An illustrative example: a drawer of unlabeled envelopes, each one a "this used to cost less" story told at a holiday, none of them with a date, a store, or a size. The stories feel precise because they are repeated. They are not yet measurements. A personal log is a small set of purchases written down so that a later year can be compared with an earlier year on purpose. It will not reproduce the CPI. It will keep a family from promoting a mood into a percent.

What the log is for

The log answers a household question: what did we pay, on this date, for this specified item, at this place? Collected the same way once a year, it becomes a private price level for a basket you actually buy. The CPI-U answers a different question: how did out-of-pocket prices change for an average urban basket, using BLS sampling and formulas. Both can be kept. Neither corrects the other automatically. The reasons a receipt diverges from the headline are laid out in why a grocery bill is not the official inflation rate. The log is how you retain the receipt side without pretending it is the headline.

Five to eight items are enough. A longer list becomes a second job and then stops. Choose items that recur, that have a unit you can write down, and that you would still recognize in five years. A specific loaf, a tank of a named grade of gasoline, a monthly transit pass, a haircut at the same shop, a kilowatt-hour if the bill states one, a rent payment with the address attached. Skip fashion and electronics unless you are willing to record the model. A "phone" in 2018 and a "phone" now are different goods, and the CPI program exists partly because that problem is hard. Your log should refuse items you cannot specify.

BLS faces the same specification problem in public and says so. Average prices are published only for relatively homogeneous things — foods, a few fuels, utility gas, electricity — and not for cars, apparel, furniture, or electronics, "due to the high degree of product differentiation," in the words of the average price fact sheet. Borrow that restraint. If the national statistical agency will not publish an average price for a category because the items are too unlike, a household percent change for "a shirt" is theater.

The fields that make a row usable

Each row needs six fields, and a row missing any of them should stay out of the comparison.

Field Write this Leave the row out if
Date The day you paid You only remember the season
Item The specification, including size or grade The label is "groceries" or "a meal"
Place Store, landlord, or biller "Somewhere downtown"
Unit Per pound, per gallon, per month, per visit You cannot say what one means
Amount paid The number on the receipt, with tax called out if it was included The figure is a round memory
Note Brand change, sale, move, household change You need the note later and did not write it

Sales tax is worth a note because BLS food average prices are published without it, while utility and fuel average prices include taxes. If your grocery row includes tax and the public series does not, a gap can appear that is only the tax. Write "including tax" or "shelf price" and keep the rule for that item forever.

A sale price is a real price you paid and a poor candidate for a once-a-year tradition unless you always buy that item on sale. The note field is where "loyalty-card price" goes. Comparing a sale year with a non-sale year measures the promotion as much as it measures the market.

Once a year, on a date you already keep

Tie the update to something that already happens: a birthday month, a lease renewal, the week you do taxes. The point is repetition, not intensity. Pull the same items, in the same units, and add a row. If an item has disappeared from your life, stop the series and start a new one with a new name. Do not quietly substitute a different good and keep the old label. That substitution is exactly what makes a sloppy index unreadable, and you do not have a BLS-style quality adjustment to fall back on.

Compute the change only along a series with two complete rows. The percent change is the later amount minus the earlier amount, divided by the earlier amount. Multiply by one hundred if you want a percent. Use the same unit in both rows. A price per pound against a price per loaf is not a percent change in anything. Keep one extra digit while you divide and round at the end. If you have three years, you can compute each step and the change from the first year to the last. The last of those is not the average of the steps, and you do not need it to be. Report the span you actually calculated.

Do not divide that percent by a count of years and call the result "annual inflation" unless you are explicit that it is a household item's average pace, not the CPI. Do not average unlike items into one household inflation rate unless you also choose weights and write them down. An unweighted average of a rent percent and a banana percent treats bananas and rent as equal shares of your life. They are not. If you want weights, use rough shares of your own spending — rent's share, food's share — and say they are yours. If you do not want that project, publish the items separately and let them stay separate. A table of eight honest lines is more informative than one blended number you cannot reconstruct.

Set the log beside an official series without forcing a match

For foods and fuels on the BLS average-price list, you can put the agency's national or regional level for a nearby month in an extra column. Read it as a level for a specified item, not as a grade of your shopping. BLS cautions that about a quarter of the CPI sample rotates each year and that this rotation can move an average price without moving the index, because the index does not compare a new outlet with an old one. Your store is one outlet. Their average is a weighted sample. A year in which you disagree is ordinary.

For the all-items picture, read the CPI-U and, if you want the index the Federal Reserve uses for its longer-run goal, the PCE price index. They differ for structural reasons covered in CPI, PCE, and why headlines disagree. Neither one should be pasted into the log as "the correct percent" for your banana row. The useful comparison is qualitative: did your specified item move in the same direction as the closest official component, over the same months? If it did not, the note field is where the reason goes — a different brand, a move to a new city, a one-time repair.

Shelter rows need the housing distinction. A rent row compares with CPI rent only in the loose sense that both are rents; the CPI rent sample is not your building. A house purchase belongs in a different log, next to a named house-price index, and must not be dropped into the consumption list. Mixing them is how family stories invent a cost of living from an asset. The map is house prices, rent, and the same number trap.

An inflation calculator is the wrong machine for the log. Calculators scale a dollar amount by a national index. Your log already has the dollar amounts. Running a receipt through the BLS calculator answers what the CPI-U basket did, which you can look up directly. It does not validate the receipt. The setup of those tools is in how inflation calculators actually work.

How the log earns its keep in conversation

At the holiday table, the rule is small. A claim about the past needs a row. "Milk was nothing" does not survive contact with a dated unit price, and it does not need to become a lecture. You read the two rows, name the unit, and stop. People can still prefer the old price. They cannot honestly remember a different one while the row is on the table.

Children can add a row for one item without taking on the whole basket. The teaching sequence that keeps that exercise from turning into anxiety is how to explain what money is worth to kids. One specified food, a date, and a store is the assignment. The percent can wait until there are two dates.

Travel adds a currency field, not a new philosophy. If you paid in another currency, record the currency, the amount in that currency, and the provider rate you actually received if you know it. Converting the old foreign price at today's rate and calling the result "what it cost" replaces a historical rate with a current one. A purchasing-power comparison across countries is a further project, and it is not required for a domestic log. The glossary distinguishes a nominal exchange rate from purchasing power if that trip row starts an argument.

Store the log where you will find it: a single sheet, a notebook page, or a file with a boring name. Back it up when you back up other household papers. A log that lives in one person's memory is the envelope drawer again.

What the log cannot decide

A personal price log does not say whether a wage was fair, whether a move was wise, or what you should buy next year. It does not predict inflation. It records specified purchases so a later comparison has inputs. Official series change their samples and methods; your notes should change only when the thing you buy changes, and then the series name should change with them. This is household record-keeping and general education, not financial advice.

Which item could you write down today with the date, the place, and the unit still attached?