Two versions of one observation
A CPI number can carry a reference month, a publication date, an adjustment label and a later revision date. Losing any of those fields makes a comparison difficult to explain. This worksheet compares two saved versions of the same observation. It does not estimate the next inflation report or turn a revised number into a trading signal.
The practical result is a short record: what you originally saved, what the later source says, whether the two entries are comparable, and the arithmetic difference in the correct unit. Keep the original entry even when a newer value becomes available. That preserves the information you actually had when writing your earlier note.
Reference month is not release day
The BLS CPI schedule separates its reference month from the date of publication. On the schedule checked October 20, 2026, the October observation is due November 10 at 08:30. The monthly calendar supplies the Eastern Time label. This is scheduled publication evidence; it is not an observed October CPI value.
In your own record, store the month as YYYY-MM and the document date separately. Give the time you checked the document a third field. A later visit to an old release does not change its publication date. If a source does not show a publication time, leave that field unknown instead of copying your browser clock.
| Record field | Example or rule |
|---|---|
| Reference month | October 2026 |
| Scheduled publication | November 10, 2026, 08:30 Eastern; checked October 20 |
| Actual published value | Not supplied by this schedule |
| Your observation time | When you personally read the relevant document |
Sources: BLS - Consumer Price Index release schedule, BLS - November 2026 release calendar.
Match the labels before subtracting
Compare the same series, geographic coverage, population, item category, seasonal treatment and index base. A national all-items observation is not interchangeable with a local or category-specific observation. A screenshot labelled simply CPI may omit enough information that comparison must stop.
This is a matching exercise rather than a confidence score. Copy a series identifier when the source provides one. For each side, retain the exact value as displayed and record whether it is an index level or a rate. If one input is a level and the other is a monthly percentage change, the worksheet must reject the pair. A small numerical difference does not repair mismatched labels.
Sources: BLS - CPI frequently asked questions, BLS - Calculating percent changes.
Identify the kind of change
BLS describes an annual recalculation of seasonal factors that can revise the previous five years of seasonally adjusted indexes. Its 2026 page identifies a February 13 release covering January 2021 through December 2025. That process should not be attached automatically to every CPI number.
The CPI FAQ distinguishes ordinary CPI-U/CPI-W publication, rare error corrections and the separate revision process for C-CPI-U. Record the explanation given by the actual source. Do not label a changed figure an error unless the publisher does. If you only have two screenshots and no revision notice, record different saved values with the cause unknown.
Sources: BLS - Seasonal Adjustment in the CPI, BLS - CPI frequently asked questions.
A worked comparison without a live data feed
Every number in the following table is invented for arithmetic practice. These are not published BLS observations. Both level entries in the first row refer to the same hypothetical month, series and adjustment basis. Subtracting the earlier version from the later version gives the revision difference; dividing that difference by the earlier level gives a relative difference between versions.
A rate comparison needs a different label. Changing a displayed rate from 0.2% to 0.3% is a difference of 0.1 percentage point. It is not an index-point change and should not be described as an additional 0.1% rise in consumer prices caused by this worksheet. The arithmetic compares records; it does not establish an economic cause.
| Hypothetical comparable inputs | Calculation | Correct description |
|---|---|---|
| Index level: 300.000 to 300.300 | 300.300 - 300.000 = +0.300 | +0.300 index point; +0.100% relative to the earlier version |
| Rate: 0.2% to 0.3% | 0.3 - 0.2 = +0.1 | +0.1 percentage point |
| Index level: 300.300 to 300.000 | 300.000 - 300.300 = -0.300 | -0.300 index point; approximately -0.099900% relative difference |
| October versus September | Comparison blocked | Different reference periods are not revisions of the same observation |
Use a strict comparison contract
Enter finite numeric values without currency signs. Index levels must be positive. Rate values can be zero or negative, but both sides must use the same rate definition and period. An empty field is missing, not zero. Keep a negative revision signed so the exported record can distinguish an upward change from a downward one.
For an index pair, calculate later minus earlier in index points, then divide by the earlier value and multiply by 100. For a rate pair, report percentage points only. Preserve more precision internally than you display, and state the rounding used. Do not calculate an official month-on-month inflation rate from two vintages of one month; that calculation requires observations for different months on a compatible basis.
Sources: BLS - Calculating percent changes.
Keep both vintages in your notebook
Use two rows or two clearly named columns. Alongside each value, save the source URL, publication date when shown, date checked and the exact reference period. Add a note explaining why the pair was accepted or rejected. Your comparison can remain useful even when the conclusion is not comparable.
For a repeatable review, retain the earlier export instead of overwriting it. Add a new comparison when a later source changes the record. A correction to your own typing should be labelled as your correction, with the earlier input preserved. This proposed recordkeeping method is ours; it is not a BLS certification or an official archive service.
What this result can support
The finished worksheet can explain a disagreement between two saved notes, make the units of a revision explicit, or show why a comparison should be rejected. It cannot validate a screenshot without its source, recover a historical vintage you never saved, predict a policy decision or determine how a market will react.
Recheck scheduled dates at the official publisher before using them. After a scheduled day passes, keep the example as a dated example until you verify the actual document. The optional link below leads to this publisher's existing Telegram offer. The comparison and worked table are usable without joining it; this page makes no claim about service eligibility in the United States.
Worked examples
Example 1
- original value
- 300
- revised value
- 300.3
- unit
- index_level
- matching labels
- True
- comparability
- comparable
- signed difference
- 0.3
- difference unit
- index_points
- relative difference percent
- 0.1
Synthetic values for one observation; not a month-on-month inflation calculation.
Example 2
- original value
- 0.2
- revised value
- 0.3
- unit
- percent_rate
- matching labels
- True
- measurement
- monthly_percent_change
- comparability
- comparable
- signed difference
- 0.1
- difference unit
- percentage_points
- relative difference percent
- unknown
Subtract the two displayed rates; relative percentage change is intentionally omitted.
Example 3
- original value
- 300
- revised value
- 301
- original reference period
- 2026-09
- revised reference period
- 2026-10
- comparability
- not_comparable
- signed difference
- unknown
Different reference months must not be treated as revisions of one observation.
Example 4
- original value
- 0.2
- revised value
- 0.3
- unit
- percent_rate
- original measurement
- monthly_percent_change
- revised measurement
- annual_percent_change
- comparability
- not_comparable
- signed difference
- unknown
Matching percent signs do not make monthly and annual rate definitions comparable.
Questions before you use the result
Is a later download always a revised CPI observation?
No. It may contain the same value, a different series, a different adjustment basis or a corrected record. Match the labels and inspect the publisher's explanation.
Can unadjusted CPI never change?
Do not make that claim. BLS distinguishes routine publication from rare error corrections and treats C-CPI-U separately. Check the exact series and notice.
Does the relative difference measure inflation between two months?
No. Here it measures the numerical difference between two saved versions of one observation. A between-month inflation calculation is a different comparison.
Primary sources and scope
- BLS - Consumer Price Index release scheduleChecked 2026-10-20. Reference month October 2026 is listed with release date November 10, 2026, at 08:30 AM.
- BLS - November 2026 release calendarChecked 2026-10-20. The calendar labels its times as Eastern Time. It lists October CPI and Real Earnings for November 10 at 08:30.
- BLS - Seasonal Adjustment in the CPIChecked 2026-10-20. Annual seasonal-factor updates may revise the preceding five years of seasonally adjusted indexes. Factors released February 13, 2026 covered January 2021 through December 2025.
- BLS - CPI frequently asked questionsChecked 2026-10-20. CPI-U/CPI-W do not routinely revise for late-arriving data; errors can be corrected. C-CPI-U follows a separate revision process. Index points and percent changes are different units.
- BLS - Calculating percent changesChecked 2026-10-20. Percent change is difference divided by the earlier index, multiplied by 100. Comparable calculations use the same index reference base. A twelve-month comparison uses the same month in consecutive years.
Prepared from the sources listed above. Examples and calculations explain a method; they do not predict prices or certify an investment. Geographic time references do not establish where a trading service is available. Dated schedules can change: consult the named publisher before relying on a future event.