Wednesday, February 17, 2010


It increased by how much?

In yesterday’s meetings, someone presented some data, showing us rates of sending spam, month by month, broken down by country. The summary page showed the change from six months ago to last month. Here’s an example of the summary table, with made-up numbers:

Spam by country, as % of total
CountryJul 09Jan 10Change
Gumana2%< 1%-1%

Maybe you already see the problem with this.

A change from 4% to 8% does not represent a 4% increase. It’s a 100% increase — the spam from Slobovia has not gone up by 4% in six months, it’s doubled. And similarly for the other numbers: Parador showed a 56.4% decrease.

What is true is that Slobovia’s contribution to the total has increased by 4 percentage points (which is not the same as saying that it’s increased by 4 percent). And if it’s clear that that’s what you’re saying, it’s a fine thing to say... we can then debate which number is more useful, and the answer to that will depend upon what we’re using the numbers for.

But it’s very misleading to list those numbers in a table like that, and entirely wrong to say that “the amount of spam from Slobovia has increased by 4% since last July.”

Gotta be careful with percentages.


Sue VanHattum said...

I agree with your thought, but if the total spam has increased, then your numbers would be wrong too, wouldn't they?

Barry Leiba said...

Yes, that's true: the table doesn't have the information needed to say how much it's increased, in raw numbers... only in its relative contribution to the total.

Brent said...

"Just say no" to percentages of matter what you think you mean, someone will misunderstand you.