Compare

46% less effort, the comparison summary

Using the Workman key cap scores, Norman's score is 46% less than QWERTY's. If you prefer finger travel metrics, Norman has 44% less finger travel from the home row than QWERTY. Both metrics beat Colemak, Dvorak, Klausler, Minimak, QGMLWY, Workman, and Asset. The Andong analysis produces much different metrics, but still puts Norman in the lead over others with 23% less effort than QWERTY.

Workman key cap scores

Forgoing sophisticated mathematical models, I reran a similar Carpalx-style ranking test using the scoring used per key cap in the Workman layout. Every time a letter is used, it adds to a simple counter. In scoring a body of literature with Workman, letter S adds 1 point to the total, D is 2 points, and so on.

When scoring the following sentence, lower is better:

The quick brown fox jumps over the lazy dog.

• QWERTY: 90
• Asset: 71
• Capewell: 73
• Colemak: 73
• Dvorak: 62
• Klausler: 76
• Minimak: 70
• Norman: 70
• QGMLWY: 69
• Workman: 68

For this analysis, the following were concatenated into a single document: Drupal 7.17, The Hunger Games by Suzanne Collins, CISSP 5th ed exam guide by Shon Harris, The Montessori Method by Maria Montessori, Alice in Wonderland, Dracula, Great Expectations, Huckleberry Finn, Moby Dick, The Picture of Dorian Gray, Pride and Prejudice, Sense and Sensibility, The Adventures of Sherlock Holmes, The Adventures of Tom Sawyer, The Count of Monte Cristo, Ulysses, Walden, the front page of Slashdot on November 15, 2012, and a sample from the global feed of App.net.

By counting the frequency of each character in the combined text, the letter frequency was the following:

etaoinsr hd lcumfpygwbvkxjqz

Spaces were added for emphasis. Note that the first 8 letters are all in the home row of Asset, Colemak, and Norman. The 8 home row letters account for 62.9% of all letters typed, excluding whitespace, numbers, and punctuation; the first 10 are 71.8%. Considering just the placement of letters, a huge majority of typing in Norman is either on the home row or in the natural stroke of the middle fingers.

Here are the scores for the concatenated document using the Workman key cap model. Lower is better.

• QWERTY: 87,341,770
• Dvorak: 57,389,636
• Klausler: 54,102,945
• QGMLWY: 52,374,051
• Minimak: 51,193,339
• Asset: 50,583,227
• Capewell-Dvorak: 50,344,818
• Colemak: 49,709,750
• Norman: 46,936,221
• Workman: 46,582,636

If you want to try some analysis of your own, I posted the PHP script I wrote for examining various text inputs in a gist.

Carpalx metric

The Norman layout scores about the same as the Workman or Dvorak layouts using the Carpalx analysis. The analysis output is in the Norman download.

• QWERTY: 3
• Dvorak: 2.098
• Norman: 1.993
• Workman: 1.993
• Minimak: 1.930
• Asset: 1.894
• Colemak: 1.842

The Carpalx score varies depending on the input corpus text. During my testing, I got anywhere between 1.9 and 2.1 by changing the text that was analyzed. The score was 2.011 using the same text used with the Workman key cap score analysis.

Note that the Carpalx analysis doesn't account for some human factors, as Martin Krzywinski admits.

The Workman layout incorporates advanced human factors and discusses their application to Dvorak and Colemak.

In recent conversation with Stephen O'Connor, who has analyzed the Workman layout, I've been persuaded to seriously reconsider the parameters in my effort model. In particular, I do not consider that the index, middle and ring fingers have different prefered motions, for a given travel distance. For example, most will agree that the ring finger prefers to extend for the W rather than curl to the equidistant X. On the other hand, the index finger has easier access to V than R.

Key frequency heat mapping

Patrick Gillespie has a keyboard layout analyzer which can accept custom keyboard layouts. Skip to the keyboard heat map in his analyzer. Then decide for yourself from the heat map if the other metrics make sense to you. The analysis from this section was based on a combined sampling of blog articles I typed myself, hosted on https://deekayen.net.

What these metrics don't show is Norman's QWERTY similarity and what I think is an increased chance of converting successfully, since most keys stay on their QWERTY finger. I'm including these to show how dramatically finger travel decreases with Norman while still being no worse in the other metrics than its competition.

Finger distance (in feet)

• QWERTY: 3122.5
• Minimak: 1942.9
• Dvorak: 1926.7
• Klausler: 1828.8
• Workman: 1822.3
• Norman: 1818.7
• Asset: 1790.5
• QGMLWY: 1761.1
• Colemak: 1761.1
• Capewell: 1728.0
• Arensito: 1715.2

Looking at these metrics, you might say the Norman layout is optimized for distance rather than finger-to-key frequency combinations. Though I don't have anything scientific to point to, it makes sense to me that almost halving the finger travel over QWERTY could contribute to an overall plan of hand-injury prevention more so than a strategy of combination avoidance.

Home row usage (in percent)

• QWERTY: 25.7%
• Minimak: 51.1%
• Norman: 51.4%
• Workman: 51.4%
• Arensito: 52.5%
• Capewell: 52.6%
• Dvorak: 54.1%
• Klausler: 55.8%
• Asset: 56.1%
• Colemak: 56.1%
• QGMLWY: 56.1%

Note, Norman and Workman make a point of de-prioritizing the center column of keys (G and H in QWERTY) whereas Colemak considers the center column part of the home row. Workman reduces overall usage of the two middle columns by about 50% compared to Colemak.

Consecutive finger use

• Norman: 4.7%
• QWERTY: 4.6%
• Minimak: 3.1%
• Asset: 2.6%
• Workman: 2.4%
• Dvorak: 2.2%
• Klausler: 1.6%
• Capewell: 1.4%
• Arensito: 1.3%
• Colemak: 1.3%
• QGMLWY: 1.3%

In this case, Norman's right index and middle fingers are overloaded compared to the other modern layouts, which gives up the improvement those layouts strived for. Norman tries to assign common characters to strong fingers and I think those are my two most agile, strong fingers.

Consecutive hand and thumb use

• Capewell: 31.0%
• Arensito: 29.7%
• QWERTY: 29.5%
• Norman: 28.4%
• Asset: 28.3%
• Workman: 27.0%
• Minimak: 26.4%
• QGMLWY: 25.8%
• Colemak: 25.8%
• Klausler: 18.5%
• Dvorak: 17.6%

Dvorak is sometimes criticized for this outcome. Longtime users sometimes comment that the Dvorak layout feels jittery when alternating from hand to hand. In designing Workman, OJ preferred high same-hand utilization.

This heat map was generated using the 1,000 most common US English words.

To make your own, import the following Norman layout along with some text of your own creation.

{"label":"Norman","fingerStart":{"1":29,"2":30,"3":31,"4":32,"5":56,"6":56,"7":35,"8":36,"9":37,"10":38,"11":56},"keyboardType":"standard","keys":[{"primary":96,"shift":126,"finger":1,"id":0},{"primary":49,"shift":33,"finger":1,"id":1},{"primary":50,"shift":64,"finger":2,"id":2},{"primary":51,"shift":35,"finger":3,"id":3},{"primary":52,"shift":36,"finger":4,"id":4},{"primary":53,"shift":37,"finger":4,"id":5},{"primary":54,"shift":94,"finger":7,"id":6},{"primary":55,"shift":38,"finger":7,"id":7},{"primary":56,"shift":42,"finger":8,"id":8},{"primary":57,"shift":40,"finger":9,"id":9},{"primary":48,"shift":41,"finger":10,"id":10},{"primary":45,"shift":95,"finger":10,"id":11},{"primary":61,"shift":43,"finger":10,"id":12},{"primary":8,"finger":10,"id":13},{"primary":9,"finger":1,"id":14},{"primary":113,"shift":81,"finger":1,"id":15},{"primary":119,"shift":87,"finger":2,"id":16},{"primary":100,"shift":68,"finger":3,"id":17,"altGr":-1,"shiftAltGr":-1},{"primary":102,"shift":70,"finger":4,"id":18,"altGr":-1,"shiftAltGr":-1},{"primary":107,"shift":75,"finger":4,"id":19,"altGr":-1,"shiftAltGr":-1},{"primary":106,"shift":74,"finger":7,"id":20,"altGr":-1,"shiftAltGr":-1},{"primary":117,"shift":85,"finger":7,"id":21},{"primary":114,"shift":82,"finger":8,"id":22,"altGr":-1,"shiftAltGr":-1},{"primary":108,"shift":76,"finger":9,"id":23,"altGr":-1,"shiftAltGr":-1},{"primary":59,"shift":58,"finger":10,"id":24,"altGr":-1,"shiftAltGr":-1},{"primary":91,"shift":123,"finger":10,"id":25},{"primary":93,"shift":125,"finger":10,"id":26},{"primary":92,"shift":124,"finger":10,"id":27},{"primary":20,"finger":1,"id":28},{"primary":97,"shift":65,"finger":1,"id":29},{"primary":115,"shift":83,"finger":2,"id":30},{"primary":101,"shift":69,"finger":3,"id":31,"altGr":-1,"shiftAltGr":-1},{"primary":116,"shift":84,"finger":4,"id":32,"altGr":-1,"shiftAltGr":-1},{"primary":103,"shift":71,"finger":4,"id":33},{"primary":121,"shift":89,"finger":7,"id":34,"altGr":-1,"shiftAltGr":-1},{"primary":110,"shift":78,"finger":7,"id":35,"altGr":-1,"shiftAltGr":-1},{"primary":105,"shift":73,"finger":8,"id":36,"altGr":-1,"shiftAltGr":-1},{"primary":111,"shift":79,"finger":9,"id":37,"altGr":-1,"shiftAltGr":-1},{"primary":104,"shift":72,"finger":10,"id":38,"altGr":-1,"shiftAltGr":-1},{"primary":39,"shift":34,"finger":10,"id":39},{"primary":13,"finger":10,"id":40},{"primary":16,"finger":1,"id":41},{"primary":122,"shift":90,"finger":1,"id":42},{"primary":120,"shift":88,"finger":2,"id":43},{"primary":99,"shift":67,"finger":3,"id":44},{"primary":118,"shift":86,"finger":4,"id":45},{"primary":98,"shift":66,"finger":4,"id":46},{"primary":112,"shift":80,"finger":7,"id":47,"altGr":-1,"shiftAltGr":-1},{"primary":109,"shift":77,"finger":7,"id":48},{"primary":44,"shift":60,"finger":8,"id":49},{"primary":46,"shift":62,"finger":9,"id":50},{"primary":47,"shift":63,"finger":10,"id":51},{"primary":-16,"finger":10,"id":52},{"primary":17,"finger":5,"id":53},{"primary":-91,"finger":5,"id":54},{"primary":18,"finger":5,"id":55},{"primary":32,"finger":5,"id":56},{"primary":-18,"finger":6,"id":57},{"primary":-91,"finger":6,"id":58},{"primary":-93,"finger":6,"id":59},{"primary":17,"finger":6,"id":60}]}

Andong analysis

For analysis, I concatenated a bunch of microblog posts from the front page of App.net as real-world text to analyze.

Overall effort

• QWERTY: 10,090
• Dvorak: 8,824
• Workman: 8,396
• Colemak: 8,388
• Norman: 8,269

Effort imbalance between fingers

• Dvorak: 657
• Colemak: 628
• Workman: 605
• QWERTY: 556
• Norman: 525

Don't trust me? Try it yourself. Here is the custom layout definition I imported into the Andong analyzer, which is no longer online.

`1234567890-=
#qwdfkjurl;[]\
#asetgynioh'*N
*Lzxcvbpm,./*R
##*S#
*L
#######&**()_+
######JURL:{}|
######YNIOH"
######PM?
#
*R
~!@*#$%^
#QWDFK
#ASETG
#ZXCVB
#

The meaning of it all

Many of these metrics show that Norman is not always the premier performer in all the various possible metrics people have designed across the Internet, but they also show that the differences are small. Changing the input texts does change the analysis outputs and the rankings. Take a moment to consider how sensible your keyboard design is.

• Do you think the center column is part of the home row?
• The majority of typing happens with ASETDHNIOR. Are those keys in a position that makes sense for you?
• How much frustration will you get changing from QWERTY to your new layout?
• How much QWERTY similarity do you need for shortcuts?

Whether you choose Norman or something else, leaving QWERTY behind is an obvious choice for dramatic change in your typing effort.