Newspaper site search is broken.It’s never been anything but broken.
Search is the primary way people navigate the web, and the way a lot of people (about 50 percent) do or want to navigate any given web site.
With all due respect to my friends at Planet Discover, PD has not fixed newspaper site search. The algorithms have never been great, and the way it is implemented on most newspaper sites isn’t exactly user friendly.
Most people are used to search working the way Google works — a simple interface with one set of results, but as Google is demonstrating, the web is no longer text based. It is a combination of text, databases and multimedia.
That is an issue, to its credit, that PD tried to address with its original “integrated search” solution, what Google calls “universal search.”
For newspapers, you have discreet collections that people might want to search:
- Current articles
- Archives
- Classifieds
- Verticals (auto, real estate, jobs)
- Events/Calendar
- Local web sites
- Local businesses/Advertising
At a minimum.
If you’re aware of the intention-driven way people use the web (think pull, rather than push), you’ll understand why this can’t all be shoved together in one neat search solution.
But I’m not aware of any newspaper site that has yet really figured out how to create a satisfying search experience within the context of a robust news and community web site.
Starting in Q4, this will be one of my projects. If you have any suggestions, want to brainstorm on this a bit, leave a comment or contact me through Facebook or LinkedIn.
Absolutely right, Howard. We have used PD for a few years and have had generally good experiences because it combines site search with regional searches (ours searches 16,000 local searches simultaneously – a cool feature, but users don’t “get” it yet, and that’s our fault). While not perfect, Planet Discover is a lot better than what we’ve used in the past.
Further complicating matters is we (the newspaper industry) have so many co-branded third-party sites (everything from Legacy.com to Cars.com) that don’t generally get included in our search results in an easy-to-use way.
We can kick butt on local coverage all day long, but if the user can’t find our stuff with keyword searches, it doesn’t do much good.
Also remember that our pals at Google have set the de facto — and thereby much imitated — standard for how broad Internet search should work and how results should be displayed.
The noble efforts of Ask notwithstanding (and even Ask tips its user-experience hat to Google on its new results page designs in several key areas): if the way Google groups, organizes and displays results is what people expect — that’s what we’d better give them.
[…] Howard Owens says newspaper sites need better search. Boy, howdy! I commented that we have to consider the de facto standards for search keyword matching and results display established in recent years by Google and now often imitated. […]
Hi, Howard. As someone who has thought a lot about search (teaching and writing related to Google) and who also is a programmer at a news site (working for WPNI), I think getting search right is a very hard problem to solve. Certainly, universal search or grouped results are not hard, but as you mention here, that’s only one step in the right direction and still leaves a lot to be desired.
The problem lies squarely with relevance. How do I know by the keywords users enter which of the results from which collections of data would be the most useful to readers? Solve that problem and you’re getting closer to a decent solution. I think one step in the right direction is following user habits from past searches (i.e. which results did they click on) as well as factoring in most popular or most emailed. I don’t know, though, that seems only a step in the right direction, too. I think it’s tough to get right.
The web has a clear authority system as Google has shown, back links. But how do we determine authority across our own sites? Only questions from me, I realize, but search is an issue I hope to put some thought and work into as well.
Excellent job of describing the problem. These are certainly issues that every content rich site is experiencing.
One of the difficult things is that people have become lazy in their information seeking behavior. They basically use Google and nothing else. I’ve noticed on our own sites that local search is almost never used. It is used so infrequently that you almost have to question giving the search box such a prominent position on your pages.
Creating a better search is one issue. I think that is solvable. Getting people to use it is a completely different issue.
Deryck Hodge had hit the nail on the head – site search is all about relevance. At SLI Systems we use the idea he mentions of following click throughs to improve relevance. It works very well.
Search on news sites need to give highlight newer content. We’ve used a couple of approaches to achieve that. At DM news the default sort order is by date – but we offer the ability to sort by popularity. At the NEJM we segment the search results page – showing the most popular at the top and the most recent lower on the page.
WRT search different types of content – by default the site search should search all the content you have but offer the ability to restrict the search to a particular type of content.
Better search on newspaper sites means more choices, more ability to narrow down a selection like some shopping sites do. If you’re buying shoes, on a good site you can choose, say, “black,” size 9.5, “tall” or “ankle.”
Another important tool would be the ability to use booleans for those who know them, and some kind of techno crutch for those who have no idea what they are, making it possible for them to narrow their search as well.