Your sales dashboard shows a spike the week your new ad ran. You believe the ad worked.
Or your sales spiked after your CEO was interviewed on a local TV news segment and gave out your website address. That media coverage contributed to the new sales, right?
Maybe.
Or maybe sales always spike that time of year.
Maybe a competitor raised prices the same week.
Maybe nothing about the ad or the interview did it at all.
If you can’t rule out those other explanations, you don’t actually know what caused the spike — you’ve just noticed two things happened around the same time.
Media scholars who study effects face this exact problem.
Before they’ll credit anything with causing an effect, meaning actually producing it, not just coinciding with it, they run it through a three-part test. It’s worth borrowing for your own reporting, whether you’re crediting an ad, a press interview, or a social post.
This is a topic I’ll be discussing in the coming weeks with my Media and You students at ETSU.
The Three-Part Test for Causation
Causation means one thing actually produced another, not just that they happened together.
And proving that requires three things:
- The influencing factor (the ad, the interview, or whatever you’re crediting) has to precede the effect.
- There has to be a demonstrable relationship between the two.
- Other explanations have to be ruled out.
Here’s a simple way to see why all three matter.
Students returning to campus every fall causes the days to get shorter. After all, the timing lines up perfectly, year after year, right? But, believe it or not, students coming back to campus doesn’t cause the days to shorten. The Earth’s axis tilting away from the sun does. (I know, that’s a revelation to you.)
The claim meets the first two conditions (the “cause” precedes the effect, and the association shows up reliably) but fails the third condition in the previous list. There’s a much better explanation that wasn’t considered in this example.
That’s the trap.
Meet the first two conditions but not the third, and you’re arguing for what researchers call a “spurious conclusion,” one that sounds airtight but overlooks a better explanation.
It’s a mistake that attribution modeling makes constantly, and one worth catching before it lands on a client’s or supervisor’s desk, or shapes next quarter’s budget.
Marketing Results Are a Lot Like Weather
Let’s use a simple analogy to drive this home.
The influence media has on audiences is similar to weather. Weather is always there, but it takes many forms. It’s very difficult to predict with precision, because the factors behind it are large in number and their interaction is complex. Supercomputers can tell us how much rainfall a region will get this year, but not who gets wet on which specific day.
The media work the same way on you and your customers.
Powerful computers, and in your case, marketing platforms and attribution tools, can tell us that certain kinds of messages lead to certain kinds of behavior in general. But they can’t tell you with precision which individual customer bought because of which individual touchpoint.
And just like you can’t control the weather but can carry an umbrella, you can’t control every factor influencing your audience.
But you can control how carefully you interpret what moved them.
Variables & Factors
Two more distinctions matter here.
Researchers separate attribute variables (fixed characteristics like demographics) from active variables (states that can shift), and separate fixed factors (traits a person cannot change) from changeable factors (patterns they can change).
The research is more useful when the variables are active and changeable. A demographic like age is only a rough stand-in for what’s actually driving behavior, so what people are doing and the context they’re in tells you more than who they are on paper.
Content: It’s a Factor
Content itself is a factor, not just a delivery vehicle. For example, take the context surrounding violence in a message.
Whether victims are shown suffering and perpetrators are shown being punished is largely what determines whether exposure translates into aggressive behavior afterward.
The same principle applies more broadly to your own messaging.
How something is framed, not just what it says, is part of what determines the effect it actually has.
Same offer, different framing, different result. A discount pitched as a thank-you to loyal customers lands differently than the identical discount pitched as a blowout to clear stock.
How Researchers Rule Out the Other Explanations
That third condition is the hardest one to meet. So how do researchers do it?
In an experiment, they split people into at least two groups. One group, the treatment group, sees the message being tested. The other, the control group, sees a version without it.
If far more people in the treatment group show the effect, the message likely caused it.
You can borrow the same idea. Compare the people, places or time periods that saw your ad or heard your CEO’s interview with those that didn’t.
Ran a promotion in one market but not another? Sent an offer to one email list but not a similar one? If sales jumped in both, your promotion probably isn’t the reason.
What This Means for Your Reporting
Next time an attribution model or a dashboard credits a single touchpoint for a result, check it against all three conditions, not just the first two.
Did the factor you’re crediting actually come before the change? Is there a demonstrable relationship? And have you ruled out other explanations, like seasonality, a competitor’s misstep, or plain coincidence?
Marketing attribution can overlook that third step. It’s easy to miss and costly to get wrong.
3 Big Takeaways
- Causation means one thing actually produced another, not just that they happened together. It requires three things: the influencing factor has to precede the effect, there’s a demonstrable relationship, and other explanations have been ruled out. Many “the campaign worked” claims only clear the first two.
- Marketing results, like weather, can be predicted in general but not with precision for any one person. Don’t mistake a broad pattern for proof about an individual customer.
- Focus on what you can change — behavior, context and framing — not on fixed traits like demographics.
Do you know what caused your last promotional campaign to work? Do share!
Stay authentic — and don’t credit a tactic until you can authentically tie it to causation.
Jeffery E. Pizzino, APR (seen here in a vintage photo circa 1983 serendipitously doing a Clash impersonation in a since-forgotten location) is a spin-free public relations pro who is passionate about telling the why of your story with clarity, impact and authenticity. He began his PR career in 1987 at Ketchum Public Relations in New York City but has spent the majority of his career as a solopreneur. He’s the Chief Authentic Officer of the Johnson City, TN-based public relations firm, AuthenticityPR. He also functions as the fractional CCO for his clients.
Jeff has an MBA in Management from Western International University and a Bachelor of Arts degree in Communications — with an emphasis in PR — from Brigham Young University (rise and shout!). He’s also accredited in public relations (APR) and an adjunct media and communication faculty at East Tennessee State University. This Milwaukee, Wisconsin native holds an Italian citizenship and plans to live and work there someday. Jeff and his storyteller wife Leticia have four children and five grandchildren. In his extremely limited nonwork hours, he studies Italiano, practices guitar, writes songs, gardens, works out, plays pickleball, disc golfs, reads, listens to New Wave music, serves as an assistant communication director in his church, watches BYU football, watches the original Mission Impossible TV series, and plays board games (mostly Dominion and Seven Wonders). No, this guy’s never bored and looking for something to do. Email Jeff.
