Arps more reliable on the many edge cases where it has always struggled. Easier access can simply produce more of those results.
The clarification is useful. It allows the technical community to focus scarce attention on the problems that remain hard - producing better information for the decisions that decline-curve analysis is meant to support - rather than on the ones that platforms and coding tools are already solving.