03 / 05
Four cases the whole industry learned from.
Brands everyone has heard of, mistakes and wins that were investigated in public. We did
not run these campaigns. We take them apart so you can use what they proved.
BMW · Germany · 2006
The shortcut that cost a car maker its index
bmw.de served search engines a page stuffed with the German phrase for used car, then
bounced people to a different page. Google removed the site and its PageRank showed zero.
It was back inside a week, once the doorway pages were deleted rather than defended.
The lesson: showing a crawler something a person will not see is the one line that gets
you removed rather than demoted, and the fastest repair is deletion.
Announced by Matt Cutts of Google, February 2006 JCPenney · United States · 2011
Number one for everything, until a journalist counted
A retailer sat at the top for dresses, furniture and more. The New York Times counted
2,015 pages carrying paid links, on sites about nuclear engineering and Bulgarian
property. Google confirmed the violation and rankings fell past page five within hours.
The lesson: rented rankings are a liability that settles on somebody else's schedule,
and the bill arrives in one afternoon rather than over a quarter.
The New York Times, The Dirty Little Secrets of Search, February 2011 Zapier · pages at scale
Tens of thousands of pages, one template
Zapier's integration directory is the most copied structure in organic search: one
template, one page per pairing of tools, each answering a question somebody actually
types. You can open the directory and count the pattern yourself in two minutes.
The lesson: scale only works when every generated page answers a real query. The same
template with nothing behind it is the thin content Google has spent fifteen years
filtering out.
Structure is public and checkable at zapier.com/apps Any case study · including this page
How to check a case study in ten minutes
Traffic figures quoted for the same well known site differ by a factor of four between
agency blog posts, because most are third party estimates rather than measurements. That
is not a scandal, it is what estimates are. It does mean you have to check.
The lesson: ask who measured it, with what, and on what date. If a case study cannot
answer those three, treat the number as a story rather than evidence.
Method, not a claim. We apply it to our own numbers too.