Commentary
Walmart Did It - Why Not Facebook?
- by Laurie Sullivan , Staff Writer @lauriesullivan, August 31, 2012
A fine-tuned on-site search engine can drive up conversions and page views. Wal-Mart Stores said its Wal-Mart Labs research and technology hub built a new search engine for
Walmart.com from the ground up. The engine relies on semantic search technology to anticipate the intent of a shopper's search, similar to Amazon.com's engine.
Through the project called Polaris, Walmart reported experiencing an increase of 20% for online product views, and between 10% and 15% increase in shoppers completing a purchase after searching for a product using the new search engine.
The technology was developed in 10 months with a little more than a dozen engineers, which makes me -- and I'm sure, others -- wonder why Facebook has not improved on its on-site search features. The engine groups related terms and phrases, rather than keywords that searchers use to describe products.
It's not clear whether Walmart developed its own patents just for the project, but the U.S. Patent and Trademark Office recently granted Walmart a search-related patent titled "Including hyperlinks in a document."
The patent describes "Techniques for including a hyperlink in a document is disclosed. A document is received via a communications interface. An entity pair is determined by a processor. The entity pair includes a concept included in a concept taxonomy and a textual representation included in the document. As output, a hyperlink is provided."
Ecommerce sites need a good search engine. If Facebook wants to move down the marketing funnel and compete as an ecommerce destination, it will need to improve on-site search.
While Polaris should help Walmart better compete online with Amazon.com, it takes more than a better on-site search engine. It requires integration of online and physical services, which as of last December the company didn't offer.
Sales associates in the store couldn't honor the less expensive online prices. The two point-of-sale systems kept separate inventory and pricing. By now, I'm pretty sure marketers understand the importance of cross-channel continuity.
Laurie Sullivan is a writer and editor for MediaPost. You can reach Laurie at lauriesullivan@gmail.com.
Laurie -
There's a huge difference in site search against a product catalog and Facebook. There's also a huge difference in searching against your own product catalog and a search product generalized enough to work against any product catalog.
To be specific, if the data is totally structured, meaning that you can view it organized in a table with columns (name, color, price etc.), you can easily search against that dataset with an open-source search product like Lucene (http://lucene.apache.org/core/). My guess is Polaris probably is in fact using Lucene. If you add a couple of data entry people to a) make sure that the structured data is up to date and b) match search queries with results (for e.g. someone searches for NFL blankets and the product is called NY Jets Blankets, you will want to make a manual pointer in one of the fields), you can get very good results. There's some scale issues but they can easily be addressed with adding more hardware.
When ecommerce companies don't want to allocate manpower to keeping the database current, the problem is a bit harder and requires a system for crawling the ecommerce site and then organizing the data into a structured form. There's companies that do that like SLI Systems but in Walmart's case, where they would be justified in maintaining the database with manpower, that approach isn't necessary.
My point is that Walmart "building their own search engine" isn't that impressive. And moreover, comparisons to Facebook aren't valid given the difference in the type of data being searched and the scale on the data side (not the request side).
But I always enjoy your articles....
Thanks,
Walmart obviously has a much higher share of its searches expressing commercial intent (per Jeffrey's comment) -- but the sheer volume of Facebook searches should mean some number have decent monetization potential. And FB could continue to innovate its search to make it more broadly applicable to its users in ways that drive more commercial intent. Will be interesting to watch.
Reaching right consumer with the right message https://sites.google.com/site/cuoirent/home Behavioral Based Social Media System for the Cable TV Market Cable has long history of failing to develop 1-1 target marketing, the Holy Grail of targeted advertising. http://tech.fortune.cnn.com/2011/01/03/the-56-billion-ad-question/ Excerpt from above link on January, 2011 Fortune.com – “Advertisers will spend $56 billion putting ads on TV this year,...The cable industry thought it would be a big opportunity too, but its efforts have fallen short. http://www.businessinsider.com/jason-kilar-here-are-my-thoughts-on-hulu-and-the-future-of-tv-2011-2 Excerpt from above link on February, 2011 Business Insider Identifies advertising market being missed by Cable TV operators In the near future, advertisers will demand the ability to target their messages to people rather than targeting their messages to TV shows as proxies for people.” The obvious alternative, with the least cost to implement is an independent Cloud CRM solution designed to cross index cable subscriber households with their corresponding social network interests. The current regulatory and privacy issues experienced by cable TV operators gathering unauthorized data from set-top boxes could be minimized, by validating subscriber and even eliminated by essentially having an opt-in plan (provided conveniently by the social media). Access along with profile and interests of households would be controlled by the subscriber’s social media platform of choice. Facebook has high consumer acceptance and could be used for household profiles, product interests, social interests, and viewing entertainment interests. There would be incentives to the subscribers to opt-in including notification and reminder of viewing favorites, Groupon type ads, and specific ads matching interests with infomercial type group discounts and urgency to buy. The current design of target marketing advertising ventures is fundamentally flawed. They focus on demographics, and fail to identify the individual behavioral current and future household interests. I would propose using a data cross indexing similar to a data warehouse project I was involved with at iN Demand. http://www.indemand.com/ . Project would involve developing a bidirectional Cloud interface program using a CRM application between the social media and MSO subscriber records and communicating behavioral marketing - business advertising, discounts, specific videos/groups, family albums – providing subscriber awareness of TV programming -- movies, products, etc. similar to Amazon and Groupon. This would make subscriber stickier and substantially reduce turnover. To paraphrase a comment I made in the CED 1999 publication about the Internet, cable TV operators need to become the new best friends with the 600 million members of social media.
Read more: http://www.mediapost.com/publications/article/182053/the-future-is-targeting.html#ixzz25X7gFJAM