Friday, November 30, 2012

The Climate, Week Nine: Holiday Ratings Resiliency

Today in The Climate, a few observations about the ratings downturns during Thanksgiving week.


Week Ending TPUT y2y bc y2y LeAv y2y
19/30/201232.4 -6% 9.2 -16% 2.50 -15%
210/07/201232.8 -5% 8.7 -15% 2.26 -17%
310/14/201233.7 -1% 8.6 -10% 2.33 -9%
410/21/201233.4 -3% 8.6 -6% 2.30 -6%
510/28/201234.2 -0% 9.1 -3% 2.12 -16%
611/04/201234.5 -0% 9.2 -9% 2.32 -11%
711/11/201235.1 +1% 9.3 -9% 2.30 -7%
811/18/201233.5 -5% 8.4 -14% 2.13 -11%
911/25/201232.4 -7% 8.4 -12% 2.02 -16%


Week Ending TPUTy2d y2dy2y bcy2d y2dy2y LAy2d y2dy2y
19/30/201232.4 -6% 9.2 -16% 2.50 -15%
210/07/201232.6 -5% 8.9 -15% 2.39 -16%
310/14/201233.0 -4% 8.8 -14% 2.37 -14%
410/21/201233.1 -4% 8.8 -12% 2.35 -12%
510/28/201233.3 -3% 8.8 -10% 2.31 -13%
611/04/201233.5 -3% 8.9 -10% 2.31 -12%
711/11/201233.7 -2% 8.9 -10% 2.31 -12%
811/18/201233.7 -2% 8.9 -10% 2.28 -12%
911/25/201233.5 -3% 8.8 -11% 2.26 -12%

Because of the way the TV calendar broke, Thanksgiving was frustratingly part of week 9 this year and part of week 10 last year. So week 9 had the largest drop-off in overall viewing of any week this year and the entertainment average was at close to worst-of-season levels. (Broadcast viewing looked a little less comparatively ugly due to an extra NFL game on Thursday and a massive college game on Saturday.) It seems likely week 10 will go totally in the other direction, since that will compare this current week of significant broadcast upswing with last year's Thanksgiving week. Ultimately, the season-to-date numbers should be more apples-to-apples next week, and I wouldn't get too invested in either the week 9 or week 10 week-by-week numbers.

Just want to note a couple things about holiday ratings. Most people hate to see originals on Thanksgiving Eve because it means low ratings. But really, nothing's getting axed because of a Thanksgiving Eve result. It's actually one of my favorite nights of the ratings season because I like seeing regularly scheduled shows out of their "comfort zone," so to speak.

Here are a few general observations that I may use in the future to help illuminate ratings drops on holidays:

1. You drop more the younger you skew. I've thrown out this generality on holidays many times in the past. Here's a quick example: the oldest-skewing show that aired an original on Thanksgiving Eve was CSI (on average this season, 29% of its audience is in-demo). It also had the smallest demo drop on Thanksgiving Eve (-4%). The second-oldest show Criminal Minds (32%) had the second-smallest drop (-10%) and third-oldest show Survivor (34%) had the third-smallest drop (-12%). The X Factor is the youngest-skewing show (46%) and had the largest drop (-24%).

Awhile back another site tweeted something along the lines of, "The holiday drop in 18-49 usage was 10%, and all show drops beyond that are just #FanExcuseBingo." You might intuitively expect it to be that way. In fact, in my stupidity I initially felt like maybe shows should drop by less than the usage drop. People casually watching syndicated shows and news shows on cable should be more likely to fall away from the TV than entertainment programs to which people are committed every week, right?!

However, the reality is the opposite. Most shows on broadcast drop by noticeably more than the drop in overall usage. At some point thisis consistent enough that it stops being an "excuse." The Thanksgiving Eve 18-49 usage drop was about 10%, and as you can see from the "Demo" column above, almost everything dropped by more than that. The median entertainment original drop was -15%.

It seems holiday drops are much less about "commitment" to the shows and much more about "availability." The people who have to travel or participate in holiday activities are going to do that, and no level of love for or commitment to a TV show is gonna change that. It seems that drop in availability is larger among the younger set; and most of broadcast primetime is still relatively young-skewing.

2. Everything dropped more in the demo than in total viewers on Thanksgiving Eve.

Demo Vwr
Private Practice -23% -11%
Survivor Fall -12% -7%
Criminal Minds -10% -5%
CSI -4% -2%
Whitney -21% -7%
Guys with Kids -15% -11%
Chicago Fire -13% -4%
Law and Order: SVU -24% -10%
The X Factor Wed -24% -11%

Some of these differences (Chicago Fire, X Factor) were actually so extreme that it appears there may have been more lost demo viewers than lost total viewers (meaning the shows actually rose a bit in out-of-demo viewers).

This is simply another example to illustrate that the in-demo viewers are much more likely to bail on TV than the out-of-demo viewers. And though we focus on 18-49 ratings, that has a practical use within that subset; 18-34s are more likely to bail on TV than 35-49s.

Where am I going with these two observations? I've always wanted a way to be able to measure some sort of expected "resiliency" in my True Strength formula. The True2 for CSI on Thanksgiving Eve was easily a season high because it held up well amid a big viewing drop, but for The X Factor it was a low (aside from some very oddly scheduled eps around baseball season). It seems like bringing total viewership into the mix might be a quantifiable way to help even those out somewhat. A show's "skew" tends to have a pretty solid correlation with its tendency to drop/not drop in what should be a terrible situation.

3. Is the Thanksgiving Effect larger than two days? I've always thought of the viewing downturn as really only affecting Thanksgiving Eve and Thanksgiving Night. And indeed, those are still by far the two worst nights. I estimated a -11% 18-49 usage drop week-to-week on Thanksgiving Eve (an official number in a press release said -10%) and -13% on Thanskgiving despite a huge-rated NFL game.

But the Tuesday before Thanksgiving also had a bad night; an estimated -5% in 18-49 viewership and a median entertainment drop of -6%. And the shows that held up the best? You guessed it: the really old-skewing ones. Dancing with the Stars was up, NCIS was barely down, NCIS: LA was even and Vegas was up. Monday was less conclusive; viewing was down just 2% and while The Voice had a terrible night, nobody else changed all that much.

The real tip-off that suggests those days are "mini-holidays" was the huge upswings on Monday and Tuesday of this week. Looking back on post-Thanksgiving weeks in years past, there is a general tendency to grow in that week, including on Monday and Tuesday. However, there are also a lot fewer originals aired in this week most years. This is really the first year in which it's been obvious over a large sample size.

Click to expand for more on the "climate" numbers used herein.

TPUT - This is an ESTIMATED average of how many people are watching TV from 8:00 to 11:00.
  • I derive these numbers by adding up all the ratings and dividing by all the shares in each of the 42 half-hours each week. That means there is some error relative to the numbers Nielsen actually releases. Sadly we don't regularly have access to those. I always advise not to rely heavily on these numbers for any one show in any one week, but the hope is that the error is minimized across a 42-timeslot sample every week.
  • I include the Old Methodology adjustment, which makes the number more like a measurement of how many people watch primetime programming Live + SD, rather than a measurement of how many people watch any TV (including old DVR stuff) from 8:00 to 11:00. This makes the number perhaps less intuitive in a vacuum, but it's pretty much a wash when making week-to-week and year-to-year comparisons, which is what we're really interested in.
bc - This is an average of how many people are watching national broadcast TV from 8:00 to 11:00.
  • This does NOT include the 10:00 adjustment used in the True2 calculation which attempts to account for Fox/CW programming and stronger cable. Again, that perhaps hurts the number in a vacuum, because the 10:00 numbers being used only include three networks, so I'm averaging timeslots that are somewhat apples-to-oranges. But again, it's a wash when making comparisons because I treat it that way all the time. It would not really change week-to-week or year-to-year comparisons, and that's what I mostly care about.
  • Another important note here is that these numbers include the preliminary averages for "sustaining" programming like presidential debates and commercial-free benefit concerts whose numbers are typically omitted from traditional Nielsen averages. I might eventually omit these from this particular calculation, but they're needed on my spreadsheets to 1) make PUT calculations in those timeslots and 2) create a competition number for the entertainment shows that air against them.
LeAv - This is a measurement of how many people watch the average moment of original entertainment series programming on the big four networks. Meaning, no sports, no reruns, no specials, no movies, no sustaining programming included.

No comments:

Post a Comment

© 2009-2018. All Rights Reserved.