Crashing into growth marketing: a CEO journey. Part 5 — Attribution

What does it feel like?

The CXL course has been an integral part of my life for a month now. And will be for another two months. Yes, I’ve rescheduled most of my life to accommodate the amount of time that I need to spend on this.

Notes: attribution is not about math, but people.

Is Excel a core survival skill?

I went to an art university, and nobody has ever taught me how to use excel sheets. I learned to mix audio on ProTools production software and program custom effects on Grand MA lighting boards — both extremely useful skills! But I was never introduced to spreadsheet formulas.


The main topic of the week has been attribution. Let’s be honest, I did not really know this was a marketing word before I started the course, and I had to google what it means.

  • Lots of conversions (we don’t)
  • Lots of paid marketing activities (we don’t)
  • Which activities should you invest in? (We’d love to know!)
  • In which order did they do the things?
  • How much time passed between doing those things and buying your stuff?
  • target different activities for those really likely to convert right now
  • target different activities for those not likely to convert (yet)
  • figure out how to convert in less touch points

It’s not about math, it’s about the people.

While it may sound attractive to just track and record a bunch of data on user activity, the course makes a really clear point — do NOT invest in attribution unless you are sure you have the skills and resources to actually use the data and change the way you run things based on the outcomes.

Attribution models

At the core of attribution is the model — what is the sense-making way of crediting the different attribution channels — how much should they be praised for bringing in converting customers. While some marketing activities target the beginning of the customer journey (focused on awareness) and others chime in near the end (focused on instant conversion), different models come into play.

  • Last interaction
    Whichever route the customer arrived that time they made their purchase, is their attribution channel and takes 100% credit.
  • Linear attribution
    This model gives the same value to all visits, as if they all impacted the decision equally. If the client was referred through four different channels and ended up spending a hundred euros, each channel would be credited 25. Although sometimes in use, the wise men do not recommend this as a good attribution model.
  • Time decay
    The values are distributed back from the last interaction, giving less and less of a share the further away the interactions are from conversion.
    a) static time decay — gives fixed percentages to channels in order (i.e. 3% — 7% — 20% — 30% — 40% at conversion)
    b) variable time decay — calculates percentages based on the time passed between the interactions
  • Position based, or the “bathtub model”
    This model gives most credit to the first interaction (awareness) and last interaction (conversion), while leaving a bit for the interactions in between. This aligns well with the actual human behaviour, where people put most emphasis on “the first time I learned about the brand” and “the time I actually purchased”.
  • Machine learning
    Ideally, the model should be run by machine learning, as the marketer shouldn’t get to decide what’s important. Usually this brings up the cost a lot, so it’s most likely that one of the basic models are used.

LTV strategy

One of the concepts that most struck a chord with me was lifetime value strategy. This is something we’d.. forgotten? Stopped consciously thinking about? Didn’t think to write on a wall?

CEO of Zelos Team Management — an app for gig work and communities.