May 11, 2004
SUMMARY: This is must-read info if you run any direct response ads online (or plan to in the future.) An Atlas DMT study reveals that a solid third of all online ads are being served up to viewers who've already seen them at least nine times.
However, after five impressions (the same person seeing your ad five times) response plummets.
So one third of your media buy is reaching people who aren't remotely likely to click. Yes, this article includes three specific suggestions to help you begin to solve this problem:
On average, 33% of all online ad impressions occur after a user has already seen an ad campaign 10 times -- which is a waste of the advertiser's money, according to Young-Bean Song, Director of Analytics, Atlas DMT.
"If you have a choice between delivering a user a 20th ad impression, and taking that impression and giving it to someone else," Song says, go for the new user. "If the user hasn't converted by then, giving them another impression is a waste."
For example, if a campaign had an average frequency of five impressions per person, don’t assume that there would be very many people who saw five advertisements.
Usually, the actual impression distribution is extremely lopsided. If you do see the data, you may notice that there is a small group of people who have been exposed to your campaign many, many times. They are wasting your media plan.
With direct response campaigns, those wasted impressions could potentially mean a big loss of revenue. Song wanted to know at what point impressions stop being effective.
"A lot of people use the term optimal frequency, what's the right number of ads to deliver to a given audience," he says. "But it's mostly used as far as branding goes; nobody's really done that for the direct marketing side."
So, last fall he ran an Optimal Frequency study that looked at each frequency level for 38 advertisers running direct response campaigns in order to find the conversion rate (conversion to sale, to fill out a lead generation form, or to register) at each specific level (i.e.: each time a specific user saw a specific campaign).
Ameritrade was the first advertiser to change media buying based on study results. "When they implemented the recommendations [from the study], they saw a 15% increase in total conversions from the campaign. Same campaign, same messaging," says Song.
Here are more useful details on the study learnings and recommendations:
Key Learnings: After Five Impressions Response Plummets
Learning #1. First impression had highest conversion
"The first impression conversion rate was almost three-and-a-half times higher than the overall conversion rate," Song says. "So they were getting the biggest bang for the buck on the first impression."
However, the first impression isn't *always* the charm. While it generally makes the most impact on a cold impression -- someone not used to purchasing that type of product -- for users more ingrained with a certain type of product or purchase, conversion may be more about opportunity and timing.
"It's always a moving target," Song says. "There are always people not ready to go to Disneyland one week and then are ready to go the next."
Learning #2. Optimal frequency is five
The point of diminishing returns (the frequency level at which the cost per acquisition makes it no longer profitable) was after the fifth impression. "You're still getting incremental conversions but they're more expensive conversions."
Learning #3. First three impressions all had at least 100% lift
In aggregate, each of the first three impressions was over 100% better at converting than the average impression, while exposures four and five were also better than average. After that, there were no frequency levels that generated a comparable conversion rate.
However, looking at individual campaigns that made up the 38 tested, there were particular brands that did see cycles in which higher frequency levels did generate higher conversion rates. Seasonality, ad executions and timing seem to be key factors for these products.
3 Specific Recommendations for Online Advertisers
1. Analyze your own data and campaigns
Optimal frequency levels vary greatly, depending on diverse media strategies, different demographic targets, and seasonable variables. Find your own sweet spot. While the first impression on average is best, that may not be the case for your product/brand. The first impression is best generally when your media plan less targeted and relies on reach to build response. Different media strategies call for different levels of effective exposure.
By doing this, marketers will be able to pick frequency levels that maximize total conversion yields, while meeting cost-per-conversion goals.
If you reallocate the waste impressions going to users who have seen the ad more than the optimal number of times, you'll be increasing your reach, making the offer to users who haven't seen the offer already.
As campaigns go on, some users may receive hundreds and even thousands of impressions. For every user who sees an ad 100 times, a cap at 10 impressions would guarantee that at least 9 additional users see those ads.
3. Try to demand some level of frequency capping from your publisher
"Different publishers have varying degrees of capping. Most can frequency cap on sessions or individual ads or for time periods, and they do a really good job on pop-ups, but they're not really incentivized to cap CPM inventory," Song says.
"To be honest," he adds, "I'm not going to be the most popular guy within the publishing community… but it will make it more effective."
(Note: Atlas DMT does not have the ability to provide capping technology. "All of that logic is controlled on the publisher's side," he says.)
Some publishers simply run impressions indiscriminately to make sure that they look like they held up their end of the plan by fulfilling impression guarantees. But in reality, it hurts them in the long run because the clients are not seeing the results they expect from the impressions that are being delivered.
The bottom line, says Song, is that the medium can be made much more effective by distributing impressions more efficiently. "And that's by frequency caps," he says.
Quick notes on Study Methodology
The study looked at 38 advertisers, mostly in B2C sectors, including ecommerce retail, travel and vacation, telecomm, and financial services, whose campaigns ran in the second half of 2003. "Frequency" was defined as campaign, not ad, specific. "If they had 10 different messages and the user saw all ten, it would still be 10 impressions," Song says.
Each advertiser was studied over a four-week period during the advertiser's most active month.
Song calculated conversion rates at each frequency level. The study looked at what happened on everyone's very first impression, what happened on everyone's second impression, etc. If someone converted on the third impression, the study took into account that they were influenced two previous times; it also took into account users who converted but continued to see the ad subsequent times.
According to Song, this cumulative look gives a more realistic idea of what would have happened on a campaign had it been capped at each individual frequency level.
The study didn't take into account recency of impressions, users who clicked but didn't convert, and amount spent at different levels of conversion.