Attribution Modeling ••• “We Don’t Need It” (You’re Fired)

One of our attribution modeling clients had a situation: the marketing manager left, and business intelligence person also left.  No one knew if they were fired, went to a competitor, or jumped to Jamaica.

The new person showed up.  Fairly senior, and in fact the, new boss.

He called to fire us.  This happens every so often for a variety of reasons:  marketing agency changes; new management; merger, etc.   It’s one of those things that happens with the marketing industry and we part with grace.

But the new person was the boss.  Well educated, strategic, manages up (senior executives) well, but not someone in the proverbial digital engine room.  The details of a view-through pixel, and the difference between an SSP and a DSP aren’t things a boss would know.  But that’s the definition of a boss:   steers the ship, versus manning the engine room.

But for a brief period, this boss had to man the engine room as the engine room operators left, so it was a unique opportunity.

The conversation started:  “I know Pete and Blake are gone.  But we have everything we need from our own internal CRM system and Google Analytics, so I don’t think we need this.”

Attribution Modeling veep from C3 Metrics says: “totally understand.  That must be some kind of internal CRM system you’ve built.  Could you tell me more about it?”

“Well there’s not much to tell, it’s our own CRM system.  We also import some data from GA.”

Attribution Modeling veep from C3 Metrics asks: “If you don’t mind me asking, how does your internal system do viewability?  We’ve developed our viewability solution in 2009 and  continue to invest millions of dollars perfecting, and scaling viewability to accommodate massive simultaneous load and calculation…how did you build your viewability solution?”

“Uh, well…it’s not like that, it just measures clicks from display.  I didn’t know that’s what C3 Metrics did.”

Attribution Modeling veep from C3 Metrics says, “Since the click-through rates on display ads are infinitesimal that’s probably not a great method to see the impact of display.  I never click on a banner, I just open a new tab and search.  The flip side is that about 69% of programmatic display ads never show up in the consumer’s browser viewport…meaning they don’t even show up.  Imagine if you ordered 10 truckloads of something and seven trucks never showed up.  Seven out of ten trucks missing all the time is a hard way to run a business.”

Attribution Modeling veep from C3 Metrics asks: “Can I know your internal system performs Cross-Device?  67% of funnels today have multiple devices, and that’s a big data hole if you’re not joining disparate devices to the same conversion funnel.  So how are you doing that?”

“Well our internal CRM system doesn’t do that.  I didn’t know  C3 Metrics did that.”

Attribution Modeling veep from C3 Metrics asks: “In your internal CRM system, how many slots or fields do you have for converter?”

“Well just one.  One row, one field, that’s the converter.”

Attribution Modeling veep from C3 Metrics asks:  “Are you logged in right now to the dashboard showing your data, I’d like to show you something.”

“Sure, where should I go?”

Just select the ‘last week’ date range, and on the left hand navigation, click the Touchpoint Length Report.

C3 Metrics veep says:  “Do you see that histogram and the data underneath it?  Those are all your conversions this past week having 2,3,4,5,6,7,8,9,10  touchpoints and more.  See, there’s even one conversion funnel with 56 touchpoints even after we control for viewability and a few other things.  But just for grins, let’s see what percentage of your conversions have five or more touchpoints.  So 42% of your transactions have 5+ touchpoints.”

“Wow, this is the first time I’ve seen this kind of data.”

Attribution Modeling veep from C3 Metrics asks: “So now that we know 42% of the time, your funnels have 5+ touchpoints after controlling for viewability and more…how does your internal CRM system fractionalize the credit if there’s only one slot/field for  the ‘converter’?”

“What do you mean, fractionalize?”

Attribution Modeling veep from C3 Metrics says:  “Well, like a football or basketball team…we know it takes a team of touchpoints to get a transactioin.  In your case 42% of the time there are 5+ touchpoints on that team.  How does your internal CRM system fractionalize the credit among the team members, and why is there just one slot for converter?”

“Well, we don’t actually fractionalize anything.  I’m not sure how it’s decided, and I’m not sure why there’s only one slot there.”

Veep from C3 Metrics says:  “The way we fractionalize credit among touch-points in the funnel is through machine-learning Bayesian modeling .”

“Who else uses machine-learning Bayesian modeling?”

C3 Metrics veep says:  “Have you read about Google’s Self-Driving cars? Machine-learning Bayesian modeling is what’s used in those self-driving cars.  Netflix’ original content ‘House of Cards’ with Kevin Spacey:  in order to figure out what celebrity to use in their original content and what type of content to produce, they use machine-learning Bayesian modeling.”


And Now…

Eyes were opened.  It was a rare opportunity to have a question and answer session with both people having nothing to lose.

The boss’ job is to steer the ship, not man the engine room.  But to discover how inefficient the engine is, is sometimes necessary to take a good look at the engine.

This went from firing to re-hiring.

If you think your boss should know:  forward this today.