A production engineer in the Permian Basin inherits 340 rod-pumped wells. Half were designed by a senior engineer who retired two years ago. The other half came through an acquisition with almost no documentation. The new engineer has a simulation license, a stack of API design manuals, and a set of rules of thumb passed down on index cards. Which approach should govern the next round of rod string redesigns?
This is not a hypothetical. It is the reality facing thousands of engineers across North American operations. And the answer most teams land on - pick one approach and stick with it - is the wrong one. The wells that fail most often are not the ones where engineers chose the wrong tool. They are the ones where engineers never asked whether the tool they were using was appropriate for the problem in front of them.
Our position is straightforward: most teams under-use simulation on the wells that need it most, and over-apply it on wells where a good heuristic would have given the same answer in a fraction of the time. The key is knowing which well is which.
When Rules of Thumb Work Well
Rules of thumb are not guesses. The best ones encode decades of field observation into compact, actionable guidance. The API RP 11L design method for sucker rod pumping systems, first published in 1965 and refined over subsequent editions, is built on the Gibbs wave equation for rod string behavior (Gibbs, 1963). That equation describes how stress waves propagate through a rod string, and the design curves derived from it remain valid for the conditions under which they were developed: vertical or near-vertical wells, single-phase or low-GOR fluids, and conventional rod grades operating within the modified Goodman fatigue envelope.
Consider a straightforward example. A vertical well in the Midcontinent at 4,200 ft measured depth, producing 28 API oil with a GOR of 150 scf/bbl and a fluid level at 3,100 ft. The operator needs to select a rod string for a 1.5-in plunger pumping at 8 SPM with a 74-in stroke. An experienced engineer can pull out the API RP 11L design tables, select a 76-54 taper (76% of 7/8-in rods on top, balance in 3/4-in), verify the peak polished rod load falls within the allowable stress range for Grade D rods, and have a workable design in under 30 minutes. A full wave-equation simulation of this well would yield nearly identical taper percentages and stress values - Gibbs and Neely (1966) demonstrated this convergence for vertical wells operating in steady-state conditions. The simulation would take longer to set up and would not meaningfully change the outcome.
Rules of thumb also earn their keep in rapid field decisions. When a pumper notices a change in dynamometer card shape and needs to decide whether to reduce SPM from 7 to 5 before the next engineering review, they are not going to run a simulation. They are going to apply the guideline that reducing SPM decreases both peak stress and the risk of fluid pound, and they are going to be right. Similarly, the rule that minimum pump submergence should be 200 to 400 ft to avoid gas interference is a reasonable first pass in most conventional reservoirs. It will not be precisely correct for every fluid composition, but it will keep you out of trouble while you gather better data.
The real danger with rules of thumb is not that they are wrong. It is that they walk out the door with the people who understand their boundaries. When a senior engineer retires, their rules - including the unspoken qualifiers about when not to apply them - often retire too. What remains are the rules without the judgment, which is worse than having no rules at all.
When Simulation Is Essential
There is a class of wells where heuristics consistently underperform, and the cost of that underperformance is measured in rod failures, stuck pumps, and unplanned workovers. These are the wells where simulation is not a luxury - it is the minimum standard of care.
Deviated and horizontal wells are the most obvious category. The API RP 11L design curves were developed for vertical geometry. When a well has a kickoff point at 6,000 ft, builds at 8 degrees per 100 ft to horizontal, and lands at 9,800 ft TVD with a total measured depth of 14,500 ft, the rod-tubing contact forces, coulomb friction, and bending stresses in the build section fundamentally change the load distribution. Lea and Minissale (1992) documented rod failures in deviated wells that were designed using vertical-well rules, finding that friction-induced loads at dogleg locations exceeded the modified Goodman limit by 15 to 30%, even when the vertical-equivalent design appeared conservative.
A specific case illustrates the point. An operator in the Delaware Basin had a well at 12,400 ft MD with a 42-degree maximum inclination and two doglegs exceeding 5 degrees per 100 ft. The initial rod string was designed using standard taper tables - an 86-76-64 combination of 1-in, 7/8-in, and 3/4-in Grade D rods. The well experienced three rod parts in the first 14 months, all in the upper dogleg zone near 7,200 ft MD. A wave-equation simulation incorporating the directional survey revealed that cyclic bending stress at the dogleg, combined with the dynamic rod-tubing contact force, pushed the local stress state well outside the Goodman envelope. The redesigned string used higher-grade rods (Grade KD) through the build section, added centralizers at 30 ft spacing through the dogleg zones, and shifted the taper point 400 ft deeper. That string ran 26 months without a failure.
Gas interference presents another domain where simulation outperforms heuristics. When a well produces at a GOR above 500 scf/bbl and the pump intake pressure drops below the bubble point, gas breakout in the pump barrel reduces volumetric efficiency in ways that are highly sensitive to pump speed, plunger clearance, and downhole separator geometry. The interaction between gas volume fraction, valve dynamics, and pressure pulsation in the pump is a coupled, transient problem. Heuristics like "keep pump submergence above 300 ft" or "slow the pump down" point in the right direction, but they cannot tell you whether 5 SPM or 7 SPM is the correct operating point for a specific well's IPR and completion configuration. Simulation can, and the difference between those two setpoints might be 30 barrels per day of incremental production - or an extra year of pump life.
Tapered rod string optimization is a third area where simulation adds clear value. A well producing from 8,000 ft with four possible rod diameters (1-in, 7/8-in, 3/4-in, 5/8-in), three rod grades (D, K, KD), and variable taper percentages has hundreds of feasible combinations. The API tables provide a handful of standard tapers that work well for average conditions, but finding the design that minimizes peak stress across all rod sections while meeting production targets and energy efficiency goals requires parametric analysis. Everitt and Jennings (1992) showed that optimized rod strings designed through simulation achieved 8 to 12% lower peak stresses compared to standard taper designs, directly translating to longer fatigue life and fewer failures per year.
The Dangerous Middle Ground
The failures that cost the most are not the ones caused by choosing the wrong method. They are caused by using the right method carelessly. Two patterns show up repeatedly in post-failure analyses, and both involve a mismatch between the tool's requirements and the inputs it receives.
The first pattern is rules of thumb applied outside their valid envelope. A West Texas operator ran 147 wells using a single SPM guideline derived from their most productive vertical field. When they acquired a package of 60 wells with inclinations ranging from 20 to 55 degrees, they applied the same guideline. Within 18 months, the acquired wells had a rod failure rate 3.2 times higher than the legacy vertical wells. The root cause was not that the SPM was wrong in an absolute sense - it was that the guideline assumed vertical geometry and low friction, neither of which applied to the deviated wells. The friction loads added 8 to 15% to the peak polished rod load compared to what the heuristic predicted, pushing several rod sections past their fatigue limit.
The second pattern is simulation built on unreliable inputs. A model is only as good as the data feeding it, and rod pump simulation is particularly sensitive to three inputs: the directional survey, the fluid level, and the PVT properties. An engineer who runs a wave-equation model with a deviation survey from the original drilling program (which may differ from the actual wellbore by several degrees in build rate), an assumed fluid level based on last quarter's acoustic measurement (which may have changed by 500 ft as the reservoir depleted), and PVT correlations instead of actual lab data will get a result that looks precise and is substantively wrong. The model output will include stress distributions plotted to four significant figures, but those figures will not correspond to what is happening in the wellbore.
The antidote to both patterns is calibration. Simulation results should be compared against measured dynamometer cards, actual polished rod loads, and observed pump fillage before they are trusted for design decisions. Rules of thumb should be checked against the failure history and operating data from the specific field where they will be applied. A heuristic that works in a 3,000-ft vertical field in Oklahoma may not work in an 11,000-ft deviated well in the Midland Basin, even if both produce similar fluids. The physics are different, and the rules need to reflect that.
How to Decide for a Given Well
Rather than defaulting to one approach, we recommend a decision process driven by four factors: well complexity, data quality, consequence of failure, and available engineering time.
Well complexity is the first filter. Vertical or near-vertical wells (less than 15 degrees inclination) at moderate depths (under 6,000 ft) producing low-GOR fluids with no history of rod failures are strong candidates for rules-of-thumb design. The API RP 11L method was built for these wells and performs well within its design envelope. Wells with significant deviation, deep pump settings, high GOR, heavy or viscous fluids, or corrosive environments need simulation. The interaction between these factors creates nonlinear behavior that heuristics cannot capture.
Data quality is the second filter, and it cuts both ways. When you have a good directional survey, a recent fluid level measurement, and reasonable PVT data, simulation is justified because the inputs will produce meaningful outputs. When the data is sparse or stale - a common situation with acquired wells or older assets with minimal surveillance programs - simulation can give you false confidence. In that case, rules of thumb with conservative safety factors may actually be the more honest approach, because they do not pretend to a precision that the data does not support. The key is to be explicit about the uncertainty rather than hiding it behind a model.
Consequence of failure is the third filter. A well producing 5 barrels per day where the cost of a rod failure and workover is $12,000 does not justify a two-day simulation study. A well producing 200 barrels per day where a failure means $45,000 in workover costs plus $30,000 in deferred production absolutely does. This is not a commentary on the value of any individual well, but rather a practical observation about how to allocate engineering time. The wells with the highest failure costs should get the most rigorous analysis.
Available engineering time is the fourth filter, and it is the one most teams are least honest about. If you have 340 wells and two production engineers, you are not going to simulate every well. Nor should you try. The practical approach is to tier your wells: simulate the complex, high-consequence, high-production wells (perhaps the top 15 to 20%); apply validated rules of thumb to the straightforward, lower-risk wells; and use the simulation results from the first group to refine the heuristics you apply to the second group. This creates a feedback loop where field data continuously improves both your models and your rules.
Where PetroBench Fits In
We built PetroBench because we kept seeing the same pattern: operators running wave-equation simulations on wells that did not need them, while neglecting simulation on wells where it would have prevented failures. The problem was not a lack of simulation software. It was the overhead of setting up, running, and interpreting simulations for hundreds of wells, which meant engineers defaulted to heuristics even when those heuristics were inadequate.
PetroBench's simulation engine runs the full wave-equation model, but it is designed for the workflow described above. You can import your well data - including directional surveys, rod string configurations, and fluid properties - and run parametric analyses across multiple design variables without manually rebuilding the model for each scenario. More importantly, the platform stores historical simulation results alongside actual operating data, so you can calibrate your models against real dynamometer cards and failure records. This closes the loop that most standalone simulation tools leave open.
We also encode validated rules of thumb directly into the platform, with explicit boundary conditions for when each rule applies. When a well's parameters fall outside the valid range for a given heuristic - say, inclination exceeds 20 degrees or GOR exceeds the threshold used to develop the rule - the system flags it and suggests simulation. This means the institutional knowledge that would otherwise live in one engineer's head becomes available to the entire team, and it comes with guardrails that prevent misapplication.
Conclusion
The question is not whether simulation is better than rules of thumb. Both are valid engineering tools with well-defined strengths and limitations. The question is whether you are using the right tool for the well in front of you, with data quality that supports your chosen approach, and with calibration that confirms your results match reality.
For vertical, conventional wells with good operating history and low failure rates, rules of thumb backed by API RP 11L design methods remain efficient and reliable. For deviated wells, high-GOR producers, deep pump settings, and any well where failures have been persistent or unexplained, simulation is not optional. It is the difference between engineering a solution and hoping for one.
The best teams we work with do not pick sides. They simulate their most complex wells, extract patterns from those simulations, refine their heuristics based on real field data, and apply those improved rules to the rest of their portfolio. The simulation makes the rules better. The rules make the simulation more focused. That feedback loop - not any single tool - is what separates operators who manage rod failures from operators who prevent them.
References
Gibbs, S.G. (1963). "Predicting the Behavior of Sucker-Rod Pumping Systems." Journal of Petroleum Technology, 15(7), 769-778. SPE-588-PA.
Gibbs, S.G. and Neely, A.B. (1966). "Computer Diagnosis of Down-Hole Conditions in Sucker Rod Pumping Wells." Journal of Petroleum Technology, 18(1), 91-98. SPE-1165-PA.
Lea, J.F. and Minissale, J.D. (1992). "Beam Pumps Surpass ESP Efficiency." Oil and Gas Journal, 90(44). Also referenced in SPE-24784.
Everitt, T.A. and Jennings, J.W. (1992). "An Improved Finite-Difference Calculation of Downhole Dynamometer Cards for Sucker-Rod Pumps." SPE Production Engineering, 7(1), 121-127. SPE-18189-PA.
API Recommended Practice 11L, "Design Calculations for Sucker Rod Pumping Systems (Conventional Units)." American Petroleum Institute, various editions.
