Twitter  The lessons @jimbradysp learned from Project Thunderdome nie.mn/1thZ8XD via @NetNewsCheck  
Nieman Journalism Lab
Pushing to the future of journalism — A project of the Nieman Foundation at Harvard

How a shift in perspective salvaged Boston.com’s local search project

In 2006, Boston.com launched a local search tool that was supposed to be a big part of the site’s future. The project made perfect sense on paper: Readers would get search results focused on eastern Massachusetts. Those results would mix the best of the machine and human worlds by using algorithms and editors’ picks. Next to the results would be targeted advertising, opening up a lucrative revenue stream. And Boston.com would expand its audience with a useful new service.

Or so the thinking went.

The reality is that Boston.com’s local search never caught on. Traffic lifted a little after launch, but then it plateaued. “It’s been a flat line almost since we started in terms of use,” said Bob Kempf, vice president of product and technology at Boston.com. “It hasn’t really grown.”

Kempf and his team poked at the problem for a year, but an assortment of tweaks didn’t give local search the lift they needed. They eventually reached a diagnosis: Local search was fighting a losing battle against the audience’s expectation of what Boston.com could be.

“We’ve done so well over the last 14 years as a news and information site,” Kempf said. “That’s what people are accustomed to getting from us.”

The aha moment

Those disappointing numbers didn’t lead Boston.com to abandon local search — indeed, it’s still the default option for any search query on the site. The project was just viewed with much lower expectations.

Moving on might have been an option had local search been a low-cost experiment. But Kempf said the final tally ran into six figures. The project was too expensive to just write off. There needed to be some kind of return.

The product team was posed with a dilemma: How do you get value out of something that missed the mark? Was there something else search could do?

Turns out there was. The “aha” moment, Kempf said, came when they stopped looking at search as a product and started seeing it as a platform.

That’s a fine distinction, and it’s hard to understand if you don’t spend your days amidst queries, databases and RSS feeds. So here’s the difference: A search product is built around user intent. You put your cursor in a search box and type in keywords related to whatever it is you’re seeking. The database then spits out results.

A search platform removes intent from the equation. You don’t input keywords. A page of results doesn’t appear. Rather, the underlying search engine serves up related content within a web page. The Globe SportsWire section on this page is an example. The search platform is used to find, assemble, and publish targeted content around the core information.

The shift from search product to search platform required acknowledging the site’s strengths and weaknesses — and accepting that it was unlikely to convince users to turn to Boston.com instead of Google for search. Newspaper sites can’t — and shouldn’t — fight that notion. But it meant acknowledging that a user will visit Boston.com for trustworthy local information. The product team realized it could use Boston.com’s big, robust search platform to deepen the relationship with that visitor. And better still, users wouldn’t need to change their behavior to get the benefit. This shift from product to platform gave search a purpose and solved the tricky issue of audience expectation.

The growth issue

There was still a problem. The big investment in search needed to be recouped, and the best way to do that was to grow the audience. Local search hadn’t cracked the audience-growth code the first time around.

Perhaps the mental gymnastics the product team ran through when it reimagined local search acclimated it toward thinking differently. Or maybe it just got lucky. Either way, a second aha moment came when the team started development on Boston.com’s hyperlocal sites.

This was a project aimed at growing that all-important audience. To succeed, each hyperlocal site needed to feature lots of stories targeting specific Boston Boston-area neighborhoods. Resources were limited, though. The company wasn’t going to hire a battalion of stringers and producers. To efficiently fill out those new hyperlocal sections, staffers would need to gather all of Boston.com’s content in one big database. Targeted articles would then be pushed out to neighborhood sections.

Database? Targeted content? That’s search, and that’s something Boston.com already had. The search platform in place had more than enough mojo to power the hyperlocal sites. All it took were a few customizations.

Take a look at these pages. Much of the information you see is served up by something that, according to original projections, failed in its first incarnation.

As for that big investment, Kempf believes the growth opportunities the search platform creates — both those in development and others not even conceived yet — will offset the cost in the long run.

“Just because it didn’t work the way we thought it would originally didn’t mean there wasn’t some value there,” Kempf said. “We just had to discover what it was.”

                                   
What to read next
lippmann-house-990
Ann Marie Lipinski    July 24, 2014
The Nieman Foundation for Journalism at Harvard wants to hear your idea for making journalism better. Come spend a few weeks working on it in Cambridge.
  • http://www.universalhub.com Adam Gaffin

    Very interesting post, thanks! One minor thing: Your Town doesn’t cover “Boston neighborhoods.” They have sites for lots of surrounding communities, but it’s fascinating that the Boston Globe does not have a single hyperlocal site actually in the city of Boston. [Fixed; thanks, Adam. —Ed.]

  • Pingback: Media Nation » How Boston.com got search right

  • Jim

    So they paid $100K and wound up with a CMS database? Could have built the whole site on WordPress for free.

  • http://www.sunvalleyonline.com Dave Chase

    Mac – I followed the links to the boston.com pages but it’s not obvious to me which portions of the page are generated by the search platform. Can you be more specific or circle the portions in a screenshot? It’s an intriguing article but it would help to understand what auto-generated content you are referring to. As a bonus, knowing what technology is powering it and whether it is available to other publishers would be a big help.

    Thanks!

  • http://www.niemanlab.org Mac Slocum

    @Dave: Great questions. I can make an educated guess on both accounts based on my interview with Bob, but I’m double checking. I’ll be sure to report back.

  • Jim

    Dave is essentially raising the same point as me, but with less snark. It looks like the YourTown sites and back sections are still being populated manually, which doesn’t rely on any super search algorithm that cost $100K (or more) to build.

  • http://www.niemanlab.org Mac Slocum

    Allrighty. Bob clarified how the search platform integrates with the Your Town section. Here’s how he described it:

    “… the search platform performs a ‘back office’ function as follows: the search index provides a source feed which is geo-tagged and fed to another aggregation/editing application. That last app (not part of the search platform) is what produces the content that appears on the Your Town pages.”

    I’ll also add this …

    It was beyond the scope of the article (far, *far* too techie), but Bob and I did discuss the “aggregation/editing application” he mentions. As I understand it, the interface acts as a workbench that allows Boston.com producers to pick and order stories. So while there is a manual component to the Your Town pages, the incoming content is generated by the search engine. It’s aggregation followed by curation.

    And Jim, to your point, a similar set-up could indeed be created using any sort of source database. The bigger issue — and one I think is applicable to a variety of organizations — is that *if* a big investment doesn’t pan out as intended, it’s worth considering alternative ways to put that expensive tool to use. That said, a company that’s looking to build its own search platform should of course balance features and cost. They certainly don’t have to spend a ton of money to get useful functionality.

    @Dave: the underlying search product is from FAST, which was purchased by Microsoft in 2008. The Boston.com product team added their own customizations.

  • http://www.sunvalleyonline.com Dave Chase

    Great service! Thanks for the follow-up

  • Jim

    Mac,

    Boston.com does deserve credit for re-purposing a costly idea that didn’t pay off, and I would hate to discourage any news organization from taking a risk on something new and innovative.

    However, the only thing that survived was a geotagged RSS feed. You can salvage your $40K BMW for $200 in scrap metal, but at the end of the day, you’re still out one car.

  • Pingback: How Boston.com Made Lemonade Out Of Local Search Lemons

  • Pingback: Paid Content : paidContent Quick Hits 11.30.09

  • Michael Andersen

    Neat article, Mac.

    It sounds to me as if a regional non-tech company tried to compete with a service provided by a bunch of national tech companies and found it didn’t have the scale to keep up. Boston.com search is probably better at local than Google in 2005, but Google is pretty great at local in 2009.

    Are regional newspapers making similar mistakes today? I don’t know of any. But as cash flow returns, they’ll be tempted.

  • Pingback: Living Stories can reinvent the article « Murder Manual

  • Pingback: Three ways Apple’s iAd might impact the news industry’s continued advertising woes » Nieman Journalism Lab