Back in 2007, a UK advertising campaign for Colgate toothpaste declared that;
‘More than 80% of dentists recommend Colgate’.
A reasonable interpretation of this claim is that around 80% of dentists recommended Colgate over all other brands. In reality, what the survey asked was for dentists to recommend several toothpastes and brands, not just a single choice. In other words, a number of brands and toothpastes were recommended by dentists, Colgate just happened to be one of them around 80% of the time.
The claim is misleading. But it isn’t a lie…
“Torture numbers, and they’ll confess to anything,” begins a chapter about numbers in Hector MacDonald’s book, Truth.
Like the rest of the book, this chapter carries a common theme — that several people can say very different things about the same subject, whilst all telling the truth. And yet, each person’s account on its own can seem like a lie in the face of contrasting evidence. MacDonald calls these types of truths ‘competing truths’, illustrating that there is no such thing as the truth — only a possibility for interpretation of the facts.
If we (brands/advertisers/agencies) are genuine about wanting to be transparent with consumers, we need to educate ourselves on how easily messages can be misunderstood, even — or maybe especially — when we don’t mean them to be.
The 2007 Colgate campaign, which was ultimately banned by the ASA (you can read more about it here), has since been used as a case study for both misleading headline copy and poor survey practices.
Here is a brief overview of the main points of the chapter:
Being picky with numbers.
Numbers that are taken out of context can seem impressive until they are viewed relative to other facts.
A delivery company proudly claiming that its fleet uses 1 million gallons less fuel per year is admirable…unless your fleet is made up of 100,000 vehicles, rendering the saving a drop in the ocean.
If you want to a number to look smaller, change the timescales or break it down into smaller fractions.
The Royal family costs tax payers just 52p per year…but there is a lot of tax payers in the UK. Taken together, it adds up to £35.7 million.
Getting political.
During the run up to the 2016 referendum, the treasury published a forecast suggesting that by 2030 British GDP would be 6% less outside the EU than in. Osborne’s press release read, “Britain will be worse off by £4,300 a year per household if Britain votes to leave the EU.”
This gave the impression that Britain would be worse off than it is now. In fact the forecast showed that Britain would have significantly higher GDP in 2030 than it has now, whether inside or outside the EU. (It would be worse off than it would have been inside the EU, but still better off than it is now.)
The press release also suggests that each household would be £4,300 poorer each year. But GDP, which factors in corporate investments and government expenditure, has little bearing on individual household earnings.
Using percentages out of context.
A Children with Cancer UK press release stated that cancer cases have risen 40% in past 16 years, sparking a predictable run of panic headlines.
This seemingly ignored that the number of children born in the UK had increased during the same period, that diagnostic tools have improved leading to more cases being diagnosed and treated earlier and that, ultimately, UK cancer death rates had actually dropped 24% in the previous decade.
Correlation does not imply causation.
A correlation chart shows that the more ice cream sold in beach resorts, the more people seem to drown. Is there something in the ice cream causing people to forget how to swim? Or was the shock from seeing such a tragedy causing swimmers to comfort eat?!
Of course not. People tend to eat ice cream when it’s hot. People are more likely to go swimming in warm weather. One does not cause the other. Instead, both are affected by a third factor.
This silly example reflects how correlations in data sets can be used as ‘evidence’ to support an agenda — for instance taking a rise in immigration numbers with a rise in crime to support anti-immigration measures, despite there being no evidence the two are connected.
Statistical averages: Median, mean and mode.
If I wanted to estimate the average weight of everyone in my marketing agency, I could add up all the individual weights and divide that number by the number of people in the agency, giving me the average weight. But there is more than one type of average. This average would be the mean average. I could also line everyone up and take the weight of the person in the middle, giving me the median. Finally, if I were to take the weight that occurs most often, I would have the mode average weight of the group.
In 2014–15, the UK median income before tax was £22,400. The mean income before tax for the same year was £31,800.
Taking these figures, and what we know about different types of statistical averages, both of these statements would be true:
A teacher on a salary of £28,000 is earning below the average income.
A teacher on a salary of £28,000 is earning above the average income.
It is therefore a tactic to choose the type of average that best suits your agenda.
In summary
Hector provides some useful guidance on what to watch out for when handling numbers. This is relevant for consumers and citizens, but also for marketers wanting to ensure their messages are transparent.
- Understand what each number in a debate really represents.
- Make sure you’re using relevant units and comparing like-for-like.
- Put numbers in context by comparing them to other relevant numbers to show how big they actually are.
What do you think? Do we have a responsibility to make messaging more honest? If we can’t find a positive message to communicate, what does that mean for your campaign?
