Yahoo showed us 30 days of logos. Here’s the one consumers liked best.

With a new CEO last summer and a $1b acquisition this spring, Yahoo has been changing significantly. Next up is its iconic logo: Tomorrow Yahoo will unveil a new logo to signify the new era. To draw attention to the change, Yahoo has been displaying a different logo every day for the last month.

We were curious about which logo consumers preferred as the best fit for the Internet giant, so we used the Survata logo testing tool to find out. We asked 12,725 respondents to pick their favorite of five logo variants (randomly selected from the 28 variants released prior to publication).

And the winner is…

Consumers displayed strong  and consistent opinions about the variants.  The “selection percentage” ranged from 47% for Day 10 (the most preferred) to 6% for Day 21 (the least preferred).  Day 10 was rated highly across all age, gender, and geographic groups.

Our large sample size enabled a detailed look at relative preferences for each pair of logos.  The most preferred variant, Day 10, is undefeated, having been selected in the majority of head-to-head faceoffs against every other variant.


The more logos change, the more they stay the same?

To understand why consumers preferred certain variants, we deconstructed the Yahoo logo into its major design attributes.

We then classified each variant according to these six attributes.  For five of six attributes, consumers preferred the variants with the attribute of the current Yahoo logo.

Our findings from Yahoo’s “30 days of change” campaign suggest it should stick to something familiar.  So, for Yahoo’s sake, we hope the new logo announced tomorrow will remind users of its graphical heritage.

Thinking of a new logo? Don’t guess; test.

Survata has researched consumer opinion on logos for numerous clients, and you can do so for yourself using our fast and affordable logo testing site.  Check it out next time your branding needs a face lift!

And, Yahoo, if you ever get bored of Tumblr’s look, we will happily crunch the numbers on that change as well.

Footnotes for our fellow data geeks

  1. We interviewed 12,725 online respondents from the US between August 27 and September 2, 2013.
  2. You can download the underlying data here.
  3. You can analyze the underlying data on Statwing.

Men are 50% more likely than women to support whistleblower Edward Snowden

This week’s news has been dominated by reports on Edward Snowden, the government consultant who disclosed classified information on NSA surveillance programs to the press. Americans have expressed a range of reactions to Snowden, with some considering him a hero acting to protect civil liberties from government abuse and others considering him a traitor who damaged a key national security tool.

As a market research company, we at Survata couldn’t resist probing this topic.  We wondered whether perception of Snowden differed by gender.  Turns out, it does.

We surveyed 1,207 Americans from June 12 to June 13, and asked them:

Last week, Edward Snowden disclosed to the press details of a classified National Security Agency surveillance program.  What is your opinion of his actions?

After excluding the respondents who said they were not familiar with the Snowden story, we found that men are 50% more likely than women to support Snowden.  Counterintuitively, men are also 13% more likely than women to oppose Snowden.  This is possible because women are significantly more likely to have no opinion (yet) on Snowden.

Curious to find more correlations?  You can analyze our data in Statwing or download it here.  Just be aware, of course, that someone might be watching you work…

Want to run your own public opinion survey? Try Survata today.

Footnotes for our fellow data geeks

  1. To the question listed above, respondents could answer: I strongly support his actions; I somewhat support his actions; I have no opinion / I am not sure; I somewhat oppose his actions; I strongly oppose his actions; I am not familiar with this story.
  2. For our analyses we excluded all respondents who answered “I am not familiar with this story.”

10 mistakes startups make when talking to users

Most startups know the golden rule for success: Talk to your users. One popular method of doing so is running online surveys. Survata has run consumer surveys for numerous startups to help them understand people’s opinions and behaviors.

When you run a survey, it is important avoid the common mistakes that can reduce or even negate the value of survey results. To help you write effective survey questions, we compiled a list of the most frequent errors our startup clients make when writing surveys. Avoid these pitfalls, and start talking to those users!

1.) Leading questions

What it means: Embedding an opinion or information into a question that suggests the user should answer a certain way

Mistake: Leading question

2.) Subjective units

What it means: Requiring the user to interpret descriptive unit labels

Mistake: Subjective units

3.) Excessive granularity

What it means: Asking the user to recall information at unrealistic levels of detail

Mistake: Excessive granularity

4.) Unbalanced scales

What it means: Putting answer choices on a scale that tilts in one direction

Mistake: Unbalanced scales

5.) Double-barreled questions

What it means: Addressing more than one issue in a question

Mistake: double-barreled question
Improved (Only “interested” respondents proceed to Question 2)

6.) Unnatural units

What it means: Asking the user to recall information in a way that does not match his/her memory structure

Mistake: Unnatural units

7.) Lack of mutual exclusivity

What it means: Having overlap in your answer choices, so a user doesn’t know how to mark the answer

Mistake: Lack of mutual exclusivity
Improved (Removed redundant “Every few hours” option)

8.) Lack of comprehensive exhaustiveness

What it means: Not covering all possible answers

Mistake: Lack of comprehensive exhaustiveness

9.) Phrasing to format mismatch

What it means: Prohibiting the user from answering as the question implies s/he should (e.g. limiting the user to one answer when multiple are requested)

Mistake: Phrasing to format mismatch

10.) Deadwood

What it means: Including words that add no information

Mistake: Deadwood

Have you come across any other survey writing mistakes? Send them to us – we’ll share our ever-expanding list.

Good luck with your next survey, and keep talking to users!

Inspired to run your own mistake-free survey? Check out Survata today.

Is Snapchat only used for sexting? We asked 5,000 people to find out

As evidenced by its growing media buzz and copycats, Snapchat is one of Silicon Valley’s hottest startups. The Snapchat smartphone app lets users send messages, pictures, and videos that disappear once they’re viewed. Launched last year, it already has millions of users, with a strong slant towards a younger audience.

The conventional wisdom is that Snapchat’s self-destruct feature encourages “inappropriate” uses like sexting – the sending or receiving of texts, photos, or videos of a sexual nature. And Snapchat’s own onboarding process, featuring two scantily clad ladies, seemingly promotes that use case:

Snapchat Onboarding

We decided to put this Snapchat-for-sexting speculation to the test by running a survey on our own Survata research network.

How prevalent is sexting on Snapchat?
We interviewed 5,475 US respondents aged 18 to 29 between January 19 and February 1, 2013. From that pool, we found 715 self-reported Snapchat users, and 13.1% of them admitted to sexting. Naturally, this begs the question of whether 13.1% is high or low. As a baseline for comparison, we interviewed a completely separate group of 453 self-reported text messaging users. 26.5% of text messaging users admitted to sexting, in line with prior studies.

So, our first conclusion: Sexting is less frequent among Snapchat users than among text messaging users. Consider this myth busted.

Men are from Mars…

The story doesn’t stop there, though. In our study, both gender and age affect one’s proclivity to sext. Whether over Snapchat or text messaging, men are significantly more likely than women to sext. On Snapchat in particular, sexting is more than twice as common among men than women!

Also, against the stereotype of party animal college students, we found that on Snapchat, 23 to 29 year olds actually sext more than 18 to 22 year olds.

If not for sexting, then what?

If sexting is fairly rare, then what is truly driving the Snapchat craze? To find out, we asked 260 Snapchat users, via a free response question, how they used the app. Frequently cited uses were to “send funny pictures to my friends” or “make silly faces for my friends” or “send jokes to my friends.” Perhaps this is just a new communication paradigm for a younger generation. As a Snapchat representative wrote, “There’s a reason why it’s illegal to record a phone call without permission – and it’s not because everyone is having phone sex.”

So keep on snapping, kids; we adults will try to be less suspicious going forward.

Want to challenge other conventional wisdom with data ? Run your own survey with Survata today.

Footnotes for our fellow data geeks

  1. We asked self-reported Snapchat users “Have you ever used Snapchat for “sexting”? (Sending or receiving texts, photos, or videos involving nudity or of a sexual nature)”
  2. We asked self-reported text messaging users “Have you ever used text messaging for “sexting”? (Sending or receiving texts, photos, or videos involving nudity or of a sexual nature)”
  3. The margin of error for the percent of users who sext over Snapchat is 3.7%. The margin of error for the percent of users who sext over text messaging is 4.6%.
  4. When survey results are weighted by gender, the difference between sexting prevalence on Snapchat vs text messaging is slightly more pronounced — 27.2% of text messaging users and 12.7% of Snapchat users report sexting (unweighted is 26.5% vs 13.1%).
  5. You can see the original data set here, or play with it in Statwing (Snapchat respondents here, text messaging respondents here).
  6. Acknowledgements: Thank you to Jacob Wenger, Greg Laughlin, Harendra Guturu, and Russell Morton for helpful comments.

How accurate are Americans at estimating calories?

The holiday season is officially in full swing. For some that means decorating a tree; for others, buying presents. But for everyone, it means food – lots and lots of food. Perhaps the American tradition of late December binge eating is why losing weight is perpetually a top New Year’s Resolution, and why NBC is already heavily promoting its next season of The Biggest Loser that starts in January.

Conventional wisdom holds that one main reason Americans have trouble keeping off the pounds is that they overwhelmingly underestimate the calories in food. This is presumably the logic behind laws that require chain restaurants to display calorie counts on menus. We decided to test this conventional wisdom using our Survata research network. So we asked over 6,000 US internet users to estimate the total calories in the following 8 meals and snacks:

  1. Pancake meal: 2 pancakes with butter and syrup, 2 strips of Oscar Mayer bacon, 1 cup of hashed brown potatoes (1,060 Calories)
  2. Cereal meal: 2 cups of Special K cereal with 1 cup of 2% milk, 1 Thomas’ English muffin, 8 oz of orange juice (594 Calories)
  3. Chicken soup meal: 1 cup of Campbell’s chicken noodle soup, 1 banana, 16 oz of Snapple lemonade (431 Calories)
  4. Roasted chicken meal: 1/2 breast of roasted skinless chicken, 1 cup of brown rice, 8 oz of skim milk (446 Calories)
  5. Cookies & milk snack: 3 chocolate chip cookies, 8 oz of Nesquik fat-free chocolate milk (394 Calories)
  6. Carrots & pretzels snack: 10 baby carrots, 1 oz bag of Rold Gold pretzels (150 Calories)
  7. Subway meal: Subway foot-long tuna sandwich, 1 bag of Doritos, 20 oz of Coca-Cola (1,443 Calories)
  8. Domino’s & McDonald’s meal: 2 slices of cheese pizza from Domino’s, medium-sized order of McDonald’s french fries, 20 oz of Coca-Cola (943 Calories)

The results may surprise you. Let’s dive into the numbers.

Take your best guess

In aggregate, the calorie estimates were quite accurate. The average was within 65 calories of the actual calorie content for 4 of the 8 meals.

Against conventional wisdom, Americans in our survey did not overwhelmingly underestimate calorie content. In a previously published study, Americans underestimated calorie content 82% of the time. We also observed a bias towards underestimation, but at a much lower rate – only 62%. And on a meal-by-meal basis, the median respondent:

  • underestimated for 5 meals (Subway, cereal, pancakes, roasted chicken, chicken soup)
  • estimated perfectly for 1 meal (carrots & pretzels)
  • overestimated for 2 meals (cookies & milk, Domino’s & McDonald’s)

Would you like fries with that?

We were most surprised by the results for fast food restaurants. Our study included Subway, which has a “healthy” reputation (thanks to millions of dollars worth of Jared ads), and the combination of Domino’s and McDonald’s, which have a less-than-stellar reputation for diet-friendly foods. Amazingly, even though the Subway meal had 500 more calories than the Domino’s & McDonald’s meal, on average the respondents estimated the Subway meal had 639 fewer calories than the Domino’s & McDonald’s meal! Perhaps Americans are too anchored to Subway as a healthy choice, and Domino’s & McDonald’s as distinctly unhealthy choices. (On a related note, we found 7 sandwiches on the Subway menu had over 900 calories, and that’s before you even throw in the chips and drink!)

Together we were right, but each of us was wrong

While average estimates were accurate (including within an impressive 5 calories – about 1% – for the chicken soup meal), individual respondents displayed a wide range of accuracy. This is a variant of the wisdom of crowds effect, which states that a group’s average estimate of a measure is often accurate even when most individual estimates are inaccurate. In fact, in our study, only 14% of respondents estimated the chicken soup meal’s calories within 10% of the actual calorie content. And no meal had even 1 in 5 people give a “close” answer:

Screen Shot 2014-01-06 at 4.46.39 PM

What we learned

It our study, Americans did not overwhelmingly underestimate calories for most meals. Estimates for fast food calories were significantly misguided – but in both directions! Apparently we have been misled by those oh-so-catchy jingles.

To both our under- and over-estimated readers, as we enter the heart of holiday food season: Happy Eating to All, and to All a Good Bite.


Footnotes for our fellow data geeks

  1. We interviewed 6,587 online respondents from the US on December 6-13, 2012. Each respondent provided a calorie estimate for only 1 meal. We eliminated extreme outlier responses (the highest and lowest 2.5% for each meal), leaving 6,256 responses for further analysis.
  2. You can view the raw responses and the sources for calorie “actuals” here, and analyze it with Statwing.
  3. You can read about the science behind calorie measurements here.
  4. Our survey unit looked like this:
  5. We love feedback and ideas for future statistically-oriented blog posts! Email us at
  6. Disclaimer: We’re not doctors, so we can’t speak to the overall health impact of these foods. We know healthy food is about more than just calories.
  7. Acknowledgements: Thank you to Greg Laughlin, Ryan White, Jacob Wenger, Susie James, and Harendra Guturu for helpful comments.