18 May 2010 11:10 pm   //   Filed under: Stray data, Technology

Online writing and the power of “should”

Today’s blog post is about using math to make writing more effective. You should read it!

A couple of months ago, I noticed a curious phrase showing up on lots of blogs.

“You should follow me on Twitter here.”

This phrase stands out for being terse, awkward, even rude. Most people would write “Please…” instead of “You should…” Yet this specific line of clunky self-promo copy spread like the flu. A Google search for that exact phrase returns 154,000 results! (For comparison, a search for “Please follow me on Twitter here” returns 1,690 results.)

We can trace this phenomenon to blogger Dustin Curtis, who used testing to find the optimal way to convince people to follow him on Twitter. “You should follow me on Twitter here” was proven to be the most persuasive sentence. You should read about his experiment here.

I have conflicted feelings about this. On one hand, I don’t want to endorse shoddy writing edited by machines. On the other hand, shouldn’t you use every weapon in your arsenal to make your writing more effective? You should!

I decided to try a test of my own. For the last six weeks, visitors to this blog have been part of an experiment.

* * *

On my home page and at the top of this blog, I’ve been running a link asking people to give money to a charity bike ride I’m doing this Saturday and Sunday. I wrote a simple program that randomly shows one of a series of pre-programmed messages, then records how many times each message was displayed, and how many times people followed the link from that message. I use that data to calculate the percentage of clicks each link got. Then I compared the click-through rate of each phrase against other phrases that ran at the same time.

Turns out Dustin Curtis was right. Adding “You should” to the front of a sentence significantly increases the number of people who click on a link. So does adding “You can.” Longer sentences produce better results. Adding a user benefit helps. Including a dollar value helps sometimes, but not always—I tried $5, $10, $25, but only $25 had an impact.

The results of my test follow.

But first a few notes: This was more for fun than for science. Because my site gets relatively little traffic, it would have been a stronger test if I had chosen fewer phrases or let each test run for more time. Additionally, a lot of my traffic comes from spiders for search engines; my simple program counted them as visitors. I figured spiders behave in predictable ways and were unlikely to skew the results of the test, but counting them may have contributed to higher- or lower-than-normal click-through rates. There was also a significant swing in the number of total clicks from week to week, perhaps due to the behavior of spiders.

* * *

Test 1: April 3 to 17
Phrase % clicks
You should give to my Bike MS ride. 2.31
Give $15 to my Bike MS ride. 1.95
Make a difference. Give to my Bike MS ride. 1.89
Give to my Bike MS ride! 1.80
Give to my Bike MS ride. 1.70
What’s the easiest way you can make a difference today? 1.70
Will you give $15 to my Bike MS ride? 1.64
Please give to my Bike MS ride. 1.52
Average 1.81


Lessons from test 1: I started with the most basic phrase I could think of as a control: “Give to my bike MS ride.” I tested it against the same phrase with some additions: Extra words, a dollar amount, a reason to act, an exclamation point, etc. Adding “You should” produced the strongest results. Adding “Please” produced the weakest results.

Test 2: April 18 to May 1
Phrase % clicks
You can help fight MS with a $15 gift to my Bike MS ride. 1.42
You should give to my Bike MS ride! 1.30
You should give to my Bike MS ride. 1.25
You should give to my Bike MS ride. Help make a difference! 1.21
You should support my Bike MS ride. 1.21
You should support my Bike MS ride May 22 and 23. 1.20
Support my Bike MS ride May 22 and 23. 1.08
You should support my Bike MS ride. Give $15! 1.08
Average 1.22


Lessons from test 2: I started with the control phrase “You should give to my Bike MS ride” and tried a few variations. The strongest performer included a reason to give: “You can help fight MS with a $15 gift to my Bike MS ride.” Two of the weakest performers included dates. Once again exclamation points made no difference.

Test 3: May 2 to May 18
Test phrase % clicks
You can help fight MS with a $25 gift to my Bike MS ride. 2.31
You should give $5 to my Bike MS ride to help fight MS. 2.15
You should give $25 to my Bike MS ride to help fight MS. 2.07
You should give to my Bike MS ride to help fight MS. 1.95
You can help fight MS with a $15 gift to my Bike MS ride. 1.91
You can help fight MS with a gift to my Bike MS ride. 1.82
You should give $15 to my Bike MS ride to help fight MS. 1.67
You can help fight MS with a $5 gift to my Bike MS ride. 1.62
Additional breakdowns for May 2 to May 18
“You should…” messages 1.96
“You can…” messages 1.91
$5 messages 1.89
$15 messages 1.79
$25 messages 2.19
Messages with no dollar suggestion 1.88
Average 1.94


Lessons from test 3: Here I split my 8 phrases into two groups, one consisting of “You should…” and the other of “You can…” There was no significant difference in click-through rate. I also included some versions with suggested donations, varying the amount I suggested. The only change in results was seen in adding the value of $25.

* * *

Summary: “You can help fight MS with a $25 gift to my Bike MS ride” was the most powerful phrase of the bunch. However, the real point is not whether it motivated people to click, but whether it motivated people to give. Sadly, it did not. During the time I ran this test I received 14 donations. (P.S.: Thank you!) I am fairly sure sure all of the people who gave found out about my bike ride from e-mail or Facebook postings, not from this web site. You could argue that some might have seen this message here as well, and that reinforced their behavior. But still: Nobody followed a link and then decided to give.

But that can change right now! You can help fight MS with a $25 gift to my Bike MS ride.