adaraliao
commented
6 years ago My first impression is this is a high-value and actionable question that Microsoft and Google can benefit a lot from, so I like the commercial value of this. What I understood is that your project aims to help Sales teams increase value from existing customers by showing them which customers have the best upside which is measured by customers’ ability to scale. This is the best way to help Bing achieve 42% growth in advertising revenue.
I like your explanation of how you arrived at this actionable question for the Sales team. You went from big corporate goal ($10Bn business), explained why market share is not as easily achievable as increasing value per customer and what makes your specific strategy worthy of attention. This is good for proposing to management to invest in your method, because they like to see what the top-line impact is.
One additional point to consider is to address what is the cost to Sales team of investing time on a customer they are trying to scale-up? If there is spare capacity, then hiring more Sales team to cover net-positive customers makes sense.
On the methodology, maybe elaborate on link between traffic log files and advertiser’s scalability. Perhaps also examine what makes a company willing to invest in advertising spending. Your algorithmn seems to address this by looking at industry classification, maybe can expand to other characteristics.
All in all, this is a great business idea with big upside potential that a Sales person will find very actionable!
In general, I like the idea a lot (my company essentially does this ;)). I would add a bit more in the middle about the actual data collection. Third party agencies and other media agencies may indeed be sitting on a bed of data, but a lot of the times, they don't know how to bring the data together (for example, is this paid search conversion actually contributed to paid search rather than an online video?). The overall modeling/intelligence engine won't be too difficult to come up with if you have a good history of the impression pipeline.
Lastly, the unintended consequence of analysts losing their jobs may not be the case. It may force analysts to become more technically capable and force them to either contribute to the data lake, or have the capability to pull from the lake and do exploratory data analysis. Finally, the beginning sentence of the the end isn't capitalized, but that's just nitpicking.
On the topic of increasing paid search for Bing, understandably logs from both IE and Bing would be useful, but could their also be an incentive in place for both the advertiser and the target? Perhaps gathering price level data on adspots/locations and giving better benefits/higher prioritization may be useful?
Overall, great idea and great article. I'm excited to see how you'd implement.