Skip to content

Systems and records

Utility policy effective date dataset: which tariff rule applied when

By SourceX Editorial · Reviewed by Noah Loul ·

10 sections

Short answer

To find the tariff or service rule that applied on a date, select the latest version whose approved effective date falls on or before the service date, that was not suspended, withdrawn or superseded by then, and that the account was enrolled in. Apply the tariff's proration rule when a bill period crosses a change, and rerun the selection when a meter reading is corrected.

Record timeline

Utility policy effective

This page covers utility policy effective date dataset: which tariff rule applied when.Illustrative

Key takeaways

  • Each rule version carries filed, proposed, approved and end dates; only the approved effective date selects the version.
  • FERC filings use 12/31/9998 when an effective date is unknown, so placeholder dates must be flagged, not used.
  • Bill periods that straddle a change need the tariff's own proration rule recorded with the split.
  • Corrected readings can move consumption between versions, so keep both results linked by revision.
  • Public rate databases are a cross-check; the regulator's records and the utility's billing history are the authority.

How do you identify the tariff or service rule that applied on a specific date?

Pick the latest version of the rule whose approved effective date is on or before the service date, that was not withdrawn, suspended or superseded by then, and that the account was actually enrolled in on that date. Then apply the tariff's own proration rule if the billing period straddles a change. Every other shortcut, such as the date on the tariff sheet, the date the filing was made or the rate shown on today's website, picks the wrong version often enough to poison a dataset.

That is why a utility policy effective date dataset needs more than a copy of the current tariff book. It needs every version of every sheet or record, the four different dates attached to each, the account's enrollment history and the reading and bill periods it is applied to. With those, a person or a model can answer "which rule applied on that date" and show the work.

This guide gives the rule version task design an energy company can use to build that record, a resolution procedure, a worked example with a corrected meter reading and the points where the task should stop and go to a rates specialist.

Filed, proposed, approved and effective: why one rule has four dates

A tariff change carries several dates, and only one of them decides which version governed service. Under FERC's rule at 18 CFR 35.3, rate schedules and tariffs are generally filed not less than 60 days nor more than 120 days before the service or change is to take effect, so the filing date and the effective date are deliberately apart.

FERC's eTariff implementation guide makes the proposed effective date a required field used to determine which version of a tariff record is effective, tells filers to use 12/31/9998 when the effective date is not known at filing, such as when it is contingent on Commission action, and says changing a proposed effective date requires a new tariff filing. Each version also carries a unique record version number.

State commissions add their own timing. Under the California Public Utilities Commission's General Order 96-B, a Tier 1 advice letter is effective pending disposition, Tier 2 is effective after staff approval (or deemed approval when unprotested and not acted on within the review period), and Tier 3 requires a Commission resolution. Other states use different procedures, so the record must store the authority and procedure for each version rather than assuming one model.

Filed, proposed, approved and effective: why one rule has four dates
DateWhere it is foundCommon trapHow to store it
Filed or submitted dateFiling cover letter, docket or advice letterTreated as the start of the new rateKeep, but never use to select a version
Proposed effective dateFiling, eTariff record, advice letter requestPlaceholder dates such as 12/31/9998 read as realKeep with a flag when it is a placeholder
Approved effective dateOrder, resolution, acceptance letter or deemed approvalDiffers from the proposed date after suspension or settlementPrimary selector for the version
End or superseded dateNext version's effective date or a withdrawalLeft blank, so two versions overlapDerive from the successor and check for gaps
Customer enrollment datesBilling system rate history for the accountCurrent schedule applied to old billsStore per account with start and end

The rule version task design: what each record must carry

The table below is the asset: the fields a rule version record and its linked account record need before the question "which rule applied" can be answered and checked. Example values are illustrative, not taken from any utility.

The rule version task design: what each record must carry
FieldHoldsWhy it mattersIllustrative value
rule_idStable identifier of the schedule, rule or sheetLinks every version of the same ruleSchedule GS-2, Sheet 14
version_idRevision number or eTariff record versionTwo versions of one sheet must never share an id3rd Revised Sheet 14
supersedesThe version this one replacesBuilds the chain used to find gaps and overlaps2nd Revised Sheet 14
authority_refRegulator, docket, advice letter or order numberLets a reviewer confirm the approvalIllustrative advice letter, Tier 2
effective_fromApproved effective date and its basisThe selector for service datesApproved, deemed after review period
effective_toDay before the successor took effect, or openPrevents two versions applying at onceOpen
status_historyFiled, suspended, approved, withdrawn, rejected, with datesA suspended version must not be selectedFiled, then approved
proration_ruleHow the tariff splits a bill period across a changeDecides charges on straddling billsDays-weighted, per tariff rule text
account_enrollmentAccount, schedule, start and end datesThe rule must be one the account was onAccount 001 on GS-2 since an earlier year
reading_revisionMeter read, read date, revision number and reasonCorrected reads change what each version is applied toRevision 2, estimated read replaced by actual

Where do superseded tariff versions and rate histories actually live?

In at least three places, and a dataset needs all of them joined. The regulator's filing system holds what was filed, suspended and approved; the utility's own tariff book or archive holds the sheets as published; and the billing system holds the effective-dated rate tables that actually priced each bill, along with the account's schedule history.

These sources disagree more often than teams expect. A billing rate table may have been loaded a few days after the approved date and back-billed, a sheet archive may keep only the current revision of each page, and a regulator's record may list a filing under a docket number the billing team never recorded. Joining them on rule_id and version_id, and storing the authority reference with every billing rate row, is what turns three partial histories into one checkable record.

Export the billing system's rate tables with their own effective dates and load timestamps rather than a report of current rates. If the system overwrote old rates in place, rebuild them from bill line items and flag those versions as reconstructed, so no one later treats a reconstructed rate as a primary record.

From service date to one rule version: the resolution procedure

Run the same steps in the same order for every question, and record the result of each step, so a reviewer can see which step decided the answer.

  • 1. Identify the account and the service date or period in question, from the reading or bill record, not from the invoice date.
  • 2. Find the schedules the account was enrolled in on that date from its rate history.
  • 3. For each schedule, list every version with an approved effective date on or before the service date.
  • 4. Drop versions that were suspended, rejected or withdrawn as of that date, and versions whose effective date is still a placeholder.
  • 5. Take the version with the latest approved effective date; if two share a date, apply the documented priority or option rule and record it.
  • 6. If the bill period crosses an effective date, apply the tariff's proration rule to each part and record the split.
  • 7. If the reading was revised, rerun steps 1 to 6 on the corrected reading and keep both results with the revision link.
  • 8. Record the answer with the authority reference, so the rule can be checked against the order or advice letter that approved it.

Billing periods that straddle a rate change

A bill period that starts under one version and ends under another is where most disputes and most bad labels come from. The safest default is that the old version governs service before the new version's effective date, and the tariff's own proration language decides how a single period is split.

A California Public Utilities Commission staff document on complaints about a water utility shows why this belongs in the record: it describes a utility applying newly approved rates to an entire billing period that began before they took effect, and states that until new rates become effective the previously approved rates are the only lawful rates a utility can charge. A dataset that stores only the bill total and the current schedule cannot reveal that kind of error, and a model trained on it will repeat it.

Store the proration method as text quoted from the tariff and as a coded value, and keep both the days and the consumption assigned to each version. Where the tariff is silent, mark the record for a rates specialist instead of guessing.

Worked example (illustrative): one bill, two versions and a corrected reading

This example is invented to show the procedure; no real utility, customer or tariff is described. A commercial account on a general service schedule has a 31 day bill period. A revised version of the energy charge sheet was approved partway through, after its originally proposed date was suspended, and the first reading for the period was an estimate later replaced by an actual reading.

  • The exception: if the new version's approved effective date had still been a placeholder when the bill was produced, the correct label is "pending", not the old or the new rate, and the record goes to a rates specialist.
  • What a model should learn here is the order of operations, not the arithmetic: enrollment, then version status, then proration, then reading revisions.
Worked example (illustrative): one bill, two versions and a corrected reading
StepWhat the record showsDecisionWhy
EnrollmentAccount on the general service schedule for the whole periodUse that schedule's versions onlyThe account did not change schedules
Candidate versionsOld version open; new version proposed for day 1, suspended, approved for day 18Old version days 1 to 17, new version days 18 to 31Approved date, not proposed date, selects
ProrationTariff text splits by days in each partIllustrative split: 17 days and 14 daysQuoted rule recorded with the split
Corrected readingRevision 2 replaces the estimated start readRerun the split on corrected consumptionBoth results kept, linked by revision
Review flagNone after rerunClose with authority referencesEvery step has a source document

Can a public rate database replace the tariff book?

No. Use public rate databases to cross-check, not as the authority. The U.S. Utility Rate Database on OpenEI publishes rates with a start date marking the rate's effective date, an end date and a field naming the rate a record replaces, and its version 3 changes added historical tracking of rate changes. That structure is a useful model for your own fields and a quick check on obvious gaps.

It is still a secondary compilation. It may not carry every rider, special contract or suspension. The tariff sheets, orders and advice letters on the regulator's site, plus the utility's own billing history, remain the records a dataset should be built from.

Which dated fact changes the permissible next step?

Three kinds of dated fact change what a person or an agent may do next, and the task should stop at each rather than infer. A version subject to refund, settlement or rehearing can still change after the fact; a customer-specific contract may override the general schedule for one account; and a corrected reading or asset record can move consumption between versions after a bill was issued.

The task design is also bounded on purpose. It answers which administrative rule applied and why. It does not decide disconnections, safety-related service rules, field work or any operating action on equipment, and it should not be used to give customers legal conclusions about refunds. Those go to the utility's rates, legal and operations staff.

Confidentiality is the last boundary. Customer-specific contracts, account identifiers, service addresses and usage histories are confidential business and customer information. Access to a billing system is not permission to reuse its contents, privacy permission differs from ownership of the records, and both differ from contractual rights to license them; removing identifiers does not by itself create a right to license.

Licensing dated rule and billing histories

Rule version histories linked to real bill periods and corrected readings are the kind of operational evidence AI developers need to test whether an agent applies the right rule for a date rather than the current one. A utility that wants to license dated rule and billing histories can do so through SourceX, which runs the rights review, contracting, preparation, delivery and payment.

Account records are delivered only after the utility approves the scope and signs, and eligibility is checked beforehand through a short assessment that asks for no billing data.

Frequently asked questions

Is the effective date on a tariff sheet always the date the rate applied?

Not always. A sheet can show a proposed date that was later suspended or changed, and FERC filings use 12/31/9998 when the date is unknown at filing. Use the approved effective date from the order, resolution or acceptance and record where it came from.

How should a dataset handle a bill period that crosses a rate change?

Split the period using the tariff's own proration language and store the days and consumption assigned to each version. A California commission staff document records a dispute where newly approved rates were applied to a whole billing period that began before they took effect.

What is the difference between a filed date and an effective date?

The filed date is when the utility submits the change; the effective date is when the change governs service. Under FERC's rule they are normally 60 to 120 days apart, and state procedures such as California's advice letter tiers set their own timing.

Can I use the OpenEI Utility Rate Database as the source of truth?

Use it as a cross-check. It records effective and end dates and links a rate to the rate it replaces, which helps find gaps, but the regulator's tariff records and the utility's billing history are the authority.

When should a rule-version question go to a rates specialist?

When the approved effective date is still pending, the version is subject to refund or settlement, a customer-specific contract may apply, the tariff is silent on proration, or a corrected reading changes which version governs.

Sources

  • 18 CFR 35.3 requires rate schedules and tariffs to be filed not less than 60 nor more than 120 days before the service or change is to take effect. Source
  • FERC's eTariff implementation guide uses the proposed effective date to determine which tariff record version is effective, requires 12/31/9998 when the date is unknown, requires a new filing to change it and gives each version a unique record version number. Source
  • CPUC General Order 96-B sets advice letter tiers: Tier 1 effective pending disposition, Tier 2 effective after staff approval or deemed approval, Tier 3 effective on Commission resolution. Source
  • A CPUC staff document on complaints against a water utility describes newly approved rates applied to a whole billing period and states that previously approved rates are the only lawful rates until new rates take effect. Source
  • The OpenEI Utility Rate Database API documents a startdate field for a rate's effective date, an enddate field and a field naming the rate a record supersedes. Source
  • OpenEI's announcement of URDB version 3 features describes historical tracking of rate changes with links to superseded rates. Source

See if your company qualifies

A short company assessment. No data uploads are needed.

See if your company qualifies