Table of Contents
The Problem with Success Stories
Post hoc ergo propter hoc. After this, therefore because of this. It is one of the oldest identified errors in human reasoning, catalogued by Roman logicians and still committed daily by engineers with PhDs and terabytes of production data. A rooster crows before sunrise. The rooster does not cause sunrise. A conformance or a gel treatment is injected; water cut drops; therefore the treatment worked as designed. Maybe. Or something else entirely was responsible, and the treatment was simply present at the scene. The error is seductive precisely because it feels like evidence. Something happened, then something else happened, and the sequence feels like causation. In everyday life this heuristic is useful: touching a hot stove is reliably followed by pain, and the inference is correct. In reservoir engineering, where dozens of variables change simultaneously, where the subsurface is never directly observable, and where no experiment can be repeated, the same heuristic becomes dangerous.
The history of medicine offers a more instructive parallel. For two centuries, European sailors reliably recovered from scurvy when given citrus fruits, and the treatment was documented, recommended, and occasionally forgotten and rediscovered. Yet physicians consistently misidentified the active mechanism: the acidity of the juice, the freshness of the provisions, the change in diet more broadly, the psychological effect of reaching port. The outcome was correct and reproducible. The causal theory was wrong for two hundred years. This mattered not because the sailors suffered from the misidentification (they recovered regardless) but because the wrong theory led to wrong generalizations. Lemon juice was sometimes substituted by lime juice, then by boiled lime juice, which destroyed the vitamin C, and scurvy returned. A correct result built on an incorrect mechanism is a foundation that fails the moment conditions change.
The oil industry has its own accumulation of such foundations. A treatment is applied, production improves, and success is declared with insufficient scrutiny of why it worked, whether it would have worked for the reasons claimed, or whether something simpler and cheaper would have worked better. In some cases a technology may have worked for entirely different reasons than advertised. In others, what appears to be improvement is simply the absence of a worse counterfactual that was never tested. And sometimes it did work but not as well as a plainer alternative that was never tried. Five examples from Enhanced Oil Recovery: microgels and nanogels used as conformance agents, Colloidal Dispersion Gels, Water- and Nothing-Alternating-Polymer strategies, polymer concentration tapering, and the bullheading of gels into injection wells without zone isolation, each illustrate a different facet of this problem.
Together they point to something deeper: in subsurface operations, where uncertainty is high and experiments cannot be repeated, the gap between “it worked” and “we know why it worked” is treacherous. And the gap between “it worked” and “it was the best available option” is even wider.
Microgels and Nanogels: What Exactly Is Being Injected?
Particles are injected into the reservoir in a compact state, propagate into high-permeability channels, and sometimes swell in response to reservoir temperature, physically blocking those channels and diverting injected water toward unswept zones.
The concept is scientifically elegant and has been piloted in numerous fields worldwide.The claimed mechanism is deep in-depth diversion. But here is a question that does not appear prominently enough in the published literature: what role does the solvent/oil carrier fluid play? These formulations are not pure particles in water. The injection fluid contains a solvent/oil component. Like for dirty water, an oil slug injected into a reservoir that will produce measurable changes in relative permeability quite independently of any plugging effect from the particles themselves. If the water phase permeability in contacted zones decreases because of relative permeability effects rather than physical pore blockage, the production response could look identical.
Carefully designed tracer-based field studies have shown genuine changes in flow patterns following such treatments: new connected volumes appeared, previously dominant channels became secondary, and previously minor pathways became primary. Production data correlated with these tracer results. These are genuinely useful observations. But even such carefully documented field results do not cleanly separate the plugging mechanism of the particles from the fluid-chemistry effects of the carrier. Both would produce the same signature in production data: reduced water cut, improved sweep. The reservoir cannot tell us which mechanism is active, and we have not asked the right question in the laboratory with sufficient rigor. This is without mentioning particle adsorption, dispersion, variability in size, concentration, etc.
The practical consequence is this: if the production improvement is largely or partly driven by a relative permeability effect from the solvent/oil, the treatment economics and the design logic are both wrong. You would need a different treatment frequency, different concentration, and potentially a different strategy altogether. And you would be incorrectly validating a deep-propagation plugging model when the real workhorse is something injected in the first few feet of the formation.None of this means these products do not work. It means we may not know why they work, and that distinction is not trivial.
CDG: When the Active Ingredient May Not Be Active
The case of Colloidal Dispersion Gels is more stark. CDG consists of polyacrylamide crosslinked with aluminum citrate at low concentrations, and its proponents claimed it was superior to ordinary polymer flooding because the CDG particles would preferentially enter high-permeability thief zones and divert flow, acting as a targeted in-depth mobility modifier rather than a simple polymer flood.
The scientific critique of this claim, laid out rigorously by several authors is quite devastating. Independent laboratories at the University of Kansas, the University of Texas, New Mexico Tech, Stavanger College, and BP all found that aluminum-citrate-HPAM colloidal dispersion gelants behave like other gelants and gels: once gelation begins, propagation deep into porous rock is extremely slow or nonexistent. CDG formulations studied had gelation onset times of only a few hours at reservoir temperatures, meaning that by the time a meaningful volume has been injected, gelation at the face of the formation has already begun. The gel does not propagate deep. It plugs near the wellbore.
The vendor’s claim that CDG would preferentially enter high-permeability zones also contradicts a basic principle of fluid flow. For radial injection into unlayered rock, Darcy’s law dictates that a viscous solution penetrates proportionately fartherinto low-permeability zones than water — precisely the opposite of what was claimed. If CDG residual resistance factors are truly higher than for polymer alone, as the vendor asserted, the gel disproportionately damages low-permeability zones, potentially harming sweep efficiency rather than improving it.
Yet CDG was deployed at commercial scale and reported as successful in field trials. How? The most likely explanation is that it worked but not for the claimed reason. Even if the crosslinker does nothing, a CDG formulation is still a polymer solution. A polymer flood improves mobility ratio, reduces fingering, and improves sweep. If CDG provides production improvement roughly equivalent to a polymer flood of the same polymer concentration, it has worked, but as a polymer flood, not as a deep in-depth conformance treatment. The crosslinker is simply a cost that provides no incremental benefit. The vendor’s claim that CDG is superior to polymer flooding is not supported; the reality is that it may be equivalent to it at best, or actually worse if the crosslinker increases residual resistance factors in a permeability-dependent way that penalizes tighter productive zones.This is the most insidious version of the wrong-reason problem: a treatment that works, but not for the advertised reason, which then gets marketed as validated science when the actual validated science is something simpler and cheaper.
WAP and NAP: When a Forced Stop Becomes a Strategy
Water-Alternating-Polymer and Nothing-Alternating-Polymer represent related but distinct ideas. WAP is a deliberate strategy: alternate slugs of polymer and water to reduce polymer consumption while supposedly maintaining sweep efficiency. NAP emerged from a more candid origin: in a heavy oil field in southern Oman, a polymer injection pilot was impacted by injectivity problems caused by surface facility constraints, water quality issues, and bacterial activity, which forced repeated shutdowns of the injectors (Battashi et al.). The producers, however, kept producing. Not only did production not collapse during the shut-in periods, it continued to improve: water cut kept falling and oil rate kept rising even with no polymer being injected.
This observation led to a legitimate scientific question: what is the physical mechanism? The subsequent modeling work provided a coherent explanation. Polymer injection creates a viscous bank that acts as a barrier between the underlying aquifer and the oil column, cutting the water cones and suppressing water influx. When injection stops, the aquifer pushes oil and this polymer bank upward toward the producers, sweeping additional oil ahead of it. Polymer adsorption creates a residual resistance factor that retards the re-entry of aquifer water. The oil mobilized into depleted water cones during polymer injection is subsequently recharged and produced by this aquifer drive during the shut-in. The mechanism is plausible, the modeling matched the data, and the economic argument is arithmetically appealing. But a harder question is shared by both WAP and NAP: would continuous polymer injection have recovered more oil, faster?
In the WAP case, the fundamental physics is unambiguous and unfavorable. When a lower-viscosity fluid (water) is injected behind a higher-viscosity established polymer bank, the mobility ratio at that interface is adverse. Water fingers through the polymer bank rather than displacing it as a coherent front. The Captain field simulation study (Johnson et al., 2023) makes this explicit: a hard switch back to water from a 1,500 ppm polymer flood resulted in “almost complete loss of the improved sweep” as water formed miscible fingers through the established bank. At 400 ppm taper concentration, unstable fingers still formed; only at 800 ppm was the front kept stable. This is not a new discovery, it is what the physics predicts. Injecting a less viscous fluid behind a more viscous one creates instability. Yet WAP continues to be proposed and tested as an economic optimization in contexts where the science already tells us what will happen.
The NAP case is more nuanced because the aquifer provides a natural driving force that substitutes for the absent polymer injection. In that specific context, i.e. strong bottom-up aquifer drive, heavy oil, horizontal injectors placed just above the oil-water contact, the shut-in period is not truly a period of no displacement but a period of aquifer-driven displacement. The NAP results from Oman are therefore not transferable to a waterflooded sandstone without a strong aquifer, where stopping injection simply means stopping displacement. Even within the Oman context, we will never know whether continuous injection would have outperformed the involuntary NAP sequence, because there is no parallel universe in which that experiment was run. The simulation comparison was not validated against an actual field trial of continuous injection at the same conditions, it compared against a modeled counterfactual calibrated to match the involuntary shutdowns. The NAP strategy emerged from operational issue, was observed to not make things worse, received a physical interpretation, and has since been pushed as optimization. That is not the same as demonstrating it is better than continuous injection.
Polymer Tapering: It Works, But Is It the Best We Can Do?
Polymer tapering consists in progressively reducing the injected polymer concentration as the flood matures and is an approach presented as an economically rational strategy, and in principle it is. The argument is that as the polymer bank establishes and sweeps the reservoir, the concentration required to maintain a stable front decreases, so the incremental cost of full-concentration polymer is no longer justified.
At the Captain field offshore UK, a taper from 1,500 ppm to 800 ppm achieved a 45% reduction in polymer consumption with no observable decline in the water-oil ratio of the swept producers. The two wells in the tapered pattern became, for a period, the most productive in the entire field. This is a genuine result. But the question it does not answer is the same one that haunts the WAP and NAP discussions: what would have happened under continued full-concentration injection?
The Captain paper is careful in some respects: the simulation work explicitly demonstrates that a hard switch to water produces viscous fingering and near-complete loss of enhanced sweep, which correctly rules out the naive cost-reduction strategy. The choice of 800 ppm rather than 400 ppm was physically motivated by the need to keep the displacement front stable. These are the right questions to ask.But the comparison presented is between tapering and stopping, not between tapering and continuing. The production data shows that WOR did not deteriorate during the taper period, which establishes that the taper did not cause harm. It does not establish that the taper was optimal. If continued injection at 1,500 ppm would have swept additional uncontacted volumes or pushed the oil bank faster to the producers, the 45% reduction in polymer cost may have been offset by a reduction in the rate of oil production during that period — a trade-off that is economically meaningful in an offshore field where lifting costs are high and time value of money is significant.
The paper acknowledges this implicitly by framing the taper as freeing polymer supply-chain capacity to accelerate injection in other patterns which is economically rational if, and only if, the marginal value of polymer in the new patterns exceeds the marginal loss in the tapered pattern. That calculation requires knowing the counterfactual recovery under continued full-concentration injection, which was not modeled against the actual field data.This is not a criticism of the work done but a structural observation about how tapering studies are designed. The question “does tapering harm performance compared to stopping polymer?” is answerable and has been answered. The harder question — “is tapering better than continuing, and if so at what concentration?” requires a comparison that simulation (which is always incomplete) can approximate but field data alone, from a single taper trial in a single pattern, cannot resolve.The honest conclusion from the Captain data is: tapering at 800 ppm did not make things worse. Whether it was the best economically recoverable strategy remains an open question.
Bullheading Gels Without Zone Isolation: When Plugging Is the Point, but Not the Right Point
The bullheading of gel treatments into injection wells without zone isolation is a different category of problem. Here the issue is not that we cannot determine why it worked. It is that it cannot work for the stated reason, and basic reservoir engineering tells us so before a single barrel of chemical is injected.The practice is widespread. An injector is suspected of channeling into a high-permeability streak connected to a producing well. A gel treatment is designed to plug that channel. The gel is then pumped directly down the wellbore and into the formation (bullheaded), without any attempt to isolate the offending interval with packers or mechanical diversion. The treatment is declared successful when the injection profile improves or when the producing well water cut declines.
The problem is elementary. A bullheaded fluid enters the formation in proportion to the injectivity of each interval, governed by permeability and the existing pressure field. The highest-permeability zone (the very thief zone the operator wants to plug) will accept the most fluid, which sounds correct until you consider what happens next. The lower-permeability zones that are not the problem will also accept some gel, which may actually set and reduce their already-limited injectivity. The net result is the opposite of the intention: you have damaged the productive intervals and left the thief zone largely untreated. This is not a modeling artifact or a theoretical concern: it follows directly from Darcy’s law. This is why zone isolation is not an operational convenience; it is a necessity for a gel treatment to contact the intended interval. Without it, the treatment design is incoherent from the first principles of fluid flow in layered systems. Another point that doesn’t help is that there will be no sweep improvement beyond the point of penetration.
Why then do bullheaded treatments sometimes appear to work? The answer returns us to the central theme of this article. If the gel partially sets in the wellbore or in the first few centimeters of formation, it may reduce total injection rate, which changes the producing well response in ways that can superficially resemble improved conformance. And if you plug, you plug — but what you have plugged is rarely what you intended. The near-wellbore region of every interval has received some treatment; the deep channel has received less than intended. When water breakthrough resumes (fingering through the gel in the high permeability zone), as it will, the re-treatment interval is shorter and the damage to the productive zones accumulates with each cycle.
The post hoc fallacy is again fully operational: the treatment was applied, something changed, success is declared. But the mechanism (selective deep plugging of the thief zone) was physically impossible from the start. That this practice persists, is routinely reported as successful in conference proceedings, and is recommended in field development plans without mention of zone isolation is one of the clearest examples in the industry of building repeated practice on a foundation the underlying science does not support.
A Common Thread
These five cases share a structure. In each, a field observation is real: conformance treatments shift flow patterns, CDG injection improves oil recovery, NAP shut-ins sustain oil production, polymer tapering maintains WOR while cutting chemical costs, and bullheaded gel treatments sometimes change injection profiles. The question is not whether something happened, but why and whether something else would have worked better.
The error arises when we infer mechanism from outcome, or optimality from the absence of harm. In reservoir engineering, this inference is systematically difficult because:
- Experiments cannot be repeated. A reservoir treated with a microgel cannot then be treated with carrier fluid alone, returned to its original state, and re-tested. The counterfactual is inaccessible. Positive production results after any treatment are consistent with many hypotheses.
- Multiple mechanisms operate simultaneously. A microgel injection that changes relative permeability and plugs channels will always produce an ambiguous result. A CDG that acts as a polymer flood and occasionally creates near-wellbore gel will always produce an ambiguous result. A NAP sequence that benefits from aquifer drive and avoids polymer loss to the aquifer is doubly ambiguous. A taper that maintains sweep and coincides with natural flood maturation is doubly ambiguous. A bullheaded gel treatment that reduces water cut and coincides with natural pressure depletion in the thief zone is doubly ambiguous. The observed outcome cannot discriminate between competing mechanisms without careful experimental design, which is hard in the field.
- The relevant comparison is not always the one that was made. Showing that a taper performs comparably to stopping polymer is not the same as showing it performs comparably to continuing. Showing that NAP does not collapse production is not the same as showing it outperforms continuous injection. Showing that CDG achieves production improvement is not the same as showing it outperforms a plain polymer flood at equivalent polymer concentration. The benchmark matters enormously, and it is too often chosen for its accessibility rather than its relevance.
- Confirmation bias is structurally embedded. Papers about EOR pilots are almost always written by people with an interest (scientific, commercial, or reputational) in the success of the technology being tested. Negative results are rarely published. The peer-reviewed record on any given EOR concept therefore overrepresents the positive outcomes, a form of survivorship bias at the level of the literature rather than the individual well. We read about the successes and extrapolate to general conclusions, ignoring the unpublished failures.
- Qualitative matches are presented as validation. A simulation that matches production history after fitting several uncertain parameters (polymer degradation rate, relative permeability, kv/kh, etc.) is not an independent validation of the proposed mechanism. It is a demonstration that the model, with enough degrees of freedom, can reproduce the observations. That is a necessary but not sufficient condition for confidence in the mechanism.
What Honest Evaluation Looks Like
None of this implies that these technologies do not work or should not be used. Polymer flooding works. Conformance treatments can work. Cyclic injection might work, sometimes. Tapering is economically rational under the right conditions. The point is narrower: the claimed mechanisms, and the claimed optimality, need to meet the same evidentiary standard as the claimed outcomes.
Concretely, this means defining the right benchmark before the trial, not after. For microgels: the comparison should include a carrier-fluid-only injection at the same volume, to isolate the particle effect from the solvent effect. For CDG: the comparison must be against a plain polymer flood at the same polymer concentration, not against water injection, making sure that the crosslinker propagates for the CDG. For NAP and WAP: the comparison must include a modeled and where possible field-tested scenario of continuous polymer injection at equivalent cost, not just a comparison against stopping. For tapering: the economic optimization must be calculated against continued full-concentration injection, not just against the option of stopping. For bullheaded gel treatments: the comparison must be against a zone-isolated treatment at equivalent gel volume, and the claimed mechanism of deep selective plugging must be demonstrated with tracers or production logging, not inferred from a surface water cut response.
Designing the monitoring to discriminate between mechanisms is equally important. Interwell tracers, pressure transient analysis, and produced fluid sampling can, in principle, distinguish near-wellbore effects from deep diversion, polymer degradation from relative permeability change, and rate effects from fluid-type effects. These diagnostics are rarely deployed with the rigor needed to resolve the mechanistic question.
Finally, resist the generalization from individual field cases to universal conclusions. A NAP concept validated in a heavy oil reservoir with strong aquifer drive and horizontal injectors does not transfer automatically to a waterflooded sandstone with a weak aquifer and vertical wells. A tapering trial in an offshore North Sea field does not transfer automatically to an onshore heavy oil flood where logistics constraints and time-value-of-money considerations are completely different. The papers presenting these concepts often have a single paragraph acknowledging this and then proceed to recommend universal adoption.
Conclusion
The oil industry is not unusual in this respect. The scurvy example is instructive precisely because the error was not one of incompetence but of insufficient experimental design: the right treatment was used, patients recovered, and the wrong lesson was drawn; a lesson that held until conditions changed and the mechanism gap became lethal. Social scientists face the same problem whenever a policy intervention is evaluated without a proper control group. Economists debate it whenever a stimulus is followed by growth. The fallacy is ancient and universal.
In reservoir engineering, the unpublished failures, the untested counterfactuals, the confounding mechanisms we did not instrument for, and the parameters we adjusted in history matching until the simulation agreed with the data — these are the invisible evidence. The fact that a technology works in some cases, observed some of the time, does not establish that it works for the reason we think and it certainly does not establish that it is the best available option.
Claiming success is not dishonest. Claiming to understand success when you do not (or claiming optimality when you have only tested against an accessible but irrelevant benchmark) is a different matter. The field deserves more of the latter kind of honesty, even when it makes for less satisfying conference papers.
References
Battashi, M., Farajzadeh, R., Bimani, A., Al Abri, M., Mjeni, R., Karpan, V., Fadili, A., and van Wunnik, J. “Insights into Oil Recovery Mechanism by Nothing-Alternating-Polymer (NAP) Concept.” SPE-207743-MS, Abu Dhabi International Petroleum Exhibition & Conference, 2021. (also published in Journal of Petroleum Science and Engineering, 2022)
Frampton, H., Morgan, J.C., Cheung, S.K., Munson, L., Chang, K.T., and Williams, D. “Development of a Novel Waterflood Conformance Improvement Tool with Potential for Deepwater Applications.” SPE-88033-MS, SPE/DOE Symposium on Improved Oil Recovery, Tulsa, 2004.
Johnson, G., Lugo, N., Neal, A., McBeath, J., and Farthing, D. “Improving the Economics of Polymerflood EOR Through Polymer Tapering at the Captain Field.” 84th EAGE Annual Conference & Exhibition, Vienna, 2023.
Seright, R.S. “Impact of Dispersion on Gel Placement for Profile Control.” SPE Reservoir Engineering, 6(3): 343–352, August 1991. https://doi.org/10.2118/20127-PA
Seright, R.S. and Brattekas, B. “Water Shutoff and Conformance Improvement: An Introduction.” Petroleum Science, 18: 450–478, 2021. https://doi.org/10.1007/s12182-021-00546-1