US Inflation Tracker: Why One Number Is Never Enough
1. Why is there more than one US inflation number?
Because there is no single true one. Inflation is not a fact you can go and measure, like a temperature. It is an average of millions of price changes, and the answer depends on which prices you count and how much weight you give each one. The US inflation tracker draws every measure on one chart, on the same axis and as of the same moment, so you can see where they disagree.
Two numbers get quoted most. The first follows what a normal household buys: food, rent, fuel, haircuts. That is the consumer price index, usually shortened to CPI. The second covers a wider list of spending and lets people swap one item for a cheaper one. That is PCE inflation, and it is the measure the Fed sets its 2 percent goal on. They almost never match, and the wider one usually runs a little lower. That gap is not an error. It is what happens when you build two measures to do two different jobs.
From there it multiplies. Take the jumpy bits out and you have core. Take housing out on top of that and you have supercore. A trim does something else entirely, throwing away whichever prices went wild in the month you are looking at, and if you sort prices by how often they change at all you end up with sticky and flexible. The tracker carries 38 lines in all, grouped so you can find the one you want without knowing its name first.
2. What is core inflation, and why does it leave out food and energy?
Core inflation is the same shopping list with food and energy taken out. Those two jump around for reasons that have nothing to do with the economy running hot: weather, harvests, a pipeline, a war. Remove them and you get a steadier line. Core CPI and core PCE are the two you will hear argued about, and core PCE is the one the Fed leans on hardest.
Core is useful, and it is a blunt instrument all the same. It removes the same two things every single month, whether or not those two were the odd ones out that month, and a fixed list like that only works if whatever you removed is always noise and everything you kept is always signal. Neither is true. A one-off jump in something core keeps, like a tobacco tax rise, lands straight in the number. A calm month in fuel gets thrown out anyway, even though it had nothing to hide.
That is why the page never shows core on its own. It sits next to measures built on a completely different rule, so you can see when the rule is costing you something.
3. What is supercore inflation, and is there an official one?
Supercore takes core one step further. It keeps only services, then takes housing out as well. What is left is the things people do for you rather than sell you: haircuts, repairs, dental work, insurance, car servicing. The idea caught on after a central bank speech in late 2022. The theory was that wages are the biggest cost of a service, so this slice should track the jobs market. The link has looked weaker than people hoped ever since.
There is no official supercore CPI. Nobody publishes services with energy and housing stripped out as a ready-made series, so we build it ourselves, and the page says so in three separate places rather than letting you assume otherwise. We take the published services aggregate, subtract the published housing aggregate, and weight that subtraction using the official basket weight published for that month, not the weight it carries today applied backwards. Then we chain the monthly results into an index.
Our version starts in March 2012, because that is as far back as month-by-month basket weights exist in machine-readable form. Holding a single weight fixed all the way back to 1990 would have been silently wrong: housing was 31.4 percent of the basket in 2012 and it is 35.3 percent now.
Supercore PCE is different. That one is officially published and we do not touch it. The tracker draws both, so you can check our home-built line against the official one and see how close they run. The full recipe is written out on the methodology page, and the FOMC page shows what the rate setters are doing about any of it.
4. What is a trimmed mean, and what does it throw away?
A trim lines every price up by how far it moved this month, cuts both ends off, then averages what is left. Nothing is banned for good, because the cut list is rebuilt from scratch every month. That is the whole difference from core, which removes the same two things regardless of what they did.
Three trims sit on the tracker and they cut very different amounts. The 16 percent trimmed mean CPI takes the bottom 8 percent and the top 8 percent off, which leaves 84 percent of the basket. The 8 percent is measured by share of spending, not by counting items, and the whole thing is built from 45 spending groups.
The trimmed mean PCE cuts 24 percent off the bottom and 31 percent off the top, so only 45 percent survives. The two ends are deliberately different sizes. That uneven split was chosen by testing which one tracked the underlying trend best, not because it looked tidy.
Median inflation is the extreme version. Rank every price change by size, weighted by spending, keep the single one sitting at the halfway point, and throw the rest away. It is the steadiest line on the page. It also tends to sit a little above the headline figure, so judge it against its own past rather than straight against 2 percent.
You cannot always see what got cut, and the page prints that. For the wider basket the month's included and excluded list is published as a spreadsheet. For the CPI trims only the rule is public, not the monthly casualty list. What you can say safely is that jumpy things, such as fuel, hotel rooms and second-hand cars, land in the cut ends far more often than something steady like rent. Short plain English explainers for each measure sit in the learn section.
5. What is the difference between sticky prices and flexible prices?
Sticky and flexible is not a trim. It does not ask how far a price moved, it asks how often a price changes at all, then keeps one half and drops the other. The cutoff is 4.3 months. An item that changes price less often than once every 4.3 months counts as slow, and anything quicker than that counts as fast. That is a settled fact about the item, not a judgement about this month, so petrol is filed as flexible even in a quiet month.
Prices that change slowly
Haircuts and salons go about 23.7 months between one price change and the next. Medical care runs 14.0 months, education 11.1, rent 11.0, eating out 10.7, and car insurance 5.9.
Prices that change fast
Petrol changes about every 0.7 months, so roughly every three weeks. Fresh fruit and vegetables run 1.3 months, gas and electricity 1.6, meat, fish and eggs 1.9, new cars 2.0, and hotels 3.1.
Anyone setting a price that only moves once a year has to guess what the whole of that year looks like, so the slow half carries more about where inflation is heading. The fast half is jumpy by design, and mostly tells you what already happened to fuel.
6. How do you read a yearly inflation rate against a three month rate?
The tracker gives you four windows: over a year, the last 3 months, the last 6 months, and the last month. The three short ones are shown at an annual rate, meaning where inflation would end up if the last few months carried on exactly like that for a whole year. That is what lets a 3 month figure and a 12 month figure share a scale. It also blows one odd month up into a big looking number, which is why the last month button is the noisiest thing on the page.
The yearly rate tells you where you have been. It is slow, and it is still carrying last autumn. The 3 month rate tells you where you are heading, and it is noisier but it turns first. So 3 months well below the year means inflation is cooling faster than the headline admits, and 3 months above the year means it is reheating and the yearly figure has not caught up yet.
The forward looking lines, meaning the market bets and the forecasts, are already yearly rates. The window buttons leave them alone, and the page says so by name whenever one is on screen. Every figure here is seasonally smoothed, the yearly ones included, so all four windows agree with each other. The yearly figure quoted in the news is not smoothed, so ours can sit a fraction away from it. We would rather print that than hide it. Readings also follow later revisions instead of freezing the first print, and the economic calendar tells you when the next one lands.
7. What is actually pushing inflation up right now?
The contribution panel answers that. It splits the whole shopping list into four bands that add up to the entire basket, then shows how many percentage points of the last 12 months of inflation each band accounts for. If inflation is 3 percent and housing shows 1.1, then housing accounts for 1.1 of those 3 points. A minus number means that part pulled inflation down.
As of July 2026 housing was 35.3 percent of the basket. Services that are not housing came to 24.9 percent, food and energy 21.0 percent, and goods 18.8 percent.
A contribution is the share of spending a band held at the start of the window, multiplied by its own price change across that window. Two of the four bands are exact arithmetic off published aggregates rather than estimates. The bars are signed from a real zero line, because energy goes negative often enough to matter. The panel also adds its own bars up and checks the total against the headline rate, and it flags a mismatch only when there is one.
It is easy to read the panel the wrong way round. A band matters both for how much its own price moved and for how big a slice of spending it is. Housing dominates even in a quiet year, because it is more than a third of everything. Plane tickets can double and barely register. Which band is doing the pushing is also what moves interest rate expectations, because a housing problem and a fuel problem call for very different answers.
8. How do I use the Helious inflation tracker page?
The page has three controls, and the order you touch them in matters.
Start with a preset
There are seven, and each is phrased as a question you might actually have. Start here gives you the number most people mean plus the number the Fed watches instead. The four main measures puts the four you will hear argued about on one chart. The others cover the trims, CPI against PCE, the housing problem, goods against services, and what markets expect next.
Pick your own lines
You can have up to 5 on the chart at once. Every measure has a plain one line description sitting next to its own switch, so you never have to know the jargon before you click. Lines are told apart by dash pattern as well as colour, so they still read if you are colour blind. The labels sit at the end of each line, and they push each other apart when two measures land on the same value.
Set the window and the history
Four buttons set the window, and a separate control sets how far back the axis runs, at 2 years, 5 years or 10 years.
No figures were collected at all for October 2025. Most charts draw straight through a hole like that, which turns that month change into zero and corrupts every 3, 6 and 12 month window touching it. Here the solid line stops, a faint dashed bridge carries the shape across, and the value stays empty everywhere: in the chart, in the table and in the feed. The page also flags separately when a window has to reach back into that missing month, because that is a different problem and deserves its own sentence.
There is also a 2 percent line drawn across the plot, and the page is precise about what it is. The target is the yearly change in headline PCE over the longer run, not core, not CPI, and not any single month. The whole thing works with JavaScript switched off, including a 24 month table with a tooltip on every month, and it sits alongside the rest of the market data pages.
9. How often does the inflation data update, and is there an API?
New readings land every month, each measure on its own schedule, and the page picks them up as they arrive. It follows later revisions too, so a figure can change after it first prints. That is the data being honest, not the page being unstable. The chart shows 10 years by default, and the full history back to 1990 comes through the feed.
Everything on the page is free, with no login, no key and no paywall. There is a machine readable feed for every series, and an endpoint your own AI assistant can read directly, both written up for developers. If you would rather just be told when something moves, each reading also lands in the news feed.
10. Who is this US inflation tracker for?
Anyone who has read that inflation was 3 percent and wondered which 3 percent. You do not need the vocabulary to start. Every measure explains itself in one line, every preset is a question, and no sentence on the page assumes you already know what core means.
In practice it gets used by four sorts of reader. Someone deciding what to do with money cares most about the gap between the yearly rate and the 3 month rate, because that gap is the whole argument about whether borrowing costs come down. A trader wants to know which band is doing the pushing before the next reading, not after it. Then there are students and writers, who need one honest chart instead of six screenshots from six places, and builders who want the series inside their own tools.
If all of this is new, the guides start from the beginning and assume nothing. Live market data sitting next to it is what the paid tiers add. The inflation page itself stays free either way.