Avoiding Statistical Bias in Decision-Making Practices.

“99 percent of all statistics only tell 49 percent of the story.” ― Ron DeLegge II

Avoiding Statistical Bias in Decision-Making Practices.

When developing marketing strategy, strong infographics, images, and video content often empower our campaigns. Vast training courses and resources are up-skilling many, improving their working knowledge of how to display content effectively to influence decision makers.

Running a business can often feel like running a gauntlet. We often review what technology we use, to ensure we have the most powerful and suitable tech for our projects. With so many options now available, for almost every function, it's common to question how best to equip our team.  Whilst this insight will be applicable to thought bias of multiple areas, we'll focus on software decisions for the purpose of conveying a concise example.

Every day our designers work with image editing tools for multiple clients. Already champions in this field, Adobe's Photoshop holds pride of place among some of our go-to applications. However, carefully crafted (and targeted) adverts often land on our screens urging us to invest in the new 'best' applications or software. These adverts are backed up with impressive statistics that almost seem too good to be true.

So, how can we tell when it is too good to be true?
Or, perhaps, how can we spot bogus statistics before wasting our time down a rabbit-hole of warped information aiming to lure us into a quick nonrefundable purchase and bitter regret at being misled.

“Facts are stubborn things, but statistics are pliable.”
- Mark Twain

Here a few common pit falls to avoid:

The "Gee Whizz" Graph.

As attention spans dip, and time seems to be the increasingly valuable commodity slipping away from us, advertisers use easily digestible infographics to convey messages at speed. The issue we may face, is how easy it is to spin statistics to our own advantage. The 'Gee Whizz' graph is an example of using scales selectively to further a certain goal. In the pictured example, we find the same data set displayed in three ways with three very different outcomes.

  1. A basic, expected, display range of 0-24 (Billion Dollars) in increments of 2 on the Y axis is supported with a monthly increment on the X axis. This shows that the growth was steady, but arguably slow from 20 - 22 over a 12 month period. Given the large empty space on the graph, our minds will often deduce this as a positive but mainly unremarkable increase.
  2. With a shortened Y axis only ranging from 18-24 (Billion Dollars), the same increase suddenly looks much more dramatic. By narrowing the range, we are given the sense of a much more positive and notable change.
  3. With the Y axis zoomed to increments of 0.2 billion dollars, the growth is much clearer. The graph now suggests a huge boom of (assumably) profit throughout the year.

A forth idea, would be to increase the X access from monthly increments to quarterly, or biannual steps. The results would be squashed into a more overwhelming sudden spike - promoting a sense of sudden dramatic growth rather than slow and steady progress.

This 'Gee Whizz' concept works for comparisons also.

Imagine two companies operate without creative agencies in 2018. In 2019 one company (A) starts work with a main competitor of Mäd, whereas the other (B) make the smart move of teaming up with us. By the beginning of 2020, A has increased profits to $46'000, whereas B is up $48'000.  

All things considered, the profit gap seems negligible and hardly our go-to for sales data. However, if we display the difference at a scale (on the Y Axis) of 45.5k to 48.25k in increments of .25k, our success looks huge.

Putting the two graphs side by side shows how ridiculously skewed data can be presented, and yet, this is extremely commonplace in the media. Our eyes are drawn to the results and not the scales - which allows advertisers to get away with clear misrepresentation.

The Missing Figures

Another easy trick, is simply removing information.

Given the nature of the above graphs, we could simply create a graph showing a different in profit between companies - yet without the Y axis labelled, it's meaningless.

It may seem common sense, but when our brains aren't tuned to look for all the information, it's very easy to subconsciously absorb suggested results - giving us biased opinions unconsciously.

Advertisers can use this trick in their copy too, rather than purely through clever imagery. For example, the age old claim of '95% of dentists prefer Brand X' may sound strong - but it leaves a lot of questions to be answered, often in a forum whereby we can't ask questions. What are they comparing Brand X to, how many were surveyed, where are these dentists from and what are their backgrounds, do any of the dentists work with Brand X directly or have an affiliation?

Correlation is not Causation

“One of the first things taught in introductory statistics textbooks is that correlation is not causation. It is also one of the first things forgotten.”
Thomas Sowell, The Vision of the Anointed: Self-Congratulation as a Basis for Social Policy

In contrast to missing data, it is sadly common for unrelated data to be combined to suggest dubious conclusions. A company may make a large amount of profit in a certain quarter - this could be due to an economic boom, a particular big project, a new market trend or for any number of other reasons. At the same time, in that quarter, the company may have started using a new piece of software, or hired one new member of staff (let's call them Jamie), or even introduced optimal yoga classes on Friday afternoons.

From the above consideration, it would be unreasonable to suggest that the company made their profit purely due to the new member of staff, or due to the fact the team were now able to practice yoga, or indeed that they had a new software tool to hand. However, it's not false for the company to claim: 'We introduced Jamie this quarter and our profits rose massively'.

Should other companies try and poach Jamie?
Can Jamie take credit for the economy booming?

Spin doctors can use, or ignore, multiple factors to their own end - implying correlation and tricking us to agree with their narrative. Once again, it's important to train our brain to think critically. When presented with information, it's a useful practice to briefly acknowledge the source and what their motive may be. For example, if Pepsi are telling us about the benefits of increased sugar in our diets, we'd be right to be skeptical. Or, if a research lab told us of the benefits of consuming high quantities of bacon, perhaps we should double check that they aren't funded by F&B industry giants.

Time Constraint Offers

If we rationally consider claims and data presented to us in adverts, we'll often be able to spot wild exaggerations or dubious figures. Sneakily, statistical bias campaigns will often be coupled with time constraints designed to rush us into blindly accepting claims that will bring us from A to Pay instead of A to B.

Be wary of the following:

  • The digital countdown that claims you only have 3 hours left to check-out a software at a particular price;
  • The e-mail follow up claiming you can have a special 50% discount as a bonus if you sign up today (and today only!);
  • The 'flash one-off exclusive sale' that claims to be a once in a life time amazing offer.
  • An 'upgrade' to a better package if bought immediately.

The reality, is that advertisers can use urgency as a tool to fluster potential customers into impulse buys. Good software (and businesses) don't need cheap tricks, and a rushed decision is likely a risked decision.

Of course, sometimes clever offers are paired with good incentives, so there's no blanket rule other than exercising caution and sensibly reviewing information.

Conclusion.

There's a reason that 'word-of-mouth' is thought to be the most powerful PR. Testimonials from trusted peers will always be more convincing than overwhelming advertising, and in a world where information is so easily manipulated online, we should always exercise caution.

When possible, request a trial and a walkthrough of any new softwares - and if such is applicable to whatever other purchases you're mulling over, an honest salesperson is likely to be glad to spend time showing how genuinely impressive their product is.

Surrounding yourself with industry experts always helps, which at Mäd we've done by meticulous hiring and networking. If you're unsure on a particular subject, seeking out a non-biased expert to review the potential purchase could help give you a clearer picture, other considerations or just the reassurance you need to ensure you've not been duped.

That's why, when we switch software, we tend to find the cutting edge platforms that last the test of time.