An attribution model is born.
My first hand account — week 6.
Its been a great few weeks . This past week has been the week of rule breaking and schema building. Now that our data pipeline has a skeleton structure we begin to lay and overlay more data into our stream. This will allow us to have more diverse data in our model. Think of is as painting with more color options; you can create a much more elaborate image with more color.
As a bonified Associate Analyst my team at Corsair’s Analytics is kind enough to reassure me that all analysts experience the same the bumps I have along the way. This past week, as our model became more and more solid, my confidence in the model outpaced the models own growth. For all you Futurama fans out there, you know Zapp for being over-confident. And it always bites him in the ass. Shocking! Well, once I realized my mistake, the soft floor of being humbled caught me.
To recap up until this week:
The way our client generates revenue is when an account moves from the lead status from ‘Leads’ to ‘Closed Won’. The path looks something like this:
Those familiar with the Salesforce pipeline may be wondering why it took so long to construct our skeleton. Our goal from the jump was to create a trustworthy series of tables that whenever called upon will produce numbers that match the downstream reports. The raw data included so many anomalies that our where clauses were elaborate for each of the tables. For instance, someone in marketing creates an opportunity to test something. This will show up in the opps table, so we have to exclude any events associated with that OwnerID. The biggest hurdle however was dates, I wont get into specifics here to protect the privacy of our client.
Coming from medical research, ‘cohorts’ were something we used to signify a different medical experience. People who get pill A vs pill B, stuff like that. Outside of medicine, cohorts are time dependent! The ‘Jan. vintage of customers’, that could be one of our cohorts. Using this idea the team could finally begin the analysis!
Creating wedge charts was interesting, I’ve never used them in the past but once George Earl walked me through how you can see performance of a specific field over time, I saw the power of the wedge! It is a fantastic way to identify late blooming marketing campaigns. After the fourth month or so most advertising stops generating leads and it’s evident on a wedge. The cool thing that we noticed were some of the marketing campaigns preform poorly UNTIL four months and then they begin a normal life-cycle of generating leads. How interesting is that!
I am an associate at Corsair’s Analytics — check out our growing company here:
