Search Philosophies

While writing a blog on a subject as complex as Search, I thought it would be best to start with a simple framework with a hope of drawing more of you in.

I like these three philosophies for their simplicity, customer focus, completeness, and economic view.

I will just state them here and expand on each in the next stories.

  1. Focus on the Long Tail (Retreival)

Its important to learn from the best as you only have as much capacity to experiment and learn yourself. So, this one is straight from Google.

These queries tend to be from customers with higher expectations who are trying to stretch the engine with longer or difficult queries. I bought into this one as I think I can replace me for the most part with a clear articulation of this vision to my team.

  • Drive Search Engine Loyalty: Customers stick to Search Engines that can be there for them. While tail queries can have varied charectristics such as query length or stronger subjectiveness, the former was found to have a strong correlation to stickiness in the Web Search space [1].
  • Follow Your Customers: Search Queries w.r.t. to their volume distribution over time follow a First In First Out (FIFO) model. You low frequency Tail Queries of today become your higher frequency Head/Torso Queries a few down the line as capabilities and expectations increase.
  • Improve Head Queries Too: Intelligence/understanding of tail queries can help you solve your top queries too. Firstly, the scalable algorithms in Ranking and Retreival from tail can easily port to the head. Secondly, the intelligence from tail queries can imporve the customer experience for head queries.
  • Maximize Conversion: As queries get specific, the result set gets specific as well and results in improved search success. If you can answer specific queries and make your customers as them, its good for business.

2. Be Economical With Your Customers’ Attention (Sorting -> Optimization)

Customers have limited time and attention span. Intuitvely, a Search Engine Results Page (SERP) that is optimized to help them find their answers with the least amount of effort (scrolls, eye scans, clicks, cognitive load, etc.,) should reduce abandonments and improve search success.

While there are multiple studies that talk to this, one study from Microsoft [2] on search behavior is pertinent to this point. A majority of search customers tend to be ‘Economic’ (p<0.01) i.e. they make their decisions after scanning a few results.

In practice, how that optimization is done could really depend on the use case at hand. Three major tactics stood out for me,

  1. Representative Top ‘n’: If there are specific clusters of relevance to users among the thousands of results at hand, creating a Top ‘n’ that is representative might just make your user life easier. An excellent example that comes to mind is Google Flights.

2. Hetrogenous Results: Blending hetrogenous results such as Refinements or Guiding Tools with Search Results is a time tested strategy. Tools become even more cricitcal in domains where customers are thinking broad (shorter search queries). Pinterest is a nice example where the refinements play well with the Search Results to take customers close to their answer.

3. Result Set Diversification: In a SERP for ‘belts’, even if the top selling 20 belts are leather, the marginal utility of showing another ‘leather belt’ will become lower than showing a ‘canvas belt’ at somepoint before 20 porducts.

3. Speed Matters

This is really more like a ‘don’t forget about me’. Responsiveness has been shown to have a positive effect on both click throughs and subsequent searches. The effectiveness of this lever though varies greatly based on your current state — anywhere from critical to negligible.

Here are results from one experiment [3] to help understand the effectivenss of and variability in this lever.


  1. Modeling Long-Term Search Engine Usage []
  2. Individual Differences in Gaze Patterns for Web Search []
  3. Speed Matters []