How does the length of the TV attribution window affect performance?

As a refresher, C3 calculates the lift based on when the spot airs.  When post logs are imported into C3, one of the key elements is to determine if the spot created lift and resulted in a conversion or sale--both determined by the C3 site tag and C3 conversion tag.  C3 is concerned with lift/and visits, but only if that touchpoint was part of a conversion or sale currently or at some time in the future.

This is done by calculating the lift from three specific channels against the apples-to-apple geography for which the spot ran...doing so based on the TV attribution window or latency window.

That window is usually between 2-10 minutes depending on the strength of the offer in the TV spot.  If a consumer is going to respond to a spot (unless a product targeting an elder audience calling by phone), they will respond by visiting the website of the advertiser within "x" minutes.

Some agencies have very defined windows for "x" they do not deviate from:  one of the largest independent DR TV agencies adheres to a 4-minute window, no exceptions.  Another agency will specify a window between 2 and 7 minutes based on a number of factors and historical data.  For Direct Response, C3 recommends a standard practice of 5 minutes.

If the window is too short, you miss capturing all the activity from the DR spot.  

If the window is too long, you end up with a situation that actually negates TV credit.  How can this be?

Remember that lift is calculated off the apples to apples baseline for time, geography, and three specific buckets of traffic, which result in a conversion.

In order to create lift over a 15 minute sustained window, the TV spot has to deliver results for that entire sustained period (i.e. more traffic from the buckets of organic, paid brand search, and SEO for the entire 15 minute period compared to the same 15 minute period average from the same buckets in the same geographies).  

Having a TV spot generate traffic for a sustained period of 15 minutes (unless long form) is highly unlikely, and since performance is measured by way of lift it becomes very hard to create lift for a 15 minute period--especially during M-F daytime.

Why would this be?

If a spot airs during the daytime, website traffic in total has a predictable cycle.  As mid-day approaches, website traffic increases.  By about 3pm, website traffic begins to steadily decrease until about 6pm and stays flat through midnight.  Each site will vary based on its target demographic and product offering, but generally:  traffic during prime-time TV is 2x lower than peak mid-day.

How does this affect lift?  If the window is set to a longer period, it will be harder to create lift (you have to generate high numbers off a high numeric base since the window is longer).  Some advertisers want the TV window to be as long as possible thinking that the longer the better:  TV will get more credit.  

In fact, this is almost always opposite.  By setting an abnormally long window, it requires TV to create a sustained lift of 15 minutes which is nearly unheard of...and exacerbated even more if those TV spots air during M-F daytime when website traffic is already high and viewership of TV is typically lower.

Setting the window to 5 minutes has demonstrated to provide the optimal measurement capturing two standard deviations of activity, but not so much that lift measurement gets clouded by a long window).  

The difference between setting it shorter or longer typically relates to how strong the offer is, but we suggest 5 minutes as a standard.