Dates in Interaction Design – Actual Research!
Many web sites – particularly those concerned with travel – ask users to provide one or more dates. In most cases this is done either by selecting a date from a popup calendar or by entering it as text. The text field itself will often include a template showing how the date is to be entered, for example dd/mm/yy in the UK (see the Expedia.co.uk example below).
Unfortunately, people in a hurry or those visiting several sites in quick succession – often with differing date formats – may not type dates exactly as requested. An alarming number of sites still handle this badly from a user-experience perspective. British Airways, for example, does not allow enough characters in the field to accommodate the full year (so if 22/07/2009 is entered the field shows just 22/07/20, which is interpreted as 2020 - thanks a bunch). Lufthansa, on the other hand, takes great exception to the omission of leading zeros.
The really frustrating thing, to my mind at least, is that it would be a lot easier for developers to interpret dates intelligently than to present an error message and reload the page. (Ajax or Dynamic HTML can make this a little less painful, but not much.) To compound the foolishness of web sites complaining about dates is the fact that it is really discouraging to potential customers to have to spend time dealing with such pedantries. People have been writing dates for several thousand years and while there are important variations (like the order of day, month and year) omitting leading zeroes or providing more digits in the year than asked for (as in the BA example) is only a problem when it comes to poor interaction design.
To investigate this issue, I conducted a survey in June 2009 of nearly 1,000 people, asking them simply how they would write a given date (‘the second of August this year’ was how it was phrased). They typed their answer into a free-format text field. Respondents were then invited to choose the closest ‘template’ date from a list of 32 – half of which were in day/month/year order and the other half month/day/year (‘other’ was the final option).
Participants were recruited from a number of US and UK email lists, mostly HCI, usability and user-experience related. For this reason, list members were also asked to distribute the participation request to non-technical colleagues, friends and relatives.
988 valid responses were received (a small number of ambiguous results were deleted). Responses were checked and adjusted so that the ‘template’ date selected matched the date entered in the type-in field. The results are charted below by date format: month/day/year (435 responses), day/month/year (518 responses) and year/month/day (30 responses). The percentages shown are relative to each group. (Note that the actual punctuation used in the type-in fields may differ from the templates. For example, some dates were written 2-8-09. Also, the use of ‘2nd’ has been taken simply as ‘2’. In all cases, the order of date fields, the use of month – numeric, abbreviated or full – and the presence or absence of leading zeros has been accurately reflected in the results.)
% Participants by Date Format (month/day/year order, n = 435)
% Participants by Date Format (day/month/year order, n = 518)
% Participants by Date Format (year/month/day order, n = 30)
As can be seen from the above, a wide variety of date formats were invoked. Participants who used month/day/year ordering had a slightly stronger preference for a purely numerical date with no leading zeros in the month or day (8/2/09) but the close second used the full month name (August 2 2009). These two categories accounted for 53% of that group.
Participants using day/month/year ordering were somewhat less ambivalent in their most popular format: full month name with no leading zero (2 August 2009). The next three runners-up were permutations of numerical dates with and without leading zeros.
Numerical summaries of the use of leading zeros, month and year designations appear in the tables that follow.
(Notes on the tables: Leading zeros are counted for day and month fields. There is only one leading zero possible when the month is not written in Arabic numerals. FullYear is 0 for ‘09’ and 1 for ‘2009’. Five participants wrote the year as ‘9’, for but simplicity of analysis, these were counted as ‘09’.)
Finally, to ascertain whether technically-minded participants might favour certain characteristics of date formats over others, respondents were asked if their main field or work or study was in science or technology. Responses were on a five-point scale from ‘1 = strongly agree’ to ‘5 = strongly disagree’):
This table shows that 62% of respondents had a scientific or technological leaning (categories 1 and 2). However, aside from a small number being prone to using year/month/day format, separate analyses of variance show no significant relationships with factors such as the use of leading zeros, month or year designations.
Overall, 76% of respondents made no use of leading zeros. So, it’s official – don’t reprimand users for failing to supply leading zeros in date fields. Other results are interesting but less useful in interaction design: 54% of all respondents used Arabic numerals for the month and 62% preferred full four-digit years to the two-digit variety (these are measured independently so it is no surprise that the total is more than 100%).
So, if you are dealing with dates and users have no artifact to work from (such as a credit card), numeric dates with a four-digit year might be the way to go. However, if you are asking people to enter dates from another source (credit card, form, email and so on) do take account of what they are looking at and allow for variations. In any event, only complain if you really cannot make sense of what has been entered.
(Do not take this as a license to use three drop-down fields. This is very inefficient compared to typing a date in. Also, don’t forget that your solution needs to work for assistive technologies such as screen readers.)
Many thanks to the hundreds of participants who completed the study. I hope it will lead to a brighter, safer user experience!
For the sheer fun of it, tag clouds of the three date orderings appear below. All punctuation has been removed since no tag cloud generator could be found that processed punctuated dates correctly. (These and any other parts of this page may be copied freely as long as credit is given. Please provide a link to this page.)
Tag Cloud of Month/Day/Years Dates (Punctuation Removed)
Tag Cloud of Day/Month//Years Dates (Punctuation Removed)
Tag Cloud of Year/Month/Day Dates (Punctuation Removed)