Intensity in consumer complaints about banks

30 Oct

Analyzing the language used in consumer complaints tells you about both the topics that people are complaining about and their severity. An appreciation for what people are saying can help you build better products, save valuable customers, and fix problems earlier. In the case of financial service complaints, customer language can also expose what’s known in regulation circles as “unfair, deceptive, or abusive acts or practices” (UDAAP). There were $2.5b in UDAAP settlements in 2014, up 30% from 2013.

In this post we take one small but revealing aspect of language: intensifiers. There are a lot of ways that people show intensity–in speech, they increase their volume, in text they may use ALL CAPS or rows of exclamation points. But right now let’s look at words that are traditionally called “intensifiers”–like very and really. Explicit accusations of deception often come with intensifiers–but as is often the case with language, a word that accompanies explicit accusations also helps pinpoint implicit ones. And outside of accusations of deception, intensifiers also help identify highly emotional content.

In daily conversation, people usually use intensifiers about positive things. People talk about really enjoying things and how they are really neat. They say thanks very much and that things are very interesting. That said, people’s everyday speech also has a lot of very important and very difficult. In Spanish speech, the words that usually occur with muy are bien, poco, importante, and difícil. These are common in Portuguese, too–muito (‘very’) also goes with bem (‘well’), importante, and difícil. Regardless of your native language, if you reflect on where  intensifiers appear, you’ll see they aren’t just used to intensify verbs and adjectives–they’re used to intensify a speaker/author’s commitment to a claim.

Take a look at how they are used in customer complaints lodged against financial institutions. Looking closely at intensifiers identifies issues with customer service as well as unfair, deceptive, and abusive acts and practices:

Chase’s lack of appropriate and timely processing of my family’s request is literally forcing us into foreclosure but I struggle to keep my mortgage current b/c of the adverse professional ramifications.

Please help me they prey on people that are poor and withouta car I cant work. I have gotten soooo mad and it is not good for my health

I explained to him I want to pay my loan I just can not afford the xxxx withdraws of $240.00 bi-weekly, he was extremely rude and ridiculed me saying he could not help me with anything until my account made it his way

tHey are now out of business filed bankruptcy sold their portfolio to a third party and cant be found. PRO-COLLECT IS ILLEGALLY TRYING TO COLLECT ON ILLEGAL BILLING STATEMENTS THAT ARE TOTALLY FALSE AND WITHOUT MERIT.

Overall, about 30% of complaints against financial firms include intensifiers. Reddit provides an interesting contrast set because they have tens of thousands of forums focused on very different matters. The median percentage of posts-with-intensifiers in Reddit forums is 15%.  Only 5% of all of Reddit forums have as many intensifiers as complaints about banks–for Reddit, these are highly emotional topics having to do with problems in romantic relationships and debates on religion or gender. In the financial service complaints data, the very highest percentage of intensifiers is in Mortgages–that’s when people are talking about their families losing their homes, so it’s no wonder that it’s so high.

We can get more granular than Mortgages. Across all different kinds of financial products, let’s look at what sort of issues customers use intensifiers with disproportionately:

  • Can’t repay my loan
  • Loan modification, collection, foreclosure
  • Application, originator, mortgage broker
  • Dealing with my lender or servicer
  • Problems when you are unable to pay
  • Problems caused by my funds being low
  • Communication tactics

In other words, people are using intensifiers in highly-fraught situations when their homes or possessions are on the line, as well as when they feel like there is problematic communication. This recalls one of the major findings about one-star ratings in Yelp reviews–they are rarely about food, they are about awful service.

For a sense of contrast, here are categories where consumers use fewer intensifiers than we’d expect if everything were just random:

  • Incorrect information on credit report
  • Improper use of credit report
  • Unable to get credit report/credit score
  • Credit reporting company’s investigation

This also means that while people issue complaints to credit bureaus, they don’t use that many intensifiers–so complaints against TransUnion, Experian, and Equifax have low rates of intensifiers. The highest rates of intensifiers in complaints go with companies like Green Tree Servicing, Enhanced Recovery Company, Ocwen, NationStar Mortgage, and Wells Fargo. That’s particular because while bad credit ratings definitely affect people, it’s not as intense an emotional situation as a home being on the line. Automated processes are also seen differently than direct contact with humans (loan officers, etc).

Intensifiers are just a tiny aspect of assessing risk. Ideally, you want a system that considers all kinds of words and phrases–actually, you want to detect these automatically and give them weights based on the statistical strength of their signal. To learn more about the ways that adaptive machine intelligence works to do this, check out this blog post or our use cases page

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: