Kabrich said:
Let's take it from the top.
let's. This is the type of dialog that helps find solutions.
1) PPM is MRC accredited in Houston only - not Philadelphia. The sample size for Houston was signed off on by the MRC. Arbitron has fallen short of those sample sizes. Even Steve Morris admits that in the letter.
I'm not arguing, just observing and clarifying: this is a Houston board, and the issues I'm adressing are Houston related, as I don't get the Philly data and can't see the sample reports. The Houston daily in-tabs have only been affected since Arbitron attempted to improve carry rates and ended up losing total panel size. The sample since January looks good (not perfect) except for the last weeks of June and July and, of course, the first weeklies for August.
Should Arbitron have first had new panel housholds ready? Yes, and I can't understand why the problem escalated unless there was a daily compliance crisis we have not been fully informed about that made booting households necessary. In theory, a panel study has replacement panelists ready at all times in all stratification variables. Obviously... and the Arbitron folks are good researchers... they did not see this wave coming, and the fishing boat sank.
2) Arbitron's effort to move out households from the panel that do not carry the meter is a joke. This could have been accomplished months ago - and in the latest weekly numbers we see no signs of the turnaround they have promised since late May to occur by September.
The situation was certainly amplified by the onset of Summer, and crashed when combined with the panel daily carry rates. I saw specific instances of problems from late in April; I don't know if Arbitron thought they would be easily fixable then. The fact that they added staff, as Steve Morris said, indicates that they had to make major changes in the recruit and compliance functions.
On the other hand, I would not expect the fix-up to be instant, as bringing new panelists on board takes several weeks, even when expedited. In a couple of areas, August Week 3 has some improvements, but those could be wobbles, too. Obviously, this wobble-less panel thing is not truly without wobble.
In fact in Philadelphia 18-24 Females In-Tab fell by 10% just this week alone - and this was one of the demos they were focusing on improving. In Houston they are down 20% in the last 4 weeks. And please don't come back and say, "but they are eliminating households who do not carry the meter consistently". You don't eliminate those who carry the meter in this process - especially when they make up 4.8% of the sample and 5.6% of the population.
Not to take an Arbitron defense posture, there are two reasons for panel problems here... the elimantion without prior replacement of bad households, and then there is the possiblity of high natural attrition from the panels. Both should be instantly solved by backup households, though. However, the panel is household based, so what we have with 18-24 is non-compliance by certain members of a household. I'm totally guessing, but arbitron may have been optimistically trying to get one or two family members back in compliance so as not to lose the whole household, and failed.
Even Males 25-34 and Males 35-44 are WAY OFF the population (about 1.5 points and 1.0 points respectively) or, 20% and 10% compared to the in-tab.
The overall issue in specific demos is whether the remaining sample can be weighted into proportionality within a tolerable margin of error. 18-24 women is one area where i think the sample was below reasonable weighting based correction.
3) The lower in-tab has happened since the 3rd week in March and in most demos within several people of the lowest levels ever - even today. Arbitron in their "don't question the king attitude" told the trades this was due to Seasonality which they stated was seen in the diaries with lower response rates in the Summer.
I see the most serious issues in April, starting around week 3. Again, the issue is whether Arbitron thought this would come into compliance using the techniques in use at the time... in any case, April or March, we were still in the test non-currency phase. So, the fact that broadcasters getting the data did not start screaming then may make us as responsible as Arbitron.
However, once a real trade reporter looked at the 20 year response rates, he found that that was not true in Houston and Philadelphia - surprise, surprise, Arbitron Spinmaster lied and was busted.
The did not have this issue in Summer of 2006, although the panel was not 100% complete then. I am willing to give a bit of benefit of the doubt to the difference in carrying a meter in the Summer and filling out a diary on the kitchen counter in the evening.
Thus, again, Arbitron's Quality Control was lacking at best - and then they had to come up with a new excuse. Think about it - they have over 90% recruited for the New York Market 30-45 days ago for a market that hasn't gone live yet - yet they cannot get the sample up to speed in the live markets of Houston and Philadelphia (where they have been "testing" the meter for 6+ years???? ). Think about it real hard.
The panel was never expected to be 100% perfect in the two-year test of Houston... and one thing nobody kenw... the reason the test was repeated in Houston... was the unknown factor of Hispanic compliance. It's interesting that Arbitron does not give in-tab data by demo for any ethnic group, something several NABOB members mentioned to me in Columbia two weeks ago.
4) There has been no drop in TSL? ROFLMAO. Do you know how ratings are formulated? You take Cume with TSL and come out with ratings and shares. Arbitron has publicly touted how great this is because the Cume levels are SO MUCH HIGHER. So, what new math are you doing that says you can take a Cume that in many cases is 100% higher for a station and keep the same TSL and come out with ratings that drop 30%?
Yeah, I do know how ratings are tabulated. I wrote ratings software for the company that surveyed Puerto Rico prior to Arbitron's entry. But I never said TSL had not dropped. It has dropped about 40% from the diary, because listening time per person has gone from 19:15 in the last diary study (if I remember right) to closer to 11 hours in the PPM. But increases in cume based on below-P2 preference levels always reduces TSL. It's the reason why big cume promotions that work often affect TSL negatively as "testers" of a station listen less until they leave or become regular users.
The 5 to 10 hour TSLs of the diary are 2 to 4 hour TSLs in the PPM, and the range from highest TSL (4 hours in August week 2 among top 15 stations) and the lowest is less than 2 hours.... vs. 5 to 6 hours for the diary. But cume and TSL are all a meter detects. You can get the same share with low TSL and high cume or low cume and high TSL. The PPM obviously compresses TSL, so shares are compressed, too.
5) The fact is the agencies are not getting what they think they are getting. There is so much missed listening by the PPM that it is ridiculous. In fact if you want to have fun with numbers in the 16-18 hours an average American is awake, 33% of the PPM meters are not included in an average day, of the Additional 66% all it takes is 8 hours to be included, thus it is possible that of the 66%, 33% undock for 8 hours and 33% dock after 8 hours. As thus, you could have upwards of 66% of the average day listening missed by the panel. Last time we heard, the median undock time for Houston was 7:41AM - in a market with the longest commute time in America. If you are a morning drive personality, don't expect the daypart to be worth the big bucks you were getting in the past - as, simply stated, mornings are not that important any longer.
Remember, the edit rule puts all detection one hour prior to undock into the tabulation, although I believe that this is unduly short; many families with children spend as much as several hours getting up, getting children ready and then leaving for work themselves. I have questioned how the decision to create this edit rule came about; I would find that a few detections by people who sleep with the radio on would be less evil than making an arbitrary edit rule. I have not had an explanation for my question of how this rule was arrived at, though.
6) There is no issue of ratings stabilizing? You really have chugged the Kool-Aid. Arbitron is using Rating Points to make this arguement - and rating points are not share points (rating points are much larger and thus less prone to change) . In fact, it is possible to see stations move 8 or more rank positions in share ranking over several weeks with the same panel in place.
The move in rank si due to compression. A change of 0.2 to 0.3 can move rank by 5 or 6 positions when there is a much tighter range of shares in the top stations than there was before. I don't think that movements of that magnitude are errors per se, as it is unreasonable to think that every station every day has absolutely identical audience levels in a market with so many options.
InTIMadate is right on the money that agencies are sitting out waiting to see the ratings stabilize (in fact, some are still using the old diary numbers for buys - Arbitron has sent out cease and desist letters when it finds out about them). Most want radio to drop their rates to make up for the rating point shortfall.
I can't see how Arbitron could sanction agencies for using released data of any kind. In fact, agencies generally use 4-book averages so this would not be something Arbitron could even do.
Regardless, once again, the spinmasters tried to obscure the facts on Philly's Revenue numbers, saying the fluctuate every month so they are not sure of a trend. Unfortunately, they are looking at it through the eyes of a single market. In Philadelphia for example, the market went live with PPM numbers released in mid-April 2007. The first full month of PPM was May. For the first time in 25+ years (and most likely forever) in May 2007, Miami's Revenue beat Philadelphia's Revenue, something the locals the reporters who spoke to the local stations did not know. In fact, Miller Kaplan's show that Miami has now beaten Philadelphia in terms of revenue for every month since the PPM went live. That is major.
That's an invalid comparison. Miami is a growth market, fueled considerably by the organic growth of Hispanic media; Miami is roughly half Hispanic and the Hispanic stations are having an incredible year there. Philadelphia has been in declining revenue since 2004, with small but significant decreases to 2005 to 2006. That 2007 is down would be un-surprising. Miami has grown over 20% in revenue since 2001, so at some point Miami was going to surpass Philly, with or without the PPM.
We all know that Urban stations do not fare as well with PPM. Radio One gets 45% of their revenue from National business which is more rating sensitive and last week they were pacing down 22% for September with the PPM going live in mid-July.
Staitons that appeal to groups that may have done more rounding in the diary are very subject to a methodology that is to-the-minute precise. I know form more diary reviews than I can remember that Hispanics tend to round to the hour and, at best, the half hour. This may be the case with ROIAK's audience, too, but I have not looked at the data. A stem and leaf charting of diary incident start and stop times shows that Hispanic listening times were considerably rounded up at both ends, so the change is not a surprise... it's a defect (or "bonus gift" of the diary method).
Finally, to think that any survey will not have some weighting is ingenuous. But we do need much closer goal complaince.
