There has always been a strange fascination in polls, those central pillars of the market research world.  They have a thousand and one uses, and of all the tools of the trade they are the closest thing we have to a Delphic oracle or Roman seers, who once found fame reading the entrails of birds and petitioning the gods to reveal great secrets of the universe.

It’s not an exaggeration to say that polls can seem like magic.  When used carefully – with randomized respondents, carefully selected demographic groups, and finely-honed questions, they can have a seemingly miraculous ability to define future political or commercial trends, such as the polls which correctly predicted the outcome of the 2017 UK general election.

However, just like the oracle, polls can be as much a tool to shape reality, as to measure it.  With equal care and attention – say a slightly leading question, or a response group made up of your best friend, your nan and the family dog, polls can also be guided to find the ‘right’ answer and lend it a nebulous air of scientific credibility.  Sometimes this is deliberate, if not completely ethical, and it’s not hard to think of examples in everyday media where this has been the case (we’re looking at you, 80% of dentists polled).

Sometimes, though, researchers can be so set on the goal of their exercise, this kind of bias can even creep in unintentionally.  It’s all very well asking consumers whether they prefer their product to be red or black – provided you remember to find out whether they want your product at all.

The fact that surveys can be used in such widely differing ways illustrates how important design is, and how efficiently and consciously you have to assess questions from every angle – especially when your own views, basic assumptions and even cultural background can make the difference between an accurate poll and a huge waste of time and money.