Commentary
GRPs For Digital Marketers
- by Josh Chasin , Op-Ed Contributor, April 21, 2010
Every couple of months, I turn a corner and find myself smack dab in the middle of a debate about the efficacy of the GRP as an online metric. It usually goes like this. On the one side: "Advertisers need GRPs to place digital media in holistic media context." On the other side: "Dude, whatever. GRPs are, like, totally last century. We can't shoehorn the power of the Internet into an old metric."
I'm going to try and get us all past this. GRPs are a necessary, but in no way sufficient, metric for evaluation of online advertising. If you're in a hurry, you can stop reading now. If not...
Gross Rating Points originated in broadcast media. First radio, and then television, had audiences measured at the program level: how many listeners or viewers were in the audience for a specific show? The result was a program rating. On January 19, 1953, for example, for the episode "Lucy Goes to the Hospital," "I Love Lucy" garnered a 72 household rating -- 72% of all TV households in the U.S. were tuned to the program (at least, according to Art Nielsen.)
Advertisers, of course, ran schedules, which were simply collections of spots. Each spot ran in a program, and so could be associated with that program's audience rating. The GRP emerged as a way to express the audience to an aggregate of spots, which is to say a schedule, and its calculation couldn't have been more simple: the sum of the program ratings for all the spots in the schedule. If an advertiser bought 10 spots across 10 different programs, and each program had a 7 rating, then the Gross Rating Points -- the sum of the ratings of the spots in the schedule -- would be 70. If an advertiser had run two spots in that landmark episode of "I Love Lucy," they would have bought 144 GRPs (which would be parsed as a reach of 72, with a frequency of 2.)
By the way, this was the answer my Advertising and Media Planning professor was looking for: "Gross Rating Points are the sum of the ratings associated with all the individual spots in a campaign." There. After 30 years, I've finally nailed it.
When media choices were limited and broadcast advertising was new, no one thought about moving beyond the GRP metric; you simply bought a ton of advertising, and it worked. And if it didn't work, you fired the agency. GRPs were, and are, a measure of tonnage.
Over time, broadcast audience measurement got more granular. In the '60s, household ratings gave way to demographic person ratings. Program ratings were replaced by average quarter hour ratings (how many people were in the audience during the average 15-minute period) and ultimately, for network TV, by average minute ratings. The idea of moving beyond the program rating to the average minute rating was to make the audience estimate approximate as closely as possible the potential reach of an actual commercial: If 20 million people watched a program at some point during the hour it aired, but in an average minute only 15 million were watching, then 15 million was a better estimate of the audience an advertiser could reach with a single commercial.
Now, you digital folks out there, whom I envision clustered around the virtual campfire listening with bated breath to my tales of the olden days, are probably thinking, "Program ratings? Commercials? This just proves that the GRP is totally irrelevant online; we sell impressions, and they are served one at a time, not broadcast."
Good point.
It turns out, though, that because ratings are expressed in terms of the population (a 7 rating against a particular demographic means that 7% of that demographic was reached), Gross Rating Points have properties with respect to the population. And because the population doesn't change across media (there are 300 million people in the U.S. regardless of whether you work in TV, radio, print, or digital), the GRP has become a metric of cross-platform comparison. So here's another fact about GRPs that you should know: 100 GRPs is equivalent to the number of impressions you need in order to reach everyone in the population one time. If you think about it -- and this sort of thing doesn't give you a headache -- that's really pretty obvious, because ratings are percentages against the population, and GRPs are individual ratings summed up.
So if your digital target is women 18-34, and you know you bought 26 million impressions against that target online, then guess what? Since the population is known (about 35 million women 18-34 in the U.S.), you've just bought yourself 74 GRPs of women 18-34. (That's impressions divided by population times 100.)
(Note to Dave Smith: Yes, the total population, not the online population.)
The question of whether GRPs are a relevant metric online is simply the wrong question. As long as you can know how many impressions you bought against a target, and how many people comprise that target, your GRPs are, a priori, known. (I'm taking for granted that you can divide GRPs by the population and multiply the result by 100, but chances are you or someone in your organization has access to Excel.)
GRPs are a known, fundamental, derivable measure of the tonnage of advertising bought, and the metric allows that tonnage to be compared across media (I bought 200 GRPS of TV and 75 GRPs online.)
A better question is, are GRPs a sufficient metric for evaluating digital advertising?
And of course, the answer is no.
Back when advertising choices were limited, the GRP told you most of what you needed to know. But with today's fragmented vehicle audiences, and with the complex processes advertisers deploy in moving prospects through the funnel, the GRP tells you less than ever before. Today we need to know -- and fortunately, have ways to express -- measures like engagement, awareness, purchase intent, branding effect, ROI, and wear-out. Of course most of these metrics were developed and refined during the TV age (think of IRI and How Advertising Works.) But in the digital age, their importance relative to measures of sheer tonnage becomes even more important.
Looking ahead, the advent of social media will undoubtedly introduce new advertising and campaign metrics (including, for example, resonance and virality.)
But as I say, as long as you can know impressions and population, you've got GRPs. So the next time a "Does digital really need the GRP?" debate breaks out, let's nip it in the bud. GRPs are a necessary, but in no way sufficient, metric for describing digital advertising. Next question.
Count me in with the "digital folks out clustered around the virtual campfire listening with bated breath to tales of the olden days". With that being said, it was a very informative read
I commend the intent of the article. For years I have told anyone who would listen that all forms of communication could be converted to rating points. Every medium has an audience and every audience is a % of a larger universe.
One correction: 100 GRP's is NOT equivalent to the number of impressions you need in order to reach everyone in the population one time. It simply represents the number of rating points necessary to generate impressions equal to 100% of the universe. It could be 50% 2x or 25% 4X.
Allen - I disagree with your correction. Let me ask you a question - how many GRPs are needed in order to reach everyone in the target population ONE time? Answer - 100. The question is not how many are needed to reach everyone in the population two times, or four times.
Love this article, thank you!
This brings back fond (?) memories of broadcast television media planing processes in which we would use one-size-fits-all algorithms to project reach and frequency (and then "effective reach") from essentially two inputs: GRPs and daypart.
It was always such a SWAG dressed up in a agency-blessed procedural which attempted to convince oureselves and our clients that we had a way of divining the truth.
Then again, we had nothing better.
No offense, Josh, but you don't sell impressions. You sell potential impressions, of very dubious potential at best.
You can't have it both ways. You can't tout the click through as the coin of the online realm (as evidenced by so many early arrivers using "click" in their names) and then ditch it when it becomes counterproductive only to again champion the same soft metrics that you found so loathsome just a couple of years ago.
Furthermore, if recent trade press articles are to be believed, a huge quantity of these so called impressions are fraudulent, which means they leave no impression whatsoever.
The sad fact is that there are now more places for ads to make no impression than there are people to impress.
Am I misunderstanding something? It doesn't make sense to me to equate online ad impressions, which are not unique at the surfer level (the same ad can be and generally is served numerous times to the same surfer), to TV commercials, which are unique at the viewer level (viewer = a TV set being watched by target population) doesn't make sense to me. To me, it simply does not seem reasonable to claim that by serving 26 million targeted impressions you have reached 74% of a targeted population of 35 million people--people who may or may not even surf the sites where the ad is being served.
Kelli...Theoretically 100 GRP's could reach the entire target universe. Theoretically we can travel at the speed of light. Theoretically we can cool an object to "absolute zero". In reality these are not achievable. I am not aware of any R/F program which can generate a 100 reach. Someone is either in a coma, or out of town that day, or had an electrical blackout. I am never comfortable telling a client we can reach "everyone".
Thanks for the lively feedback. BTW, is/was your agency a member of of the 3AI agency network. If so, I knew/worked with Karen...your media director from the late '80's.
Well, Josh -- seems like we are on the same page with this one. I wrote a similar article last August: http://bit.ly/E78ZD
Kelli, you're wrong. 100 GRPs will never deliver 100% reach of a target audience because there will always be those in the target who just aren't exposed to the media schedule. 100 GRP's deliver the EQUIVALENT of your target in terms of impressions but not reach. If you want to get close to 100% reach you'd have to buy hundreds or thousands of GRP's. (Even with 1,000 GRPs you'd probably get at best a 98 R/10.2 F and that depends on the media mix).
If I were a client being told by my agency that my media schedule was reaching 100% of my target with 100 GRP's I'd fire them.
Thank you. Well said.
Rob is right. I said 100 GRPs is the equivilent of reaching everyone once. In practice, yes, that probably plays out to 25 X 4, 33.3 X 3, 18 X 5.56. And so on...
Thanks everyone for the lively comments.
Good post Josh. There is a free tutorial on how to calculate TRPs (Target Rate Points) for online display and in-game advertising media buys at SmartDigitalSpending.com
Download at http://bit.ly/aCybJC
If your ad spend is 100% online, GRPs should only be used when no other metrics are available--they can only be estimated very roughly. How do we know what the overlap is when we buy 100 spots with 10% reach each? Same problem with impressions, how do you know the difference between 1 guy viewing an ad 100x on different computers vs 100 people viewing an ad 1x?
For the marketing mix though, we must always strive to increase our reach across online and offline populations. For this, GRPs are very helpful.
Regards to your class: Your so called teacher wasn't a professor or even a teacher regardless of what the university title was. Instead of worrying about publishing crap, all teachers in actually any business field should have the majority of their income come from the subject they are teaching. The pace of the market runs so fast that someone who spends most of their time in a classroom misses the points they are supposed to be teaching. Would not be surprised if you went back to marketing/advertising classes now that you would find the same problem. Why do you think so many people reading and writing for MediaPost complain about the help. GRP's are a gauge, not an absolute. And past program performance does not necessarily predict future program performance or new program performance. Note; Program = any form of information on any media.
Demo target ad impressions / Demo target population X 100 = GRPs. This formula works the same no matter what media channel is involved. The critical thing to calculating GRPs is to make sure that the ad impressions in the numerator are consistent with the population in the denominator. That is the only hard part for people who do not have any media training to keep track of - and most digital media buyers do not have any media training so they often prefer to conclude that GRPs are obsolete even if their clients feel they are important.
GRPs can be calculated on the back of an envelope by anyone with 9th grade algebra skills.
One of the elegant things about GRPs is that Reach X Frequency ALWAYS matches the GRPs even though there are derived separately.
Reach/Frequency is a complex estimate that requires a database that can model for the DUPLICATION of advertising impressions. Combining the reach achieved through one media resource with the reach achieved through another for the same campaign in order to estimate total reach requires some advanced algebra and a working knowledge of statistical probability. It is hard to do but we did it for TV by hand before computers came along in the 1980's. Computers do it today but the concepts are the same.
No legitimate media strategist would build an advertising campaign without keeping a close eye on the GRPs and R/F for EVERY media channel. This metric is not the be all and end all of media planning - it is the foundation on which any and all other metrics are over-layed.
Thank you Rob for your correction....EQUIVALENT is most certainly right!! No media can reach 100% of your target audience.
Allen - Yes! Small world
I remember OMMA Performance Microsoft Advertising's strategist actually espoused (at a performance marketing conference no doubt) the importance of GRPs. Now, to very big advertisers whose media budgets are reapproaching tonnage levels, that's an interesting pitch. But as everyone's agreed, the focus cannot be on impressions until the value of impressions is agreed upon and we can't agree on impression value yet because not everyone's using view-based tracking systems or following lifetime value.
To add complexity, the display media market is shifting so rapidly to demand-side platforms and dynamic bidding that we are in for another 5 years or more of experimentation--on good old fashioned banner ads alone!--setting aside the value of mobile, social ads, etc.
Have fun with your GRPs Microsoft!
Woo hoo Josh. That sure fired a few people up!
I must say that while you encountered GRPs via TV, and as you correctly point out they started with radio, they are actually the bedrock of COMMUNICATIONS planning. Into a brand's communications plan we then shoe-horn the various media. So, if online wants to be seen as a part of the communications industry then GRPs is the lingua franca. Those that insist they are different because they are in the online sphere, have been blinded by the proliferation of online data (of which we now know SOME may even be remotely close to believable given the poor metrics that have abounded there over the past 15 years).
But I want to sow a seed to think about. I believe that the relationship between GRPs and Total Reach is stronger in TV than it is online. The reason I say this is that many (most?) ad campaigns, well at least here in Australia, tend to have one spot per programme, often due to cost. However, in online with individual impressions being served (at costs as cheap as chips), I suspect that with the targeting ability of online it is far more common, if not indeed prevalent or the norm, for the same person to be served the same ad impression many times. That is, the average frequency online tends to be much higher than on TV. So while, GRPs can (and should be compared) across media, should we not be extremely careful to not equate these as having equivalent reach - which is what I suspect many people would do. That is, your 74 GRPs online against W1834 would tend to deliver lower reach than 74 GRPs in a TV campaign. Further, 74 GRPs in a magazine campaign using highly unduplicated titles would deliver even higher reach than either TV or online. A 74 GRP mail-shot should deliver the whole 74% reach!
The moral of my ramble, is that most advertisers want to buy reach. They know that smashing their ad over the head of their target is counter-productive, so maximum reach at lowest frequency becomes the mantra (unless you are launching a new product of course). Of course, it also should be cheaper as it would be buying less weight - a win/win!
I'm with @mike einstein - this whole house of cards comes tumbling down with the faintest puff of post digital media winds... "impressions" as the core foundation for truly measurable/valuable online media truth is a big fat fail, imho.
I must say, I'm somewhat surprised that this, of all columns, has spurred such thoughtful and heated discourse.
As per John Grono's assertion that online media are prone to excessive frequency delivery, we've done some analysis of frequency distribution for campaigns, and have found instances where a significant portion of campaign impressions go against persons exposed over 10 times; so that if one accepts the premise that effective reach is between 3 and 10 impressions, these campaigns would be better "served" (pun intended) if some of those impressions could be shifted to new eyeballs. (One of the problems with cookies as a means for frequency capping, of course, is cookie deletion and rejection. Well, I guess that's two problems.)
Conversely, our Chairman Gian Fulgoni (who has been trying to horn in on my turf by publishing in the Online Metrics Insider) published an interesting column at the comScore blog in which he demonstrates that there is measurable impact accruing from frequency even past 10 exposures:
http://blog.comscore.com/2010/03/impact_of_ad_frequency.html
Of course one might argue that not all so-called impressions are in fact exposures (as per Mr Einstein below.) It looks like there is about to be increased uptake for online ad verification services-- third parties that can quantify that specific ads were served to people and appeared "above the fold." One wonders how the digital community will react to that...
Nice article, but have you thought of the differences on the impact. A full screen spot on tv has a much bigger impact on the target group than a leader board or a MPU on a website.
Even though you have the same calculated GRP level with all your different media plans, the impact is different. We do online planning on GRP level to be comparable with other media but we weigh online GRPs in Germany only with 0,75 (depending on the ad formats used). That makes it more comparable with TV.
Good piece Josh. It's always nice for the two of us to have common ground.
Hey Josh,
Look, impressions, exposure, GRPs...call it what you want, the fact is, they're virtually worthless. And that's not my opinion (although I totally agree), it's yours! How else do you expain the fact that these very impressions are bought and sold by the millions every day for under a buck a thousand?
This is like Woody Allen's routine about the two old ladies in the Catskills: "The food isn't very good this year." "I agree...and such small portions."
As far as "...increased uptake for online ad verification services..." goes, I predict that more and closer scrutiny will only prove what logic already dictates: In an on-demand world, no one demands more advertising. That's why your attempts to measure performance always end up measuring failure instead. Now you're measuring the metric. Good luck.
Josh, thanks for alerting me to Gian's blog - it must have come out while I was on holidays in Antartica (low internet penetration down there!)
The chart that shows increased visitation and search queries at the 10+ exposure level is very telling.
If you looked at the 'response curve' for the other media (with their varying quality audience measurement) you will see declining response with increased discrete frequency. That is, the first exposure delivers the highest response and it is all downhill from there. Well, apparently not-so in online.
However, I think that just maybe our definition or measurement of exposure in the online world may be the problem here, as you allude to.
How many of these 'exposures' are being served to machines that are actually bots or crawlers? Of the remaining 'exposures', how many are being served to inactive browsers? Of the exposures left, how many are being served to the active tab on the active browser? Now that we're down to the exposures that hit an actual screen with a person sitting at the computer, we need to consider was the ad above the fold, did the user scroll down to the ad, etc. Oh, and finally .... did they actually then notice the ad.
I'm hardly surprised that it takes so much frequency based on the existing (inflated) metrics to get a response! I'm also hardly surprised that click-through rates are where they are at, why CPMs are at rock-bottom prices. Of course, fixing the denominator in all these guidelines would go a long way to aligning online to what behavioural scientists see as the response in all other media.
As a good mate described it to me ... buying online impressions is like buying Zimbabwean dollars with it's $100 trillion bill!
Our digital industry definitely needs new standards moving away from the traditional CPM which is a printed oriented standard. Our company WWN in cooperation with Alenty is now offering new metrics such as Brand Exposure Duration / a mix CPM taking into account length of exposure for each impressions served.
If you are still selling Impressions (or worse, POTENTIAL impressions)- get out of the business. As a buyer I can tell you that impressions don't buy my products or visit my business; actual customers do. So pitch me clicks, responses or actions. I don't want another rep telling me how many impressions I can get! Get it? Give me Action not words!
Kelli,
If you have a 50% reach with a 4 frequency = 200 GRP's
Does that mean I have reached the population 200%???