Correlation does not equal causation!

by 30 replies
38
With so much emphasis on testing within IM, how much emphasis does this little observation have? Does it even play a role?

I've seen attempts to associate sales with colors, site design, content length, mind control, and moon position. But does anyone take into account the possibility of no association at all? And that, perhaps, moving a web element ten pixels higher really has nothing to do with a new sale?

#justthinkingoutloud
#main internet marketing discussion forum #causation #correlation #equal
  • Good question.
    • [1] reply
    • I never studied that, but you bring up a good question; something to think about.

      I tried everything under the sun to increase free traffic to my blog - all to no avail. In fact, even after hundreds of compliments on my blog design, look & feel, and its content, including an indie book publisher inviting me as a resident blogger on his site, Google SLAMMED my blog during the "quality content" algorithm update.

      I haven't had more than 15 visitors/day since then. Fortunately for me, this same indie publisher is correlating all of my (according to Google, less-than-quality) content into a book and I'm shutting down the blog.

      Meanwhile, the two blogs I've hardly worked on all year, the traffic INCREASED after that algorithm!!!

      Personally, I think your content has more to do with the setup of your site. As long as you keep it clean, minimalized and don't have mass quantities of ads all over the place, 'oh, the colors' may not make much of a difference. In my opinion, it's all subjective. But that's just my opinion.
  • Correlation is a quantifiable statistical concept. Correlation without causation is termed "spurious correlation". When statisticians come across situations like these, they are advised to use common sense in determining causation. Which is pretty much what people need to do with optimizing their websites.

    The only way to figure things out is to test different ideas. Sometimes traffic may increase simply because more people are searching for one of the many keywords on your site. You may attribute this traffic increase to a slight change you made in design, which is not true.
    Using common sense is the only way to get past this issue.
    • [1] reply
    • It's the scientific method that rules in testing.

      Testing is not had with one sale - one result is no indicator of anything.
      If you have 100 results, then you can determine that the single difference between A and B is the cause - what is there to correlate?

      Traffic is volume. Volume doesn't increase the rate of sales per 1000 visitors. Volume increases the amount of sales which, given the same elements, is determined then by the rate of sales per 1000 visitors (or whatever number you have to use to establish a rate). Even still, this number can correct itself with higher volumes, upward or downward.

      The greater the sample (1,000 - 10,000 - 100,000) the more consistent the resulting rate will be. The smaller the sample (1 - 10 - 100) the less consistent (accurate) your rate will be.

      Should testing be done for your desperate attempts at making money online? Then, in that case, you needn't waste your time. Because, what does correlate is the fact that, you do not want to invest the time and effort necessary to build up a long-term income flow which causes you to not do your research, to not verify market data, to not test every part of your sales funnel, and finally to feel that testing matters not (meaning that science and mathematics are irrelevant).

      After all, you can throw your money at paid traffic and fan your sales page - sure, you'll get sales. The issue, is that you could get more sales if your sales funnel was tested and optimized. Yes, you will be happy with $200 a month from your traffic, but someone else is making $2,000 using the same traffic that you are.

      This is the only correlation in IM.
      • [ 2 ] Thanks
  • This is a question of liniar and non-liniar systems.

    In a liniar system, it is easy to predict how A will determine B. 1 + 1 = 2.

    However, in a non liniar system, 1 + 1 = 3151

    As humans we tend to think in liniar form. For example, I write you a sales copy, you can already think about the money coming in. This is a liniar system.

    However, a non-liniar system is the fact that the copy will interact with your traffic, product, market conditions, etc therefore the result will be more unpredictable.

    PPC tends to be a rather liniar system. SEO tends to be non-liniar.

    In a liniar system input equals output - loss. In a non-liniar system the output can be very different to the input doing to other systems interacting with it.

    The human body is another good example. Every organ serves a purpose. Your lungs helps you breathe. Your hearth keeps the blood traveling through your body.Taken aside, they are all liniar system, causal systems where A -> B. But in the moment they interact with each other, a small change in any of the organs can cause a catastrophic change in the others.

    Hope that makes sense.

    And to answer your question directly - if you test everything in a vacuum, in theory, you can see how it affects your site. However, in the moment you test everything as a set of systems that are interconnected, then the results are not predictable anymore or at least not as predictable as when testing a single thing.
    • [ 1 ] Thanks
    • [1] reply
    • The testing advised by many in IM is "live" testing - sending traffic to your landing page and measuring the results of version "A" and version "B".
      • [1] reply
  • There is no one cause behind any sale, only the culmination of all the different factos playing in the equation.

    For example: You want some to equal 9 to make sale. 1+1+2+3+1+1= 9
    If you remove any of the numbers you end up with 8 which is not a sale. Kind of like a prospect is so close to buying but is not actually doing it because he need that extra push.

    Now imagine a squeeze page or a website as a gigantic equation with thousands of numbers. Most times the reason a prospect does not want to buy is because they want everything to look perfect, and as soon a one thing doesn't, they have an excuse to not sign up for a newsletter or buy a product.

    Now, a prospect may not notice the small imperfection (10 pixels up or down), but why risk it?

    PS: It's the little things that matter

    - Kabbas
    • [ 1 ] Thanks
  • Back in the internet early years... my boss told me that my BIG red button was just too garish, and that he wanted a smaller one. He created a small pale little button that he said looked more pro. Well, our sales were cut in half and he was screaming - put your button back in. Funny how customers will not click on what they do not see. Do not be subtle when SELLING is your purpose. Most of our customers did not realize that our products were on sale unless we told them in big red all cap letters -
    • [1] reply
    • Marketing as a form of communication is not a science, but an art. And, actually, if we need to apply stats to a system the available math can't deal with it in real scientific terms (as it happens in medicine by the way).

      So... adding intuition to correlations and an ability to feel the ever changing pulse of the market is what works.
      • [ 1 ] Thanks
      • [1] reply
  • I used to have a fishing buddy who, on one particular trip, was having no luck at all. Everyone else on the boat was catching fish, but he was still looking at the skunk. After watching him try pretty much everything in his tackle box, the captain handed him one of the boat's standard rigs, and told him his problem was that he was holding his mouth wrong. My buddy was instructed to simply hold the rod, while twisting his mouth into some kind of psycho grin.

    You guessed it. On the very first drop, he caught a nice fish. Then another, and another.

    Long after that trip, he insisted that the only way he could catch fish was if he twisted his face into the same grin.

    While the grin, in and of itself, didn't cause my friend to catch fish, it did correlate, at least for that trip. Looking deeper, what the grin actually did was get his mind off of what he would try next and gave him something to concentrate on.

    One day, he dropped his hook and held it still, but he forgot to grin. Still caught fish. Finally realized that the cause was holding the bait in front of the fish, rather than constantly yanking it away to change something.

    Too many marketers waste time and resources testing trivia. Much better to go for the big win by testing big changes - different headlines, different offers, different value propositions.

    One more fishing story...

    Back in the late sixties and early seventies, the Lindner brothers built their company and their brand by barnstorming around the midwest, catching limits of large fish from "fished out" lakes. They knew that the lake wasn't really fished out, anglers had simply removed that part of the population that favored certain spots and methods. By using new methods in new areas of the lake, they tapped fish populations that hadn't been tapped before.

    Online, that means keep testing big things - new traffic sources, new content types, and so on.
    • [ 2 ] Thanks
  • When a credible person says changing the color of So-and-So button to red will show improvement, or move a box slightly to the left/right to see changes, they say it because they've probably tested its ability to deliver consistent results.
    • [1] reply
    • They've likely tested its ability to deliver results with their combination of offers and visitors. There's no guarantee you will see the same results.

      As the car makers like to say, "your mileage may vary."

      And the stock sellers, "past performance does not guarantee future results."

      While you didn't say it, a lot of marketers follow the monkey see monkey do method. They see a gaggle of IM gooroos do something, and assume they're seeing the winning result of testing. Given that for every A/B test, half of the people see the losing result, that can be a very dangerous assumption.

      Don't just assume something will work. Try it. Then try to beat it. If you can't, great. If you can, also great. But you never know if you don't try.
      • [ 2 ] Thanks
  • hi i am shamiur.

  • My experience with IMers suggests there is very little real emphasis on testing within the industry.

    Sure, lots of people talk about it and suggest it because it makes perfect sense, but I doubt that very many IMers actually even attempt to test anything in their marketing except maybe a few different traffic sources.

    Most newcomers to IM try something and if it doesn't work out, they are off trying something else completely different.

    For effective but simple marketing tests to be worthwhile (valid), you must keep all variables constant except for the one being tested . . . and you must have a statistically valid sample size.

    How many marketers really understand the whole test - track - tweak process? Most don't . . . so they don't take advantage of it.

    If you don't think testing is important, what better way is there to optimize a marketing campaign? Or are you saying it's just best to let the chips fall where they may?

    Steve
    • [ 2 ] Thanks
    • [1] reply
    • What's the probability of a test's success based upon empirical data compared with the probability of success of a test based upon nothing?

      Whatever your choice is, Murphy still prevails--"If anything could go wrong, it will." And it can.
  • Yes, they do it all the time. Most IMers prove that not changing anything at all on their Sales Funnels results in no sales!
    • [1] reply
    • But correlation certainly does imply causation, particularly when this involves large numbers, or statistically consistent data. This is a basic foundation of the scientific method (and BTW in the professions of marketing and copy writing as well), which often is the basis for an underlying hypothesis wherein every possible causative relationship is systematically explored and tested. To ignore correlation as if it does not suggest causation at all can be just as fallacious as assuming causation. Small changes in the nuances of marketing (ie wording, colors, content length, grammar, style, etc) can and do correlate with causation - often with major consequences.
      • [1] reply
  • [DELETED]
    • [ 1 ] Thanks
    • [1] reply
    • Thanks

      I know. I don't engage because I believe the persona. I do it for the sake of others.

      Alexa is fun though, and real enough. In the past there was one lady even funnier, Laura Miller was the name. I did the tech part for that project.

      Agree with the P.S.
      • [ 2 ] Thanks
  • One thing that people don't often highlight is the need to be careful what variables you test. A variable can be related to another variable but not causally.

    For example sales of woolly mittens (variable a) no of snow men built (variable b). Statistically you could find a correlation between to the two. This could lead to the false assumption that to sell more mittens we should build more snow men.

    The two aren't causally linked. The causal link here would be the amount of snow.

    I'd assume in my hypothetical that their would also be a point of elasticity as well.

    Another thing that you see in real marketing is to get results validated by retesting and peer review much like in science. When was the last time an IMer did this?

    Instead its a jump to this happened because I did this and for £1997 I will sell it to you.
    • [ 1 ] Thanks
    • [1] reply
    • As I recall, that so-called hypothetical person behind the MYOB handle has never made a claim to making $500 million a year as an affiliate, despite repeated assertions. Check the original sources again, and see if they match up with your hypothesis.

      Wow, with such rigorous critical thinking and analytical skills, there is a chance for you to become a real scientist, just like Sandra was.

      This is known widely among specifically targeted audiences as the MYOB method.
Join the discussion

Next Topics on Trending Feed

  • 38

    Correlation does not equal causation!

    With so much emphasis on testing within IM, how much emphasis does this little observation have? Does it even play a role? I've seen attempts to associate sales with colors, site design, content length, mind control, and moon position. But does anyone take into account the possibility of no association at all? And that, perhaps, moving a web element ten pixels higher really has nothing to do with a new sale?