modeling
| | |

How I Made Attribution Modeling Simple Enough to Sell to Your Boss

Stuck in a one-click world? Can’t justify your head terms, display buy, content-based email campaign, or any other top of funnel campaign? Trust me, you’re not the only one. I’ve been through this. A lot. But I have also overcome this a lot. What I learned a long time ago is that attribution is a complex topic, especially for people not in the day-to-day execution of marketing campaigns.  I wanted to share my experiences to help other marketers overcome the challenges of selling through attribution modeling.

In my past, I have typically run into two scenarios where attribution needs to be a discussion point. One much worse than the other.

Scenario 1: Marketing tracking & reporting coming from multiple sources

Think about it this way: PPC is being reported from Google Ads (formerly Adwords) or a bid manager, email by the ESP (email service provider) the company uses, affiliate by the network, and organic search from Google Analytics. No single source. This causes massive duplication with multiple sources reporting the same order or conversion.

Example – Marketing Drove More Sales Than Actual Sales…Only with Faulty Reporting

This issue came to a head at an agency I worked at years ago. The agency had a client that sold sports merchandise and our team managed Paid Search, Email, Affiliates and Comparison Shopping Engines (CSE). It was a November monthly recap report that was the tipping point. Our team was really excited by how much revenue was driven through our marketing channels until they reviewed the monthly report with the client and numbers were something like this:

Paid Search: $1,000,000

Email: $1,500,000

Affiliates: $750,000

CSE: $500,000

Organic: $1,000,000

Total: $4,750,000

Big numbers, right? Happy team? Yup. Then the client got on the call and told the team they sold $4,250,000 in merchandise…whoops. Yeah, the agency reported that marketing was responsible for more revenue than the site drove. As you can imagine, a multi-channel, single-source platform was needed to avoid this misrepresentation in the future.  Not the point of the article, but for those that are curious, it took over a year to find a software solution as this occurred in the early days of dedicated solutions. And at this time Google Analytics did not have any multi-channel reporting.

Scenario 2: Marketing is already a single source but also single click

It’s still mostly a Last Click world (likely GA, Adobe/Omniture, or Coremetrics) and occasionally First Click (looking at you, Hubspot). You know there is a smarter way but you need executive buy in to change the way you are doing things. Your boss is pushing back because:

  • He/she doesn’t fully get it
  • It would change reporting and we won’t be able to comp year over year
  • Fear of change
  • He/she doesn’t see the benefit
  • He/she doesn’t play well with the related tool (in my case it was Salesforce)

I’ve personally encountered both scenarios on multiple occasions. Fortunately the first, more extreme, the scenario is one I rarely come across anymore. But how did we handle it back then?  

Ecommerce Client Over-Reporting

The first was to show the client what was happening. To do that, we pulled order IDs for each transaction out of all the independent systems. We made a two-column spreadsheet, one column had the order ID and the other listed the channel. With a quick pivot table, we were able to show the client how their customers were consistently hitting multiple touchpoints before they bought.

After we simplified the story so that they understood the problem, the client was gung-ho to find a solution with us. This particular example was so extreme that it was an easy sell.  

Simplify Your Message to Achieve Complex Goals

My second scenario is where I see most marketers today: a single source doing their reporting, typically one of the big three web analytics platforms. The Marketer wants to analyze their campaigns’ performances in a smarter way but the Marketing Director (V.P., CMO) has an “if it’s not broke, don’t fix it” mentality.

I was in this situation myself while working client-side. I had pitched moving our reporting to an attributed model and was promptly shut down for all the reasons mentioned above. Fortunately, we reached a tipping point when my boss got an email from a board member and we had a conversation that went like this:

Boss: Hi Keith, Mr. Random Boardmember just emailed and wants to know why we aren’t running PPC ads when he searches for [main industry head term].

Me: We have run it in the past, but people rarely convert on that term. This is a term usually used when they are just starting their search and we can’t maintain our ROI goal while running it.

Boss: Aren’t we missing out on a lot of potential sales?

Me: Yup.

Boss: How do we fix it?

Me: Remember when I showed you that stuff about attribution?

Boss: Too complicated, people won’t understand it.

Me: If we don’t give credit as appropriate to top-funnel keywords, we’ll get reamed out by the board for poor performance instead of not bidding on a certain keyword.

Boss: What about First Click?

Me: No.

Boss: You could think about first…

Me: We just got our retargeting campaigns performing well. If we go the First Click, retargeting will have a zero for ROI – by definition, it can’t be the First Click…

 Lightbulb moment! For the boss and for me.

I had pitched a really meaty deck showing all the features and advanced models that were available. I took a complex topic and made it complex – I was Calvin in the strip above, unknowingly making his parents’ life complicated. In the one minute conversation – I was able to take the complex topic and make it simple.  

Those were examples of the first step in two different cases. Step 2 came with its own set of challenges. I had the buy-in on the concept of attribution. Now we had to agree on a model.  

Back to Basics – Attribution Modeling

As a refresher, here is a quick review of the 5 most common attribution models.

 

Last Click – the value assigned to conversion is given to the channel which had the most recent click prior to conversion.

 

 

First Click – the value assigned to conversion is given to the channel which had the least recent click prior to conversion.

 

 

Linear (Even) – the value assigned to conversion is evenly spread across the series of channels that led to a conversion.


 

Time Decay – the value assigned to conversion is weighted to give more value to the channels that were more recent to the conversion.

 

 

Position Based (U-shaped) – more weight is given to the first and last channels in the path, while the remaining credit is divided amongst the middle channels.

 

 

There are more than just these five with new custom and algorithmic models popping up quite a bit recently. But if you are trying to get a company to dip their toe in the attribution waters, those are not likely to be your starting point. If you’ve made it this far, there is no point in discussing First and Last Click anymore. That leaves us with three options.  

[su_newsletter_email]

Pitching a model

Depending on the business, there are two models I typically recommend. Time Decay is a strong option when you have a long sales cycle, think six months or so, like buying a luxury car. A consumer starts the research phase well in advance of going to a dealer for a test drive, with a lot of touchpoints along the way.  

My other go-to is Position-Based. Why? It’s logical. Ok, remember when I said there was no point in talking about Last and First Click anymore? Well, I have to ask for one exception, I need them to make a point.

The last Click gives credit to the last touchpoint before a sale. As a marketer, that makes sense to me since I want to drive as many conversions as possible and report on what campaigns worked vs what didn’t. But, what if no one ever knows about our company, product, or service? How will they get to us to even make a conversion? Now we have the case for First click. Position Based takes both of these into account, assigning more credit to the First and Last click in a path and dividing up the rest in between the other touchpoints. As a digital marketer, this makes the most sense to me. The problem lies in the fact that many of us have to convince folks who are not in the digital marketing weeds that these models will allow us to make better decisions with our budgets because they allow us to look at the full picture and report on it, to secure more budget.

I’ve often found resistance to moving to these models. Things I’ve heard include:

“It’s too complicated”

“The rest of the company will think it’s marketing BS to make your team look better”

“How am I supposed to explain that to the board in one slide?”

“We’re not that big of a company, do we need the fancy stuff?

This leads us to … the compromise!

When encountering the type of resistance I just noted above, Linear/Even attribution becomes the fallback model. Simpler to understand for the resistant folks who are being forced to change.  Better than the Last Click model that most people still run. Though, like Calvin, not a compromise I’m happy with. The upside is that most situations I’ve run into where someone starts on Even attribution, within one to two years, they moved on to something more advanced.

If even attribution ends up being your compromise, you may have to repeat this process to get where you really want to be.  So be like Homer with hair and always repeat.

Takeaways

While attribution is a complex topic and can be a difficult sell to get a company to change models, my experiences have shown me a few things.  First, while complicated, there are ways to simplify the topic to get buy-in.  Second, picking a model can be an iterative approach. In my experience, even attribution is an easier sell than more complex models.  Making the compromise and settling on a model you may not prefer is better than sticking with a fully outdated approach.

Similar Posts