Demystifying Data #1: Less is More - "How to stop throwing numbers at the problem and embrace Data?"

[#1]

Oddly enough, I realized that we are never going to have enough data while I was frantically trying to stuff the day’s leftovers (the COVID Lockdown has ensured that I can’t escape my share of the household chores) into the fridge:

“Ahhh if only I had a bigger fridge”. That’s what my mom used to say.

Wait… What?

That made me realize two things: firstly, that I may not have the fridge that I deserve right now, but instead one that I need (thanks Nolan!) and second: I can make better use of what I already have!!!

I can probably guess at what some of you may be thinking: this was supposed to be a post about data and he’s just spent the first three paragraphs rambling about fridges and making oblique comic-book-movie references, report spam!

Ok I’ll stop apologizing for analogizing. The fridge is the data. We should be spending our time thinking about the data we already have.

The problem is that we often have a lot of it. Too much in fact, so much so that it overwhelms us.

How much is actually important?

Well, I usually apply something known as the Pareto Principle or the 80-20 rule. Now how do we go about applying it towards OUR data. Perhaps even more importantly: How do we leverage this in Search Engine Marketing?

Breathe. Focus. Read on towards the philosophy of Less Is More. (call it LIM if you like).

Case: “Top-level insights”

Problem: The ask is quite vague (and unfortunately quite frequent as well in the world of business), and one must balance on the edge of doing too little and risk missing out important details, or spending too much time digging without striking anything of value.

Approach: Find the Less Is More (LIM) Values i.e. the keywords that contribute majorly terms of Cost (you can do it with Revenue, Visits, or indeed any aggregated metric). I have chosen the arbitrary value of top 20%.

Note: Depending on your industry, vertical, or client positioning (i.e. leader brand, challenger etc.) our focus value may change to 50%, 80% etc.

Screenshot – Baby Steps to LIM Power (the bonus image at the bottom contains detailed setup info):

Result: Very interesting… just 2 keywords (out of 200K total keywords analyzed) consume 20% budgets! That’s 1/5th of the total cost consumed by 0.001% of the bid units.

Although this isn’t earthshaking news in SEM by any means (“…of course brand spends more yadda yadda…”), it still gives us an idea about the cost distribution, and this is something we can use that to our advantage, depending on the strategy the customer is using - fight it out with the big boys on the Head Terms, or look for efficiencies in the Long Tail.

Can we do even more with the same data? Yes. We just need different CUTS. (I’m going to be covering data cuts in more detail later in the series). We can add things like slicers for different dimensions such as Devices, Match Types, Locations, etc. to see how our Focus Terms are changing. Once we begin slicers and understanding the dimensions our data already possesses the possibilities are endless.

Questions to Think About engendered by the case:

• What are my LIM terms for just Generics?
• If I change the device, do my top terms change?
• Is my top 20% fixed or does it show seasonality?

These questions could be a good starting point to begin digging for data-driven opportunities

Superpower: The Pareto Principle is a Power Law. That means you can apply Pareto ON PARETO (look it up or comment if you need more detail). If Top 20% throws up too many terms to process, then we can look at the top 20% of that i.e. 4% of the total and find value.

---

Bonus Case: “Search Query Mining”

Problem: Mining ones Search Terms (user queries) for keywords or as it is better known – The dreaded SQR

Approach: Throw data, not numbers at your problem. Yes, there is a difference.

Questions to Think About (QTA?):

• Most expensive Search Terms:
  • Q: What are the terms that make up most of my cost? (let’s say 50% is a good place to start with)
  • A: Apply our approach of Less Is More
  • Add/Remove them according to your business str
  • Breathe a sigh of relief knowing that you didn’t miss anything of consequence

Bonus Screenshot: Only 59 out of 2295 total Generic Search Terms (that’s less than 3% btw) accounted for 50% of the total cost. Ergo: you need only sift through less than 3% to have more than a 50% impact (ok, that’s approximate).

Thank you for reading, hope it helps you reach your results faster and more efficiently. I’ll be exploring how to cut data and (perhaps more importantly) why to cut data in the coming installments of Demystifying Data so be sure to tune in for more!

Please leave a comment below if you have tips on efficient fridge management (or indeed anything else that you think can be covered).

Sayaning off!

[#2]

Very well written Sayan. Data and Fridge comparison is interesting.

[#3]

Thanks Sanjay! I was trying to make it humourous

[#4]

Interesting article @Sayan0Mukherjee. I can relate to the fridge space management part as well