Today, I'm revealing how a common oversight in campaign analysis is costing eCommerce brands millions—by neglecting the 'long tail' effect, leading to overestimated marketing ROI and misguided strategies.

Don't Underestimate the Longtail

Imagine this...

Your brand sends a flyer + coupon code by post to 50,000 customers whose next order is overdue.

You have considered everything from the customer segment to the ideal timing and the perfect offer.

Two weeks later, you measure the results:

1,000 customers (2%) reactivate their accounts, generating $100,000 in net revenue.

The question now is, is this a good result?

The campaign itself cost you $30,000 including the coupon code.

Looks good at first glance.

❗ But you haven't considered the long tail?

How many customers would have ordered in the same time period without the reactivation, even though they are overdue?

Even if the probability per customer is small, the sum of the long tail is often very large. Assuming that 1.5% of these customers would have purchased again anyway is even conservative.

This completely changes the performance of the campaign:
Instead of $100,000 in sales, we end up with only $25,000, making the entire campaign negative.

Now, you might say, "We don't send offline campaigns"

But the same effect can be applied to many other things:

Example: Email Marketing

You send a campaign to your mailing list. Klaviyo shows you an attributed revenue of $50,000 for this campaign.

The real effect of this campaign would be calculated as follows:

Since all customers received the campaigns, the first question is how many of these customers would have ordered anyway without the newsletter:

Let's assume that the share of revenue by existing customers is 50% of the daily sales (on days without a campaign).

Your daily sales are $90,000. So the sales directly attributable to the newsletter would be $50,000 - (50%*$90,000), or only $5,000.

That's a huge difference, especially when it comes to channel ROI.

It would be easy for your email provider to correctly allocate the revenue, but the truth is that they have no interest in doing so because they want to claim the maximum potential revenue for their channel.

What can we learn from this?

If we want to evaluate the effectiveness of a measure, we must first calculate the long tail that occurs anyway (and is often forgotten).

Some examples of questions to ask & validate against the longtail:

• Is a particular coupon campaign really driving incremental sales?

• How successful is my new collection or product launch?

• What is the true ROI of my marketing channel?

We know the classic application of this method from TV advertising. Since direct tracking is hardly possible, you first have to predict the background noise at the time of broadcast.

Only the delta to this background noise can be attributed to the campaign itself.

This principle must be applied to all measures where results are achieved without further action.

The price of gold has also been rising since I was born in 1986, but I have nothing to do with it 😉

That’s it for this week!