‘Up to’ Is Doing All the Work in That Headline Number

The case study headline says results improved by up to 40 percent, and you are deciding whether to repeat it or set your own target by it. Is 40 percent what happened, or the best thing that happened? Those are different claims, and the phrase supports only one.

What ‘up to’ is actually claiming

Strip the marketing off the phrase and it becomes arithmetic. “Up to 40 percent” says the highest value anywhere in the data was 40 percent, and every other value was 40 percent or less. It is a stated maximum, a ceiling, and as a description of the data it is usually perfectly true.

What it does not tell you is how crowded the space near that ceiling is. A maximum of 40 percent is consistent with results that mostly landed near 35 percent, where the ceiling fairly describes the campaign, and equally consistent with results that mostly landed near 3 percent, where one outcome sits far above the rest. The phrase cannot distinguish those two worlds. That is not a flaw in the wording, it is what the words mean.

Why the headline reaches for the ceiling instead of the average

The ceiling is attractive because it is the largest number available and a defensible one.

That ceiling usually comes from one of three places: the best week or month lifted out of a twelve month campaign, the best client account or store in a larger portfolio, or the best channel or audience segment inside a broader result. The same four words work for all three, and the headline rarely says which. It is the neighbour of the case where one outlier account can pull the whole average with it, except that this outlier stands alone.

A hypothetical case: the ceiling next to the median

Here is a hypothetical example, with numbers invented for illustration. They belong to no real agency and no real campaign, and describe no case study published on this site or anywhere else.

Picture an agency running the same paid social programme across fourteen client accounts for six months, measuring lead volume against each client’s prior six months. One account came in at 38 percent more leads. That is the ceiling, and the honest headline for it is “up to 38 percent”. Line the fourteen up in order and look at the middle of that list, and in this invented set the median sits at 9 percent.

Both numbers describe the same fictional programme. The distance between them is 29 percentage points, and a reader who took 38 percent as the expected outcome has set an expectation more than four times what the typical account delivered. Nothing was falsified. The ceiling was asked to stand in for the middle, which it cannot do.

The questions that recover the median

Four questions close most of that gap.

• Up to X percent of what group, and over what time period? Six accounts over three months is a different object from sixty accounts over two years.
• Is this the single best performing unit, or something close to the typical unit? Ask it in those words. The answer is short.
• What share of the accounts, weeks, or stores came close to this ceiling, versus far below it? “Nine of fourteen were within five points” and “one of fourteen got there” fit the same headline.
• Does the report give a median, an average, or any distribution at all, or only the highest number? A range, a median, or a chart of individual accounts means the ceiling is not all you have.

Press on the first one. Chasing “of what group, exactly” often lands you in how a small base inflates a percentage.

A ceiling claim is not the same failure as a bare multiplier

A multiplier reported as 3x or 10x hides how large the base was, so you cannot tell whether tripling meant two to six or two thousand to six thousand. An “up to X percent” claim hides something else: where the typical case sits beneath a stated peak. One claim can carry both problems, but they are caught by different questions, the starting number in one case and the middle of the distribution in the other.

Reading past the headline number

None of this makes an “up to” claim dishonest. The ceiling is usually a real measurement, and leading with your best result is what everyone does with good news. The right response is to go looking for the other kind of number in the body of the report: a median, a range, a table of accounts, a chart with more than one point.

If the report gives none of those, be plain about your position: you cannot reconstruct a median from a maximum. Treat the ceiling as an unrepresentative best case until shown otherwise, and do not carry it into your own planning as an expected outcome. For the wider habit, see the general framework for reading a case study.

Read more of them, or add one

Ceilings standing in for averages are easiest to spot with results claims side by side. Browse the case study library, where every study is credited to the agency that published it, or submit your own if your work reports its numbers properly.

FAQ

Does ‘up to’ mean the result is fake?

Usually not. The ceiling is normally a genuine maximum from real data. The problem is omission, not fabrication: the highest number is published, the middle is left out, and the reader fills the gap with an assumption the document never supported.

Is there a more honest way to state a range of results?

Yes, and it costs one sentence. Pair the ceiling with a median, or with the spread behind it, and the reader can judge the distance themselves. A bare maximum leaves that gap unknown, so every reader guesses at it privately.

How is an ‘up to’ claim different from a claim like ‘3x results’?

They hide different things. A multiplier hides the size of the base it multiplied, so you cannot tell how much movement 3x represents. An “up to” ceiling reports a real peak but hides where the typical case falls beneath it. Ask the multiplier for its starting number, and ask the ceiling for its middle.

Sources

• One Outlier Account Can Skew an Agency’s Portfolio Average, fetched 7 September 2026.
• When the Base Is Small, the Percentage Is Just Noise, fetched 7 September 2026.
• How to Read a Marketing Case Study Without Being Misled, fetched 7 September 2026.