Northern Light

» see northernlight.com as it lived, on the wayback machine

~ the obit ~

northern light launched in august 1997 from cambridge, massachusetts. the company was founded by david seuss and a small team. they built a search engine that did two things differently from altavista and lycos. first, the results were organized into hierarchical folders by topic. second, the index included a "special collection" of paid premium content (academic papers, industry reports, paywalled news) that you could buy individual articles from for a few dollars each.

i never used northern light. by the time i got online in late 1998, my search engine in turkey was altavista, then yahoo's directory, then very quickly google. northern light was a US thing that did not have much reach outside its home market. i learned about it from older programmers later who had used it for academic search.

the folder organization was the unique selling point. you searched for something. you got the usual list of results. but you also got a sidebar of "custom search folders" that grouped the results by topic. so a search for "java" would have folders like "java programming language," "java island indonesia," "java coffee," "java certification courses." you could click into a folder to see only those results.

on 16 january 2003 northern light closed its consumer search engine. they pivoted to selling enterprise search software. the public search index went offline. the special collection went away (the underlying content moved back to the publishers). northern light still exists in 2026 as an enterprise product for big companies. but the public search engine is gone.

~ the rap sheet ~

~ how the folders worked ~

the folder system was a real technical achievement for the late 90s. instead of just listing results, the engine clustered them by topic and presented the clusters as folders. this is harder than it sounds. you have to look at all the result snippets, identify what each is about, group results that talk about the same thing.

the clustering used a custom taxonomy and some statistical techniques. northern light had editors who maintained the topic taxonomy. the clustering was automatic but the topic labels were human-curated. the result felt more organized than altavista's flat list.

this was the same time period when yahoo was running a hand-curated directory and google was building pagerank. northern light was a third option: organize automatically by topic. the approach did not win against pagerank, but it produced a kind of result that has actually never come back to general-purpose search.

~ the special collection ~

the other unique feature was the special collection. northern light had partnerships with publishers (academic presses, news services, industry research firms) to index their paywalled content. when you searched, results from the open web were free, but results from the special collection cost money. typical price was $1 to $4 per article.

this was useful for researchers. you could find a relevant journal article from your search results page and pay for it on the spot. no university login required. no separate database to navigate. just a search result with a price next to it.

the model was ahead of its time and slightly off-target. researchers were used to going through their library's database access. casual users did not want to pay for articles. the special collection was profitable but not big enough to subsidize a competitive consumer search engine. and the open-web search side was losing to google.

~ how it died ~

google killed it. by 2002 google had taken most of the consumer search market through better relevance and a cleaner interface. northern light's folder system was interesting but not interesting enough to overcome google's better core search.

the special collection model was ahead of its time but did not scale. publishers were uneasy about distributing through a third-party engine. customers were uneasy about paying per-article. the pivot point was clear by 2002.

on 16 january 2003 northern light closed the consumer-facing search and refocused on enterprise. they still sell search software to fortune 500 companies. it is a perfectly fine business. but the public-facing northern light, the one with the folders, is gone.

~ what we lost ~

the folder organization. modern search engines do not cluster results by topic. google has tried various forms of related-search suggestions and topic facets but never the full folder treatment. for some kinds of queries, the folder approach is genuinely better than a flat list. nobody has it anymore.

the per-article paywall search. this could have been a model for academic and industry research. you search, you find what you need, you pay for the specific article. instead the world settled on subscription-based access (academic libraries, jstor) and pirate sites (sci-hub). northern light's micro-payment model was a path that did not get followed.

the era when search engines were still trying different things. between 1997 and 2003 there were maybe ten serious search engines (altavista, lycos, infoseek, hotbot, excite, ask jeeves, northern light, google, plus several others). each one tried something different. by 2005 google had won and the experimentation stopped. northern light was one of the more interesting experiments.

@}-->--

« previous  ·  random  ·  next »