,
38 tweets,
11 min read
Read on Twitter
The details of the lyrics controversy haven't been discussed enough in my view. There's been a lot of press but little analysis and investigation. There's nuance that needs to be teased out. I hope you find this thread informative.
1/ Licensing is an issue that comes up a lot when it comes to lyrics. As it should to ensure that the creators are being compensated. Genius did run afoul of this but came to a licensing agreement in May of 2014. Citation: nytimes.com/2014/05/07/bus…
2/ This was the same month in which LyricFind proclaimed that "every significant lyric website in English is now licensed or shut down". Citation: lyricfind.com/index.php?id=15 
3/ So as of May of 2014 licensing of lyrics shouldn't have been a major issue moving forward. Sure, you'd have to police new sites but industry organizations and watchdogs were quite attentive.
4/ One of the things I see happening is conflating licensing with accuracy. When it comes to lyrics these two things are not synonymous. LyricFind even mentions this in their response to these scraping allegations. Citation: lyricfind.com/index.php?id=3…
5/ That conflation seems to carry over to Google's Search Quality Rater Guidelines. These guidelines provide a way for Google to evaluate changes to their algorithm. Citation: cnbc.com/2018/09/17/goo…
6/ In October of 2015 Google released an update to their Search Quality Rater Guidelines that included details on how to evaluate Lyrics. Citation: thesempost.com/google-quality…
7/ Details of these Search Quality Rater Guidelines around lyrics show the confusion. Raters are instructed to give the LyricsBox is Fully Met rating, the highest, with a note that those lyrics are licensed. (Please note that captures are from the March 28, 2016 guidelines.)
8/ In another section they specify how to rate a lyrics site. In this instance it was the number one lyrics site at the time. Raters were instructed to give this a Moderately Met rating, in large part because of accuracy, which we have established is not synonymous with licensed.
9/ While not explicitly said, the inference is that LyricsBox content is more accurate than the content on lyrics sites. The reason seems to be that the content is licensed. But we've established that licensing does not equal accuracy.
10/ These guidelines already establish a tilted playing field of sorts where the content in the LyricsBox is always better than the content on lyrics sites. So lets think about what it means to take content from a lyrics site and use it in the LyricsBox.
11/ According to these guidelines that means content deemed less accurate, to be rated as Moderately Met, was being used in the LyricsBox where it is suddenly transformed into Fully Met content. This is a big deal IMO.
12/ This is exactly what happened as of December of 2015 when scraped lyrics showed up in the Lyrics Box. Citation: seroundtable.com/google-stealin…
13/ This is absolutely scraping and not some mistake in verifying specific lyrics. MetroLyrics had a feature where non-lyric content was inserted. It talked about the meaning of that lyric. Citation: yvoschaap.com/weblog/misbeha…
14/ Some tried to give Google the benefit of the doubt here. Maybe they both got these from the same source?
15/ That seems unlikely. As of December of 2015 LyricFind was providing licensing and feeds for ... MetroLyrics. At that time I don't believe Google was getting their feed from LyricFind. Citation: lyricfind.com/index.php?id=15
16/ To me the more important issue with that incident is that the scraped artifacts in question were removed. This is a tacit acknowledgement that Google knew about the issue and took steps to fix the problem.
17/ This establishes that content in the LyricsBox was scraped from other lyric sites and that Google knew this was happening. Further, three months later many more instances of the same type of scraping were identified.
18/ I reached out to Google and provided examples of scraped content. This is one of the images from the email I sent.
19/ The result was a verbal discussion that this was wrong and would be addressed. Sure enough all the instances I provided were quickly and quietly cleaned up. But as we've seen, the scraping continued.
20/ Another important date to note is June 27, 2016 which is when Google partnered with LyricFind. Citation: lyricfind.com/index.php?id=2…
21/ Yet the scraping continued. Whether it was before the LyricFind relationship or after, scraped content has regularly been featured in the LyricsBox.
22/ While LyricFind says they didn't scrape Genius content the method by which Genius watermarked those songs makes that nearly impossible. How else would the same pattern of different apostrophes make it into the LyricsBox?
23/ According the the WSJ article Genius informed Google of this problem as far back as 2017. This means that in a post LyricFind relationship world Google was again made aware of scraped content in the LyricsBox. Citation: wsj.com/articles/lyric…
24/ According the LyricFind, they too were informed at least twice that Genius had found scraped content. Citation: lyricfind.com/index.php?id=3…
/25 I believe there's clear evidence that scraped content has been used in the LyricsBox regularly for over three years across different licensing and feed agreements.
26/ I can't speak to copyright law or specific legal issues. But as a search marketer I work under Googles guidelines. And it seems like Google should operate under those same guidelines.
27/ I have two major problems with the LyricsBox controversy. The first is that scraping, something against Google's guidelines, has been allowed to take place. They knew about it and allowed it to continue for years.
28/ The second problem is that their own rater guidelines essentially describe accuracy as being important when it comes to lyrics. And the only way Google seems to measure this is if it appears in the LyricsBox.
29/ Yet the work to get those lyrics right is being done elsewhere, by other sites in this space. Sites that rate Moderately Met according to Google are the ones providing the added value in this vertical.
30/ As I've said before, if Google sourced these lyrics legitimately I'd be pissed for my client but I'd have to simply deal with the new playing field. But that's not what has been happening.
31/ So putting an attribution link to LyricFind as a fix just doesn't work for me. The track record we have shows that Google does not ensure content in the LyricsBox is not scraped from other sites. In fact, it shows the opposite.
32/ There is a large business case for keeping the LyricsBox as is to ensure that Google can capture as much free traffic to their subscription services. Services that have under-performed against the competition. Citation: engadget.com/2019/05/09/you…
33/ But I see no reason for Google to continue rendering the LyricsBox differently from any other featured snippet. Google knows which sites are licensed and should snippet the lyrics and site the source.
34/ Here's a lo-fi mock-up of what it could look like. (Just imagine the Genius reference is in blue. I'm failing to alter the color because of the inherit function and I don't want to deal with a Peter Griffin CSS episode.)
35/ And this doesn't have to be Genius every single time. It should be the site that has the best or most accurate lyrics for that query. The site that best fulfills that intent.
36/ A small coda to all of this is the fact that being first to market here is incredibly important. A new song comes out and the volume in those first week or two is enormous. So being better and faster is a huge advantage.
37/ Many of the watermarked songs are those where being better and faster deliver outsized business value. All songs are not created equal, so winning on the ones that generate the most volume is critical.
