Internet Usage By Time of Day

When do people use the Internet?  The answer is simple enough -- whenever they want to at wherevery they have Internet access.  Immediately, several obvious facts jump out.  But this is still a very broad characterization.

We will now refer to some survey data from the 2004 TGI-EGM Colombia study.  Within this survey, there were 1,620 survey respondents who reported their Internet usage time by time of day (namely, on a hourly basis).  If Internet usage were completely random, then the usage would be distributed uniformly by the hours of the day.  

However, we are interested in asking a slightly different question:  At what time of day do people first use the Internet?  This is known as a first-time-use question.  If Internet usage were completely random, then the first-time-usage would also be distributed uniformly by the hours of the day.  This is almost true, but not quite.  For example, if everyone uses two hours of Internet per day and we are counting from midnight onwards, then we would have counted everyone by after 10pm and there would not be anyone coming in at 11pm for the first time.

At any rate, the following chart shows the first-time'use numbers for the hours of day in the 2004 TGI-EGM Colombia study.  For example, 1.4% of the Internet users first appeared between 12 midnight and 1am; 1.9% appeared between midnight and 2am (of which 1.4% appeared between midnight and 1am and 1.9% - 1.4% = 0.5% appeared between 1am and 2am); and 100% of the Internet users would have appeared at the end of the day by definition.

In this chart, the 45-degree straightline represents a baseline of random uniform first-time usage.  Internet usage begins very slowly after midnight and falls far short of the random baseline.  This is obviously very much expected because most people sleep in the wee hours of the night.  The curve then accelerates sharply after 7am up through noon.  There is a short drop-off during l unch time, and then the usage picks up again rapidly up through the end of the workday.  Thereafter, the cumulative curve stays over the random baseline.

(source: 2004 TGI-EGM Colombia)

The discussion above points out that there is a relationship with people's work habits during the day.  This can be made explicit by asking people about what type of access they have, specifically with respect to home and work usage.  Of the 1,620 survey respondents, 44% said they had home access but no work access; 9% said they had work access but no home access; and 13% said that they have both home and work access.  Please note that this only adds up to 66%.  There is another 34% who have neither home nor work access, because they go through schools, public access terminals, internet cabinas, and friends' homes.

The chart below shows the first-time-usage curve of these three groups.  The three groups all sleep at night.  Upon arriving at work, the two work-access groups immediately rise up in usage.  By 9pm, the work-only group has reached 100% with no more users afterwards.  By 6pm, only two-thirds of the home users have accessed the Internet already, and the rest would drift in eventually during the evening.  

(source: 2004 TGI-EGM Colombia)

How important is the time-of-day information for media planning?  In traditional media, time-of-day information is important because it affects the audience level.  For example, the article International Comparisons of Radio Usage proves that radio usage levels are significantly higher during the morning and evening commute time periods.  Those time periods generate larger audiences, reaches and frequencies of exposures, and so they garner higher prices for advertisements.

For web advertisements such as banner-ads, this is usually not considered.  The commercial arrangement is typically based upon a number of banner ads served over a certain period of time (such a week), regardless of the time of day.  In this sense, the Internet is considered to be like print media (such as newspapers and magazines).  For other forms of promotion on the Internet (such as a mass email notice), this might be an issue for something that is time-sensitive (e.g. a specific promotional event that is scheduled to run at a specific time).

(posted by Roland Soong, 11/19/2004)