Monday, June 6, 2011

True Strength Begins: Viewing Levels 101

I've been working on a large intro to this project, but I think what I have really confuses things at this point, and most of these posts should pretty much stand on their own anyway. So rather than hold up the whole deal till I get the perfect intro, we're just gonna dive right in. The short version, as said in my pre-summer post, is that I'm going to spend a lot of this summer trying to figure out how various factors affect TV ratings. Those factors include but aren't necessarily limited to: viewing levels, competition, time of day, time of year, day of the week, lead-in. And a lot of those things overlap. I think I'll post the more formal intro within the next few weeks, as there are some things I need to establish, but the first few posts should be OK without that.

Disclaimer - I'm getting the viewing level posts over with right off the bat because, as I say on the daily posts when I post the viewing levels, these numbers are inexact and not supplied by Nielsen; I have to estimate by combining the ratings divided by shares of everything in a timeslot. Because of that, I try to minimize the error by using averages of as many data points as possible. But there's still going to be some margin of error. So what I recommend paying attention to in these posts is the broad strokes, not the exact figures. This is most useful as a set of general guidelines and for comparative purposes.


The very most general stat: in the average primetime moment of the 2010-11 regular season, about 35% of TV-owning adults age 18-49 were watching their TVs. But that's far from a consistent number in every moment. We know that differences in viewing levels (how many people are watching TV at a given time) matter to TV ratings just by looking at how TV schedulers treat them. Fewer people watch TV in the summer, so there's a lot less big-time programming in the summer. Fewer people watch TV on Friday and Saturday nights, so there's a lot less big-time programming on those nights. It might seem a bit of a self-fulfilling prophecy ("If they put the big stuff there, the viewing levels would rise!") but that's only true to some extent. The networks' tendency to "give up" during those times does widen the gap, but there'd be a gap anyway.

Day of the Week

One of the most obvious discrepancies in viewing levels is the difference between Friday/Saturday viewing and everything else. During the 2010-11 season, the average percentage of TV-owning adults 18-49 watching TV from 8:00 to 11:00 on Sunday to Thursday is 36.5. On Friday and Saturday? Just 29.9. That's a dropoff of 18% from the weeknights to the weekend nights, or a difference of 6.6 percentage points.

Even within those two types of days, though, all days aren't created equal. Friday and Saturday are about equal, with Saturday also a bit boosted by (college) football, but the weekdays still vary a decent amount.

Sunday 38.7
Monday 37.5
Tuesday 36.2
Wednesday 35.4
Thursday 34.8
Friday 29.8
Saturday 30.1

Now, a lot of the advantage for the first two days comes from football (more on this later), but not enough to put them on completely level playing fields. Sundays averaged 37.6 since the week after the Super Bowl while Mondays averaged 37.1 since the week after the BCS National Championship game. So it seems that even without those huge external factors, viewing does seem to peak on Sundays, trickle down through the rest of the week, and then leap downward on Friday and Saturday. I find that kind of interesting because we always think of Thursday as being such a huge day for advertisers. There are so many big hits on the night and we almost consider it some kind of "honor" when a big show gets moved to Thursday, yet that day has slightly lower viewing than any of the other "big" days.

Time of Day

Before I made my ratings bible spreadsheet and started coming up with viewing level estimates on a daily basis, I just had this notion that primetime viewing was symmetrical. I figured that if 8:00-11:00 was determined to be "primetime," that meant viewing probably peaked right at 9:30 and rose evenly up until that point and declined evenly from that point on. I had no real reason for thinking that, it just made sense.

It turns out it's not quite so symmetrical. I only have access to half-hour breakdowns for all the primetime ratings from 2/28/11 thru 5/21/11, but looking at hour-by-hour breakdowns across the whole year, they seem to line up somewhat well with these. Here are the half-hours.

Sun Mon Tue Wed Thu Fri Sat Week
8:00 34.0 33.1 32.0 31.8 30.6 26.6 26.1 30.6
8:30 36.7 36.2 35.1 35.0 33.9 28.9 27.6 33.3
9:00 39.1 37.9 37.3 38.4 35.6 30.3 28.5 35.3
9:30 40.3 39.4 38.6 38.4 37.2 31.4 30.5 36.6
10:00 39.4 38.9 37.3 38.2 37.0 32.1 31.3 36.3
10:30 36.6 37.0 36.2 36.4 34.8 32.5 31.5 35.0

So there are a couple clear patterns here. And if it's not clear, lemme break it down another notch.

Sun-Thu Fri-Sat
8:00 32.3 26.3
8:30 35.4 28.3
9:00 37.7 29.4
9:30 38.8 31.0
10:00 38.2 31.7
10:30 36.2 32.0

So there are two basic patterns. The one that covers most of primetime is the "Weeknight Pattern." This one is sort of close to that symmetrical notion I had, but it's much more heavily weighted toward the second half of primetime. Viewing doesn't seem to peak right at 9:30; if that were the case, you'd expect the 9:00 and 9:30 half-hours to be close to even. Instead, it seems to peak sometime during the 9:30 half-hour, perhaps in the second half (sometime between 9:45 and 10:00). It goes down in both directions from the 9:30 half-hour, but there's still a slight edge to the later ones (with 10:00 a little higher than 9:00 and 10:30 a little higher than 8:30), which is why I say the peak moment is probably sometime after 9:45. The least-viewed half hour, and it's not even close, is 8:00. So there's definitely an inherent viewing disadvantage for shows that air in that half-hour, especially for the shows (like How I Met Your Mother, The Middle and The Big Bang Theory) that air exclusively there.

Then there's the "Weekend Pattern" for Friday and Saturday nights, where viewing actually continues to increase throughout the evening. It slows down in the second half, and it seems like it's probably declining by the time you get into the 11:00 half-hour, but it's still clearly different from the way the other five days operate. We'll get a little bit more into this in the next post when we break down these percentages into broadcast & cable, but for now I'm just saying how it is.

The Two Methodologies 

Before we get into the next post about viewing across the year, I need to acknowledge that Nielsen made a change in their definition of "viewing levels" late this season. Here's a note from a TVByTheNumbers daily ratings report (you can find this on any of their daily ratings pages):
NOTE: All ratings are “live plus same day” from Nielsen Media Research unless otherwise indicated. Starting with data for Mon 3/28/11, all program ratings will now include multiple viewings of a telecast on a DVR. Previously only the first viewing was counted. HUT/PUT will now consist of Live TV viewing and DVR playback at time of play. Previously, HUT/PUT consisted of live viewing and DVR playback credited to time of live telecast.
Most people don't (and probably shouldn't) care about the HUT/PUT methodology change because it only affects HUT/PUT figures (which are only occasionally available) and share (which is always available but most people don't care about). It doesn't have any effect on program ratings, which is all most people really care about. But since I'm looking at viewing levels, I have to take it into account. So here's the best way I can think to explain it:

Ratings measure shows, not strictly time periods. Yes, there are breakdowns by half-hour, 15 minutes, minute-by-minute, etc., but those are breakdowns of the duration of the show, whenever those pieces of the show are being watched (live or via DVR). This was true before the March 28 changes and is still true. (The only difference to ratings is the multiple DVR viewings change, which I don't think has made much of an impact, at least within the Same Day window.)

Pre-March 28, persons using TV (PUT) calculations were also attached to shows, and then the shows were attached to their live timeslots. When you viewed a show on the DVR, you counted as a viewer of the show and you also counted as a viewer of TV in the timeslot where the show originally aired live. So if you're watching live, it worked out, but if you watched The Middle on DVR at 10:00, you counted as a viewer of The Middle and a viewer of TV at 8:00. Now, you always count as a TV viewer when you're actually doing the viewing; so in the above example, you count as a viewer of The Middle and a viewer of TV at 10:00. So now, all viewing level measurements are truly measurements of how many people are watching TV (either live or anything they've DVRed from earlier in the day) at a given time.

That may seem like a better depiction of reality, but the "share" statistic has become even more apples-to-oranges. The calculation for a Live + Same Day share used to be:

People watching the show live plus people watching the show later in the day via DVR                
People watching any live TV plus people watching currently airing shows later today via DVR

And now it's:

People watching the show live plus people watching the show later in the day via DVR               
People watching any live TV plus people watching today-DVRed shows right now

Maybe I've made it more confusing, but the bottom line is that I think "share" is arguably even less helpful than it was before purely as a measurement of show strength, but it can still be applied the same way to get viewing levels. And viewing levels are now, perhaps for the better, measurements of "everyone watching TV right now" rather than measurements of "everyone who watches the TV that's airing right now." (I say "perhaps" because the old methodology provided a nice way of gauging how much TV that airs in a timeslot actually gets viewed; now, you have less of a sense of what people are watching, you just know better when they're watching.)

But let's forget about share. It's just a tool to get to viewing levels. So how's the methodology change affect viewing levels? This next table shows viewing level differences between the four weeks when I have half-hour breakdowns before the methodology change (February 28 to March 27) and the eight weeks after (March 28 to May 22)

Sun Mon Tue Wed Thu Fri Sat Week
8:00 -0% -1% +1% -2% -1% -1% -2% -1%
8:30 +3% +4% +6% +2% +2% -2% -2% +2%
9:00 +4% +8% +7% +5% +6% +3% -1% +5%
9:30 +7% +9% +11% +8% +8% +5% +3% +7%
10:00 +9% +9% +14% +8% +8% +6% +7% +9%
10:30 +11% +11% +13% +16% +8% +7% +4% +10%

So this change in methodology has clearly "juiced" the viewing levels a bit, especially considering you might expect viewing to drop a bit in the late spring. The change is about +5% when combining everything, about +6% from the Sunday to Thursday period. But again, all timeslots are not created equal. It's clear that the farther you get into the evening, the more this methodology change increases viewing levels. In other words, by the time you get to 10:30, there are a lot more people watching things previously DVRed than there are people watching 10:30 shows later on via DVR. Which certainly makes sense. And it slightly skews the tables in the "Time of Day" section; it's made the increases at 8:30, 9:00, and 9:30 steeper and the decreases at 10:00 and 10:30 less steep overall than they would've been using the old methodology the whole year (or vice versa if we'd used the new methodology the whole year).

Anyway, this methodology thing is probably more relevant to the second post, where I'm gonna look at viewing levels throughout the year and the impact of stuff like football on viewing. Generally speaking, when comparing viewing at different times of year, I'm going to have to try to avoid comparing pre- and post-methodology changes (or come up with an expected adjustment). But I'm throwing it in here because it fit well with the timeslot stuff, and also because the next post will probably be pretty long too!

No comments:

Post a Comment

© 2009-2022. All Rights Reserved.