Predicting Netflix Customer Losses Using Social Media and Search Data

Back in July, when Netflix announced structural pricing changes that enraged a vocal segment of its customer base, we noticed a spike in certain search queries and social media posts that suggested a fresh wave of home entertainment comparison shopping. Tactically, we also noticed that, despite tens of thousands of public complaints, Netflix adopted a policy of “hear no evil, see no evil” in its social media communications. Last week, Netflix disclosed that approximately 1 million Netflix customers had stopped service in the two months after the announcement. This indicates that there is indeed a correlation between those data points and performance, and that the company’s unapologetic social media style is playing a role as well.

With this week’s stunning public letter from Netflix CEO Reed Hastings, we’re seeing another spike in potentially telling data. We’ll show you some of those trends and benchmark it in hopes of finding out just how much correlation there is between search and social media data and actual customer fallout.


We marketing consultants spend a lot of time advising brands to listen to their customers in social media channels such as Twitter and Facebook. We also advise them to engage as many disgruntled customers as possible in hopes of making them into happy brand advocates. On both July 12th and September 19th, Netflix announced changes that so enraged its customers that reaching every one of them would have been impossible.

Netflix pricing announcement


Despite the public outrage, Netflix decided to execute its usual social media strategy of posting lighthearted commentary about national holidays and new Netflix programming. In the case of the July 12th announcement, it responded to 86,000 mostly seething Facebook and blog comments with two new Facebook posts within seven days about other subjects. Netflix customers did not seem to be interested in discussing topics such as how the Mark Wahlburg film The Fighter could provide workout inspiration; instead, these posts racked up an additional 7,500 or so additional angry comments about the announcement on the 12th.

Netflix posts workout inspiration

Then, in what will surely go down as one of the more bizarre product announcements in tech history, on September 19 Netflix responded to widespread criticism over its pricing model with a tweet that said that it had been “listening” to its customers.

Was Netflix really listening?

It then provided a link to the Netflix blog in which CEO Reed Hastings apologized for not being communicative in July, and then, several paragraphs later, announced the decision that will split its brand and service in half. Netflix had decided to embrace the long view of its business model, and this meant more bad news for Netflix customers. As more than 26,000 blog comments (and counting) appear to illustrate, this will lessen the convenience of its service, and surely send more customers running for the exits at a time when Netflix is already vulnerable due to the widely-publicized loss of a content deal with Starz.


After the pricing announcement in July, we noted spikes in search queries that appeared to demonstrate customers actively comparison shopping. These below are only a subsection that appear in two primary patterns: those related to “Netflix alternatives” and “Netflix vs [insert competitor].”

Netflix Comparison Shopping Search Data

As for the graphs, the tall columns on the right represents search query volume for July 2011, with the other bars in the chart representing volume in the previous 12 months. As you’ll note, July wasn’t the only month with a spike in terms that compared Netflix to specific competitors; there had also been considerable activity in the spring when Blockbuster and Amazon Prime embarked on marketing campaigns to create awareness for competitive products. These queries could have just as easily been non-Netflix customers as existing customers.

However, the August figures were not only more dramatic, but also coincided with a virtually new type of search query, one in which customers appeared to be searching for any alternative to Netflix at all. Incidentally, the figures under the “Local Monthly Searches” column represent the estimated number of searches for each term in July 2011. For those new to Google keyword tools, please note that extremely round numbers and some data duplication are common with these tools; therefore the numbers should be used as overall indicators but cannot be used for complicated statistical computations.

So, understanding that these search volume estimates are far from perfect, there were in the neighborhood of 90,000 unique broad match searches for terms that appear to be competitive shoppers in July. Judging by the prevalence of these two search streams alone, there appeared to be a correlation of about 11 departing customers for every correlated search query. Note that correlation is not the same as cause-and-effect. In other words, we can’t guess which of these customers searching “netflix alternative,” for example, had already decided to leave the service regardless of what they found, or whether any did. The point is that this level of activity is surely an important barometer.

In terms of social signals, in this case blog comments and Facebook posts, there were 86,000 in direct response to the announcement, and at least 7,500 afterwards, nearly all with negative sentiment. Again, this correlates to approximately 11 departing customers for every correlated comment in these channels. Incidentally, there are also a great number of Facebook “Likes,” however in this case, since Likes are also used as a sharing tool, we don’t believe it necessarily translates to positive sentiment. We also have mountains of Twitter data, but apart from verifying spikes, the sentiment is much harder to gauge since much of the Twitter mentions of the brand are link-based without commentary.

As for the September 19th announcement, in the two days since the announcement there have been approximately 26,000 blog comments (about 5x the total number for the July announcement) and 10,000 comments on the Facebook post. In two weeks, we’ll review the resulting search data and see where we are on social signals, and update this post. We’ll also post a followup when subscriber losses are posted publicly.


If the July data is any indication, Netflix may lose more customers simply their handling of public customer service than they would based on the choice to actually split the DVD and streaming services. It also looks like the short-term fallout for Netflix from the September announcement will be heavier this time. As much as Netflix can control its public communications, it may also feel so confident in its superiority of service from its competitors that it isn’t concerned about customer comparison shopping (“I can’t quit you, baby”). The relatively small streaming content library at a service like Amazon, for example, may also help to stem the tide.

Examining the ensuing social media and search data after these announcements can’t turn back the hands of time for Netflix, but it may yield valuable economic lessons for other brands considering difficult decisions that may displease its customers, and how it responds in social media channels.







No comments yet.

Leave a Reply