--- id: VOLT-HOME-WP-083 title: "When does a simple time-of-use tariff beat a fully dynamic tariff?" slug: when-does-a-simple-time-of-use-tariff-beat-a-fully-dynamic-tariff description: "A declared wholesale-signal comparison between a simple time-of-use proxy and idealized dynamic exposure." published: 2026-08-30 cluster: "Tariff economics and consumer safeguards" status: measured evidence: /research-data/home-papers/when-does-a-simple-time-of-use-tariff-beat-a-fully-dynamic-tariff.json figure: /research-media/home-papers/when-does-a-simple-time-of-use-tariff-beat-a-fully-dynamic-tariff.webp figure_alt: "Chart for When does a simple time-of-use tariff beat a fully dynamic tariff?: dynamic-versus-simple-TOU flexibility signal, shown as TOU, dynamic." source_ids: - dynamic-tariff-viability - acer-retail-2025 - ec-retail-flexibility-2026 - energy-policy-flexibility - iea-demand-flexibility peer_reviewed: false --- ## Abstract This working paper compares two deliberately simple wholesale flexibility signals. The dynamic proxy is 35% of the mean observed daily day-ahead price range. The time-of-use proxy is 25% of the median daily range. Across 93,161 observations, the registered dynamic-minus-TOU signal is 64.61511430212212 EUR/MWh-equivalent. The positive sign means that, under this frozen idealized construction, the dynamic proxy is larger; the study does not observe a simple time-of-use tariff beating dynamic exposure. A simple tariff could still be preferable if dynamic-price access, forecasting, automation, risk, or contract costs consume more than the modeled signal advantage, but those costs are not estimated here. The 4,000 kWh synthetic household and 35% flexible-energy assumptions provide the family context, while the primary metric remains a wholesale signal comparison rather than a household bill. Taxes, VAT, supplier margin, network charges, device losses, and customer telemetry are excluded. Every monetary quantity is a declared synthetic wholesale scenario, never a measured customer saving or retail-bill result. ## Plain-language answer In the exact scenario tested, the simple time-of-use proxy does not beat the dynamic proxy. The dynamic signal is larger by 64.6151 EUR/MWh-equivalent. That number is not a saving and is not an all-in tariff difference. It measures a gap between two formulas applied to observed daily wholesale price ranges. The comparison intentionally gives the two designs different capture assumptions. Dynamic exposure receives 35% of the mean daily range. The simple TOU design receives 25% of the median daily range. The test therefore asks whether a lower, fixed capture fraction of a typical range matches a higher capture fraction of the average range. It does not build an actual peak/off-peak calendar, assign prices to fixed hours, or measure how household load responds. A simple tariff can nevertheless win in a broader decision if it avoids costs or failures that the idealized dynamic proxy ignores. Examples include subscription fees, forecast mistakes, controller downtime, appliance constraints, price-tail aversion, or a network tariff that conflicts with wholesale prices. This paper does not price those factors. Its defensible answer is conditional: dynamic leads in the frozen wholesale-signal scenario; the break-even burden for unmodeled dynamic frictions is the measured 64.6151 EUR/MWh-equivalent gap on this metric, not a guaranteed retail threshold. ## Research question The research question is whether a simple time-of-use rule can capture enough recurring wholesale dispersion to outperform fully varying dynamic exposure once simplicity is acknowledged. The registered implementation narrows that broad question to a comparison of two transparent proxies. The outcome is not causal and not a field experiment. There are no households randomly assigned to tariffs, no observed bills, and no measured automation behavior. “Beat” means “produce the larger declared wholesale flexibility signal under the two formulas,” not “deliver lower total annual retail cost.” This precise definition prevents usability, risk protection, and retail fees from being smuggled into a wholesale statistic. ## Data and provenance The public evidence derives from a frozen SELECT-only production snapshot and contains no customer telemetry. The family’s registered contracts are `day_ahead_prices`, `forecast_accuracy`, `bess_index_daily`, `zone_load`, and `zone_holidays`. The comparison itself uses daily maximum and minimum day-ahead prices to form zone-day ranges. The evidence reports 93,161 observations. Daily prices cover 2021-01-01 through 2026-08-29. The corpus also records detailed intervals from 2025-10-01 through 2026-08-29 and a long-history boundary from 2015-01-01 through 2026-08-29. The publication cutoff is 2026-08-30T00:00:00Z. The listed tables and windows define the governed evidence surface; they do not imply that every table enters this two-formula result. The snapshot SHA-256 is `7e97489fc8528c8cc8c38830e05b48d949ce1f67b98575dff26f5d7c321e4c67`. The preregistration is identified by `adb36bf6b447af9f96339249b8becaefc20422499cca1977242866347a97bd4b`, the paper registry by `7bcb91d7476d0a69fe9fa75a5c7782f8117e0153f82f9112b7e1d307d3943717`, the source registry by `07949550ac443ff673fda5c0209b99f137544f3ffecf6775f109bb9d09663bd6`, and the analysis code by `57c57de79cdab2b5b6d6c54c485cb5162598c5ba0b0bfe995da40d75e6c52ba9`. The evidence-manifest hashes are `76959fd0c79ebb505932f79cfec85a3324cace2ac54f51a60a5a3226eff8d53c` for the JSON and `5df3e289c125cc29ad3dca52182576f81a1158270b9db52d05f8ba5e6a329fb9` for the WebP figure. The extraction ran in a read-only transaction with a 180-second timeout. The references supply context about retail choice and flexibility but do not contribute numerical findings to this study. ## Method For each included zone-day, the analysis computes `max_price - min_price`. It then calculates the arithmetic mean and median of the resulting daily-range distribution. The idealized dynamic signal is: `mean daily range × 0.35`. The simple TOU signal is: `median daily range × 0.25`. The primary metric subtracts the TOU signal from the dynamic signal. A positive result favors the dynamic proxy on this wholesale-range scale; a negative result would favor the TOU proxy. The 25% assumption is registered as “TOU captures 25% of median daily range.” It is an assumption, not a fitted behavioral estimate. Likewise, the 35% dynamic share is declared rather than inferred from an appliance schedule. The broader family describes a synthetic household with 4,000 kWh annual use and 35% flexible energy. Annual kWh does not enter the reported EUR/MWh-equivalent difference. Scaling it into an annual synthetic wholesale amount would require a separate declared conversion and should not be confused with a retail bill. Neither “tariff” is implemented as an actual retail contract. The TOU side has no named clock periods, season, weekday rule, or network component. The dynamic side assumes access to daily wholesale dispersion but has no forecast, scheduling error, price cap, or risk premium. Taxes, VAT, supplier margin, and fixed fees are absent. No device is simulated, so battery efficiency and degradation, EV charging efficiency and availability, and heat-pump COP and comfort constraints are all excluded. The result is descriptive. The family preregistration provides Holm correction for inferential claims, but this paper makes no unadjusted significance claim. ## Results The registered dynamic-minus-simple-TOU flexibility signal is 64.61511430212212 EUR/MWh-equivalent. Rounded, the idealized dynamic proxy exceeds the simple proxy by 64.62 EUR/MWh-equivalent. This is a synthetic wholesale scenario quantity. It is not a measured customer saving, a retail price, or an annual bill difference. Because the sign is positive, the measured evidence does not show TOU beating dynamic pricing. The paper’s supplied interpretation says a simple tariff can outperform when full dynamic exposure is costly or automation is constrained. That is a conditional mechanism statement, not the empirical outcome of this calculation. No cost or failure distribution was added to the dynamic side, so the study cannot report how often that condition occurs. The figure values are exactly 23.8675 EUR/MWh-equivalent for the TOU proxy and 88.48261430212213 EUR/MWh-equivalent for the dynamic proxy. The regenerated evidence reports `bootstrap_95_interval: null` and `interval_method: "not reported for this estimand"`, so no confidence interval is claimed for their 64.61511430212212 difference. The result is best read as a hurdle: any claimed simplicity advantage must be compared against the dynamic proxy’s measured wholesale-signal lead, using commensurate units and explicit costs. It is not permission to subtract 64.62 EUR/MWh directly from a customer’s bill. ## Robustness and placebo checks The most important robustness grid varies the two capture fractions before outcome inspection. A TOU design with better alignment than the assumed 25% would narrow the gap. A dynamic system with less than 35% usable flexibility would also narrow it. Because both sides are linear in their declared fractions, a sensitivity surface can show where the sign changes. The frozen evidence displays only the registered comparison, so this paper does not claim a sign over all plausible fractions. A no-dispersion placebo sets each daily maximum equal to its minimum. Both signals then equal zero, and their difference must be zero. A distribution-shape check compares mean and median daily ranges. Since the dynamic and TOU sides deliberately use different summaries, skew from extreme days can favor the mean-based dynamic proxy even before capture fractions differ. Reporting that design choice is essential. A genuine TOU robustness test would define fixed local-time periods using only information available before delivery, apply them unchanged across an out-of-sample window, and value the same synthetic load under both tariffs. It would test seasons, weekends, daylight-saving transitions, and time-varying network charges. This paper does none of those things and therefore cannot validate a particular peak/off-peak product. An automation-cost stress would subtract declared forecast, control, and subscription costs from the dynamic side while applying comparable implementation costs to TOU. A risk stress would compare upper-tail daily costs, not only flexibility signal. Those are proposed extensions, not hidden components of the present result. Serial dependence matters because adjacent market days and neighboring zones are related. The corpus requires blocked day or week bootstrap methods for inferential work and Holm control within this ten-paper family. Here, the result remains descriptive. ## Limitations The proxies are asymmetric by construction: 35% of a mean is compared with 25% of a median. This is transparent but does not isolate “dynamic” versus “TOU” as tariff technologies. A different capture assumption or distribution summary could change the result. There is no standardized intraday household load shape despite the 4,000 kWh annual family assumption. Annual energy does not enter the primary metric. The test does not know whether flexible consumption is available during low-price periods or whether inflexible consumption remains in high-price periods. No physical asset is represented. For batteries, there is no capacity, inverter limit, state-of-charge, round-trip efficiency, or degradation cost. For EVs, there is no arrival, deadline, charger power, or charging efficiency. For heat pumps, there is no COP, thermal loss, hot-water demand, or comfort constraint. A simple tariff may be easier for one asset and poorly aligned with another, but this analysis cannot distinguish them. The scenario excludes taxes, VAT, supplier margin, network charges, retail risk premiums, fixed fees, and contractual caps. A constant adder changes total cost but not interval ordering; a time-varying network charge can change the preferred interval. Neither is modeled. The pooled zone-day distribution also prevents a direct conclusion for any named market. Finally, customer savings, bills, and preferences were not measured. Simplicity has operational and cognitive value, but this paper does not quantify it. The evidence reports no interval for the primary metric. ## Practical implication The practical choice between TOU and dynamic pricing should be made with an all-in, household-specific simulation. Start with the local retail tariff, preserve its actual time periods or dynamic formula, add the household’s load and controllable assets, and include forecast errors, controller availability, efficiency, degradation or comfort, taxes, and network charges. Use the 64.6151 EUR/MWh-equivalent result only as the frozen dynamic wholesale-signal lead under the registered proxies. If transparent, commensurate simplicity benefits and avoided costs exceed that lead, TOU could win in the expanded scenario. This paper does not show that they do. ## Reproducibility Verify the evidence JSON named in the frontmatter against the paper ID, slug, title, status, assumptions, date cutoff, source tables, all five provenance hashes, and the evidence-manifest hashes for the JSON and figure. Confirm the source transaction was read-only. From the frozen daily-price records, calculate each maximum-minus-minimum range. Compute the arithmetic mean and the median using the analysis code’s conventions. Multiply the mean by 0.35 to obtain the dynamic proxy. Multiply the median by 0.25 to obtain the TOU proxy. Subtract TOU from dynamic. The output should equal 64.61511430212212 EUR/MWh-equivalent. Reproduce the two figure bars from those component values. Any implementation that chooses peak hours, applies a retail tariff, changes the capture fractions, uses annual kWh, or adds device constraints is a new scenario and should receive its own evidence record. ## Disclosure Analysis and drafting were model-assisted. The paper exposes its formulas, frozen evidence, content hashes, and omitted costs. It is a public working paper, not peer reviewed, and not financial, tariff-selection, purchasing, investment, or trading advice. Every monetary quantity is a declared synthetic wholesale scenario and never a customer saving or retail bill. Volt has no live traders or live capital; C0R is the only paper strategy. This tariff-method paper is not trading research and does not authorize or evaluate trading. ## References 1. Advances in Applied Energy, [Assessing the conditions for economic viability of dynamic electricity retail tariffs for households](https://doi.org/10.1016/j.adapen.2024.100174). 2. ACER and CEER, [Rewarding flexibility: How retail contract choice can help unlock consumer flexibility](https://www.ceer.eu/wp-content/uploads/2025/11/ACER-CEER-2025-Retail-monitoring.pdf). 3. European Commission, [Communication from the Commission on the Citizens Energy Package](https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX%3A52026DC0115). 4. Energy Policy, [Welfare redistribution through flexibility — Who pays?](https://doi.org/10.1016/j.enpol.2025.114684). 5. International Energy Agency, [Scaling Up Demand Flexibility](https://www.iea.org/reports/scaling-up-demand-flexibility).