From The Ground Up

Fwd: Engineering, startups and life

Why Twitter Needs To Dig Deeper

leave a comment »

Last Friday Business Insider posted an article summarizing their conversation with Bill Gurley from Benchmark Capital, which invested a cool $35 million in Twitter in 2009. Gurley explains his bullish expectations for Twitter:

“If I can isolate the people who are into mountain biking in Marin,

the ability to put ads against that is really high.”

This one sentence has a lot of complexity behind it, involving at least 3 key aspects that are technically difficult problems:

1) Profiling — Generating enough segments with rich demographics that enough brands care about. Sure “mountain biking in Marin” is popular, but are there enough mountain bikers in Marin for a big brand to care about. Brands like Pepsi/Mountain Dew are used to budgeting hundreds of thousands of dollars, if not millions of dollars, to ad campaigns that reach hundreds of thousands of people, if not millions of people. The flip side of that coin are local brands, the mom’s n’ pop’s that Groupon thrives on and that Google is going after. This market presents a different challenge involving scaling out a massive sales force to hit the sidewalks. Google uses the search term as the profile generator and Facebook offers a less targeted display-ad experience; my guess is that Twitter will fall somewhere between these two to display targeted deals or personalized content.

2) Identifying — Once enough segments have been generated the technical challenge is to find the users who match that profile. This involves storing and indexing a whole ton of data, upwards of 8TB per day and growing non-linearly. Being able to match characteristics of the segments in (1) against 140 characters of text (hardly always English) is a hard problem. See Lucene and Sphinx, although, these tools are better for search than for information retrieval — a distributed store would be more effective. I think Twitter will start by looking at each user’s profile data and use gender, location and maybe age to identify segments, but the user experience is critical here and if any ads appear to be spam Twitter could suffer a lot of blowback.

3) Targeting — I believe identification is the toughest nut to crack here. But if Twitter does do a decent job at it, they need a targeting engine that decides who receives an ad, and when and where. This system will need to integrate feedback from previous user clicks to optimize identification.

For Twitter placing ads is all about a revenue stream. But they need to be very careful in how they implement and execute on this idea. In the end Twitter’s longevity relies heavily on consumer experience.

Twitter will need to go deeper in time to build a profile for everyone, but also cast a net wide enough for each profile such that each profile has a substantial population/demographic for brands and/or local markets to reach.

From the consumer perspective Twitter is in danger of feeling stale. They have made updates to some of the views, like now being able to see user’s stream as they see it, and different views on who you follow, who you should follow, and what the people you follow are saying as a timeline. But I think there is more rich functionality that Twitter can expose to their user by engineering a more robust backend.

While Twitter’s current focus is on real-time messaging, there is a vast and unique knowledge base already built-up and should be taken advantage of.  Brands, social media monitoring and apps would benefit from better search functions such as being able to search a specific user’s tweet history (even my own).

Exposing conversations and relationships as timelines is long overdue — user’s want to see how conversations and relationships evolve over time. One of Facebook’s major drivers in engagement is this: “stalking” their friends to see who is talking to whom about what.

From a consumer perspective I think the more content Twitter can aesthetically push into each user’s stream, the more engaging it will be. This would include being able to see pictures within the stream with a single click (not leaving the page) and share video in stream as well.

Lastly, Twitter , like Google (AdSense) and Facebook (Social Plugins), are finding ways to expand their footprint across the web (Tweet button and their own URL shortener). There is a lot of money to be made in display advertising and if Twitter can tie users’ presence across the web they will have a powerful engine for highly targeted ads with an expansive reach.


Written by Girish Rao

June 24, 2011 at 11:13

Posted in Uncategorized

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: