The Trick to Seeing Through Marketing Statistics

Tortured StatisticIf you torture the data long enough, it will confess to anything.

One of the challenges with relying on research, particularly research companies use in their own marketing, is that the statistics used have been tortured.

When you see a sales pitch full of statistics, how do you differentiate the real story from the spin? Short of dismissing all of the data presented (a good idea, but not realistic for most of us), these five methods will be a good starting point.

1. Turn Stats Upside Down

I was in a recent meeting where a sales person claimed 46% of B2B buyers have purchased a product because of a video they watched. They didn’t just notice it or remember it, but nearly half of potential buyers that watch a video actually bought a product because of it!

Before you react, turn the same metric upside down. Look at the other side of the same metric: despite all of the videos vendors are creating, more than half of potential buyers say they have never purchased a product because of a video.

Check IconLook at the other side of the same statistic to get a more balanced perspective.

But continue on, we will look at this same statistic again.

2. Put Stats Back in Context

Marketing often isolates statistics, removing them from any context that would allow for appropriate comparisons.

In the example above, the video statistic was highlighted on its own. But is 46% a lot more, or a lot less, than the same metric for other types of content?

Here is a comparison, from the same study: 41% purchased a product because they saw a banner ad on a tablet.

Wow. When compared to a banner ad on a tablet, watching a video is just slightly better. Barely better than a banner ad will never be a claim to fame.

Check IconPut metrics into context of other measurements before evaluating them.

3. Consider What Isn’t Said

Companies that field research for their own marketing activity never share all of the results. They choose what to use and, more importantly, what not to use in their marketing.

Consider carefully the comparisons that are not made or the metrics that are not reported. What would you like to know that isn’t included? It is probably safe to assume that if it was included, the story wouldn’t be nearly as rosy.

Here is a hypothetical example:
79% of end users agree that our solution is easy to use. But what isn’t said? Here are a couple of possibilities: would they recommend it to someone else? Would they buy it again? Our oven is easy to use but it is way to small. I definitely wouldn’t recommend it to someone else or choice it for myself.

Check IconIf you were doing the research, what else would you want to know? Assume that information is less favorable than what the marketer has chosen to present.

4. Learn How The Questions Were Asked

Market researchers can bias the response based on how the question is asked or by the choices that were given. Consider the potential results of these two surveys about marketing priorities:

Which of the following most closely matches your top marketing priority?
A. Improving brand metrics (including awareness, perception, intent to purchase and likelihood to recommend).
B. Demand generation (including lead generation, email marketing, content marketing, pipeline development and sales enablement).

In B2B marketing, B is probably the clear answer for many marketers.

However, if we list the answers differently, we might get a significantly different result. Consider the following list of choices:
A. Brand awareness
B. Content marketing
C. Sales enablement
D. Lead generation
E. Marketing automation

Demand generation marketers will split their answers among the last four choices and brand awareness might become the top priority based on this survey structure.

Check IconGo back to the original question, the full list of choices and the responses to see where the results were biased.

Bonus: here is a classic example of how marketers biased their research to make bacon part of breakfast.

5. Check the Sample

This is a favorite trick among magazine publishers. Publishers field a readership survey and, surprise! every publisher reports the audience likes their specific publication best.

Well, no duh. They send the survey to their own readers and it was only opened and answered by the people who care about and read the publication regularly. Anyone that doesn’t care that much about the publication, particularly those folks that get a free copy of a controlled circulation magazine, ignored the survey as well.

Social media research often has the similar problem. Instead of recruiting a cross-section of the audience, respondents are recruited through social media or their own database. No wonder the results say social media is so important, the heaviest social media users are the only people who responded!

An example I recently came across is Passle’s State of Business Blogging 2013 report, with markedly different results from the Content Marketing Institute’s 2014 Benchmarks report.

When you look at the sample audiences, they are vastly different. Which one matches the audience you are looking to understand? (For most content marketers, the answer will be CMI’s research). Hat tip to Ardath Albee for information on the Passle sample audience.

Check IconCheck that the sample and recruitment are appropriate before relying on the results.

Your Turn

It’s your turn. What would you add to this list, or what would you take off? Let me know in the comments below or on Twitter (@wittlake).

Photo Credit: Shadow Viking via Flickr, modified by author. cc

Get every post delivered directly to your inbox.

Your email address will not be shared or sold. I hate spam too.

  • Dara Schulenberg

    Great illumination and reminders Eric. We always need to consider the source of research, measure our own best practices plus benchmarks and differentiate between correlation and causation. Critical elements of context and qualitative data are vital to making research or analytics actionable in many cases.

    • http://b2bdigital.net/ Eric Wittlake

      Thanks Dara!

  • Jeff Phillips

    I can see that you tried very hard to make your case. I have no opinion about your critical thought process. My take away would be a simple caveat to measure twice and cut once. Perhaps that is what you meant to say. Have a good day and keep blogging.

    • http://b2bdigital.net/ Eric Wittlake

      LOL. The challenge with measure twice, cut once is that we often can’t measure ourselves, we have to decide which measurements from others we can rely on.

      • Jeff Phillips

        I agree – the self sees blindly and fresh perspectives can unshackle us from our comfort zone. As you say, we must decide which measurements from others we find reliable. Good point! Thanks again!

  • Pingback: The Trick to Seeing Through Marketing Statistic...()

  • Pingback: The Trick to Seeing Through Marketing Statistics | Social Savvy Business()

  • Harriet Czernobay

    A great article Eric, and thanks for referencing the Passle research. This is essentially the point we are trying to make in our blog post here – http://goo.gl/l7CSFL – that you shouldn’t take any research or statistic on face value.

    I think the danger here comes from social media, when statistics simply get shared and retweeted without reference to their original source. As the leading authority on content marketing, many people just take the Content Marketing Institute’s research as fact without considering the survey sample. Obviously by reading the research you can see that they survey their own customers – but we didn’t think this gives a truly representative picture.

    At Passle, we have in fact carried out several different pieces of research – one by trawling through websites to investigate business’ marketing efforts, across various industries and locations, and several others by surveying B2B marketers in the US using a third party. Interestingly, we found that both methods yielded almost exactly the same result.

    From my experience, it is easy to get caught up in the content “bubble” and forget that there are a lot of businesses out there to whom the concept of content marketing is completely alien. Which is why we hope our research is a refreshing break from other marketing statistics you may read! (Although of course the different results comes from our differing survey sample – neither is right or wrong.)

    Anyway, a good post highlighting the potential marketing statistics have to mislead. I hope many will bear these points in mind when reading statistics in future!

    • http://b2bdigital.net/ Eric Wittlake

      Thanks Harriet!

  • Pingback: The Trick to Seeing Through Marketing Statistic...()

  • http://www.liveconversations.biz/ Robert Lesser

    A couple of other observations::

    - Sample Size – many ‘research’ studies do not call out the sample size. A small sample size may be helpful for observations but may not be representative of the market. A lot of research doesn’t call out the sample size

    - Correlation vs. Causality – some studies imply that because two findings occurred together, that one caused the other. This result could be random or biased.

    Here is a good example from Velocify how to speak to both points by providing context to readers: http://www.marketingcharts.com/wp/traditional/lead-distribution-use-of-automated-options-linked-to-higher-conversion-rates-37073/

    Statistics may make for great headlines, infographics and tweets but many folks may be unwittingly amplifying poor or mis-leading research.

    Thanks Eric for blogging on this.

    Robert

    • http://b2bdigital.net/ Eric Wittlake

      Robert, great additions, definitely key considerations when looking at any data.

      Favorite example: do Google+ +1′s REALLY drive ranking, as some correlation results have been misconstrued to say? Based on the only causality test I have seen, the answer is clearly no.

      Thanks for taking the time to comment, I appreciate it!

  • Pingback: A/B, GE, P2P & B2B - Top Stories October 25-31, 2013 | B2BGuides.co.uk()

  • Pingback: » Alcuni consigli per l’analisi di statistiche marketing()

  • Pingback: Luigi Canali, progetti di comunicazione & marketing » Blog Archive » Cinque consigli utili per estrarre i dati reali dalle statistiche()