May 17, 2023

Lessons in search data with Seth Stephens-Davidowitz

Date & Time (GMT):
May 17, 2023 12:47 PM
Date & Time (EST):
May 17, 2023 12:47 PM

At Observe Summit 2022, we heard from Seth Stephens-Davidowitz, data scientist and best-selling author of “Everybody Lies” who told us that a lot of the data that we’re using to understand people is flawed. Seth explained how Google search data can give more accurate insights as people are extremely honest when using Google and talking about almost every topic. So, Google can provide a comprehensive dataset that’s rich and informative.

The problem with survey data and how search addresses it

One way of understanding consumer preferences is to ask them directly via surveys. But Seth explained that it’s easy to get skewed or inaccurate data this way because people lie in surveys. “More than 50% of non-voters lie to surveys that they voted,” he stated.

His method involves looking at people’s Google searches. This can give more accurate information as people are more honest: they tell Google what they might not share in surveys, to friends, family, or even therapists. Seth calls this tendency up pour your heart out to Google “digital truth serum.”

Analysing Google Trends data gives you a very different view of people, and it’s been shown that it provides more accurate information than survey data. For example, when it comes to predicting who will turn out and vote, it’s very hard with surveys because people want to say, “Yes, I’m going to vote.” But Google searches on where to vote, how to vote, voting information, and polling places give a much more accurate prediction on voter turnout than asking people if they’re going to vote.

Similarly, when it comes to measuring racism, Seth explained that if you ask people in a survey, “Are you racist?” very few people are going to admit they are. “But on Google, people make racist searches with sometimes shocking frequency,” he stated.

He gave an example of how in 2016 when Donald Trump was running for President, there was a clear correlation between racism and support for Trump. It was found that the single biggest predictor of support for Donald Trump in the Primary was racist Google searches. “People aren’t going to admit this out loud but they’re going to admit these attitudes on Google,” Seth added.

Social media data vs. search data

When it comes to social media as a data source, Seth explained that this gives you a very different view of people than search data. This is mainly because people prefer to give off an idealised image of themselves on social media  - they create a public persona. Google, on the other hand, is private so only they have to know what they type in.

For example, here’s how people describe husbands on these data sources:

  • Social media – Husband is… “the best”, “my best friend”, “amazing”, “the greatest”, “so cute”
  • Google searches – Husband is… “amazing”, “gay”, “a jerk”, “annoying”, “mean”

These responses are quite different!

Seth explained that although social media is an excellent data source, “you always have to keep in mind, “Am I getting a complete perspective or can I round my understanding with more data”?”

Search as a predictive tool for brand marketing

As it relates to the corporate world, search can be an effective tool for understanding your customers better. Because people are so honest with Google, the data might be able to tell you things that people wouldn’t otherwise admit.

“You’re so familiar with your brand that you don’t remember that the people you’re trying to sell your brand to don’t know nearly as much as you do and they might have really basic questions,” explained Seth. “And if you can answer those basic questions or speak to those basic questions, you may have a lot more success selling your product.”

You have to consider what basic questions people have about your product or service.

The risks of analysing search data

One of the biggest challenges when analysing search data is that you never know why people are making a search. “You have to keep in mind that Google searches frequently measure interest in the topic, whereas interest isn’t always what you want to measure or what you’re interested in,” he explained. “You don’t just need everybody to know about you; you want people to feel positively about you…”

Overall though, the honest data that search engines can provide you with, offers valuable insights to  help inform your business decisions.

Or view the interview on LinkedIn

This interview was recorded via LinkedIn Live, if you prefer to view on LinkedIn, click the button below.

View Interview

See related content

Webinar & Panel
[Webinar] More Than a Soundbite: Storytelling Using Social Media Data
Webinar & Panel
A panel discussing Keeping the Social Data Tap On
Webinar & Panel
IPSOS Challenge: Finding social insight beyond Facebook, Twitter & Instagram