This is an open-access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on http://www.jmir.org/, as well as this copyright and license information must be included.
Measuring the impact of online health campaigns is challenging. Ad click-through rates are traditionally used to measure campaign reach, but few Internet users ever click on ads. Alternatively, self-reported exposure to digital ads would be prone to recall bias. Furthermore, there may be latency effects whereby people do not click on ads when exposed but visit the promoted website or conduct campaign-related searches later. Online panels that unobtrusively collect panelists’ Web behavior data and link ad exposure to website visits and searches can more reliably assess the impact of digital ad exposure. From March to June 2012, the Centers for Disease Control and Prevention aired the national Tips From Former Smokers (Tips 2012) media campaign designed to encourage current smokers to quit. Advertisements ran across media channels, and the digital ads directed users to the Tips 2012 campaign website.
Our aim was to examine whether exposure to Tips 2012 digital ads influenced information-seeking behaviors online.
ComScore mined its panelists’ Web behavior data for unique codes that would indicate exposure to Tips 2012 ads, regardless of whether panelists clicked the ad or not. A total of 15,319 US adults were identified as having been exposed to a Tips 2012 campaign ad. An equal number of unexposed adults (N=15,319) were identified and matched on demographics and Internet use behavior to the exposed group. Panelists’ Web behavior data were mined for up to 4 weeks after initial Tips 2012 ad exposure to determine whether they visited the Tips 2012 campaign website or other cessation-related websites (eg, nicotine replacement therapy site) or conducted searches for campaign-related topics (eg, quit smoking).
The proportion of exposed adults visiting the Tips 2012 sites increased from 0.4% in Week 1 to 0.9% 4 weeks after ad exposure, and these rates were significantly higher than in the unexposed group (0.1% in Week 1 to 0.4% in Week 4,
These results suggest that online ad exposure is associated with confirmed visits to the Tips 2012 campaign sites and visits to other cessation websites and that these information-seeking behaviors occur up to several weeks after ad exposure. Web behavior data from online panels are useful for examining exposure and behavioral responses to digital campaign ads.
Smoking is the leading cause of preventable deaths in the United States, accounting for approximately 480,000 deaths annually [
In 2012, the Centers for Disease Control and Prevention (CDC) launched Tips 2012 From Former Smokers (Tips 2012)—the first federally funded, national tobacco paid-media education campaign. The Tips 2012 campaign advertisements aired nationally from March to June 2012 on cable television, radio, online, print, and outdoor media (eg, billboards). Campaign ads featured former smokers sharing their stories about the daily challenges of living with smoking-related illnesses. To provide smokers with resources and information about quitting, Tips 2012 television ads promoted the 1-800-QUIT-NOW telephone quitline portal and the National Cancer Institute’s (NCI) Smokefree website [
The purpose of this study is to examine the impact of the Tips 2012 digital advertisements on cessation information-seeking behaviors online. Traditionally, the impact of campaign ad exposure on short-term tobacco-related outcomes like information-seeking behaviors and awareness of campaign messages has been examined by linking gross rating points (ie, reach x frequency of ad exposure) as an exogenous observational measure of television ad exposure to survey responses (eg, [
Measuring the impact of digital campaign advertisements is challenging. Ad impressions and click-through rates (CTRs) are traditionally used to measure message reach, but they are limited because only a small fraction of Internet users ever click on ads [
A more accurate approach to measuring digital ad exposure and behavioral impact is needed. A panel-based method that unobtrusively collects Web behavior data and can link ad exposure to online information-seeking behaviors at the individual level may be a more reliable method for measuring online campaign effects [
The digital campaign consisted of display, video, mobile, and search ads that were intended to reach the target audience of adult smokers aged 18-54 and to direct them to the Tips 2012 website. All digital ads ran from March 19-June 10, 2012. Display ads were animated or static and appeared at the top or sidebar of popular websites, such as weather.com, to attract target audiences. The display ads were placed on select websites and ad networks and highlighted the stories of former smokers Annette, Brandon, Shawn, Roosevelt, Suzy, and Terrie (see examples in
Video ads are generally shown before or after other video content, such as an online television show or music video. Video ads ran on 29 websites and featured the same 30-second ads that ran on television. Video ads were tagged with the CDC Tips 2012 website so that viewers who clicked on the ad were directed to the Tips 2012 site. For video ads, approximately 407 million impressions were served, and these ads generated 4.2 million clicks for a CTR of 1.05%, which exceeded industry standards of 1.03% for video ad CTRs [
Search ads appear at the top and sidebar of search results so that when consumers type in any of the paid search terms (eg, how to quit smoking), the top result retrieved is the Tips 2012 site. Tips 2012 search ads were purchased for the two top search engines, Google and Yahoo. For search ads, 22.6 million impressions were delivered via search ads, and these ads generated 224,811 clicks for a CTR of 1.0%, which was equal to the industry standard of 1.0% [
Provides examples of CDC Tips 2012 campaign-related advertisements.
The data for this analysis came from comScore, a market research company that unobtrusively collects Web behavior data on 1+ million US Internet users to measure trends in consumer behaviors online. Panelists download tracking software on their computers that enables comScore to track their Web behavior, including every website they visit, searches they conduct, purchases they make, and ads that are delivered on sites visited, regardless of whether the ads are clicked or not. These data are then aggregated and weighted to provide national estimates on consumer behaviors online. The panel is a convenience sample with panelists largely recruited via nonprobability-based sampling methods (eg, online ads, partner websites). However, a subsample is recruited via random-digit-dialing to calibrate the post-stratification weights that comScore uses to project its estimates to the US Internet population. Panelists are provided incentives for participation such as free online games and charitable donations on panelists’ behalf.
The Tips 2012 campaign’s media contractor provided the “tags” (hash identification code, hypertext markup language source code) to each of the Tips 2012 digital display ads. Using this information, comScore mined its panelists’ Web behavior data for these display ad tags and exposure to sponsored links (search ads) to identify individuals who were exposed and not exposed to the Tips 2012 digital campaign from March 19-June 10, 2012. Video and mobile ad exposure was not examined in this study. Among comScore’s approximately 1 million US adults who were active panelists (ie, tracking software was installed and sending data) during the March 19-June 10, 2012 time period, 15,319 panelists were identified as having been “exposed” to the Tips 2012 digital display and/or search ads. For a control group, an equal number of “unexposed” adults (N=15,319) were matched to the exposed group on demographics and Internet use behavior (eg, time spent online) using propensity score matching nearest neighbor approach (
Demographic characteristics of exposed and unexposed panelists.
Demographic | Exposed panelists, % (N=15,319) | Unexposed panelists, % (N=15,319) | |
|
|||
|
18-24 | 15 | 11 |
|
25-34 | 18 | 20 |
|
35-44 | 20 | 21 |
|
45-54 | 24 | 25 |
|
55-64 | 14 | 15 |
|
65+ | 9 | 8 |
|
|||
|
White | 46 | 50 |
|
Black | 22 | 20 |
|
Asian | 13 | 13 |
|
Other | 19 | 17 |
|
|||
|
North East | 20 | 19 |
|
North Central | 20 | 20 |
|
South | 37 | 40 |
|
West | 23 | 21 |
|
|||
|
Less than $25K | 30 | 29 |
|
$25K-50K | 25 | 24 |
|
$50K to <75K | 21 | 23 |
|
$75K to <100K | 12 | 13 |
|
More than $100K | 12 | 10 |
Children in household | 27 | 28 |
We examined whether panelists in the exposed and unexposed groups visited any of the campaign sites listed in
The Tips 2012 television ads directed audiences to NCI’s Smokefree.gov website because it offers extensive cessation resources. As a result, awareness of the Smokefree.gov website may be high and people may associate this site with the Tips 2012 campaign. Therefore, we also examined visits to NCI’s Smokefree.gov, its associated websites (Smokefree Women and Smokefree Espanol), and social media pages.
Tips 2012 campaign and non-campaign websites.
Website | URL | ||
|
|||
|
|
||
|
|
CDC Tips 2012 campaign website | cdc.gov/tobacco/campaign/Tips 2012/ |
|
|
Tobacco Free Facebook page | facebook.com/cdctobaccofree |
|
|
Tobacco Free on Twitter (@CDCTobaccoFree) | twitter.com/CDCTobaccoFree/ |
|
|
CDC StreamingHealth YouTube Channel | youtube.com/user/CDCStreamingHealth |
|
|
||
|
|
NCI Smokefree website | Smokefree.gov |
|
|
Smokefree.gov on Twitter (@SmokefreeGov) | twitter.com/smokefreegov |
|
|
Smokefree Women website | women.smokefree.gov |
|
|
SmokefreeWomen onTwitter (@SmokefreeWomen) | twitter.com/SmokefreeWomen |
|
|
Smokefree Women YouTube Channel | youtube.com/SmokefreeWomen |
|
|
Smokefree Women Facebook page | facebook.com/smokefree.women |
|
|
Smokefree Espanol website | espanol.smokefree.gov |
|
|||
|
|
||
|
|
HealthWays cessation service | quitnet.com |
|
|
Alere cessation service | quitnow.net |
|
|
Legacy cessation service | becomeanex.org |
|
|
American Cancer Society cessation resources | cancer.org/Healthy/StayAwayfromTobacco/GuidetoQuittingSmoking/index |
|
|
American Lung Association cessation resources | lung.org/stop-smoking/ |
|
|
||
|
|
Nicoderm CQ patch | nicodermcq.com |
|
|
Nicotrol inhaler | nicotrol.com/ |
|
|
Nicorette gum/lozenge/mini | nicorette.com |
|
|
Habitrol patch | habitrol.com |
|
|
||
|
|
Make Smoking History—Massachusetts | Makesmokinghistory.org |
|
|
Tobacco Free Florida | tobaccofreeflorida.com |
|
|
Tobacco Free Florida—Facebook | facebook.com/TobaccoFreeFlorida |
aThese are examples only, not the entire list. In total, 101 cessation sites were examined, including 10 national cessation-related sites, 4 NRT sites, and 87 state cessation program sites. This list of sites were compiled and reviewed by tobacco control researchers at Research Triangle Institute and CDC.
We also examined visits to key national cessation sites (eg, [
To determine whether exposure to Tips 2012 online ads influenced audiences to seek out additional information about the campaign, panelists’ search behavior data were mined for the occurrence of specific (eg, Tips 2012, Terri ad) and general (eg, quit smoking) campaign-related search queries on major search engines (eg, Google, Bing) as well as general websites with search functions (eg, YouTube). A list of 2270 potential search terms were examined based on top external keywords from Adobe SiteCatalyst for the CDC Tips 2012 website and Google Analytics for NCI’s Smokefree.gov site, as well as top keywords used in the digital ad campaign.
For each time period, we calculated the proportion of panelists in the exposed and unexposed groups who (1) visited the Tips 2012 campaign-related websites, (2) visited Smokefree-related websites, (3) visited other non–campaign-related cessation websites, and (4) conducted searches for any campaign-related key terms. Proportions were calculated separately for the exposed and unexposed groups and at each weekly time period (Week 1, Week 2, Week 3, and Week 4) after initial ad exposure. Results for Week 1 represent the proportion of exposed and unexposed groups who visited campaign sites or conducted searches within 1 week after first campaign ad exposure. Results for Week 2 represent the proportion of exposed and unexposed groups who visited campaign sites or conducted searches within Weeks 1 and 2 after first campaign ad exposure, and similarly Week 3 represent Weeks 1-3 after first campaign ad exposure, and Week 4 represent Weeks 1-4 after first campaign ad exposure. We conducted
Figure 2 summarizes the proportion of exposed and unexposed panelists who visited any of the Tips 2012–related campaign sites. The proportion of exposed panelists visiting Tips 2012 sites increased from 0.4% in Week 1 to 0.9% in Week 4 after initial ad exposure. Unexposed panelists also visited Tips 2012 sites but at significantly lower rates from 0.1% in Week 1 to 0.4% at Week 4 (see
Shows the change in visits to CDC Tips 2012-related websites over the course of the campaign by digital ad exposure.
Visits to CDC Tips sites.
|
Week 1 | Weeks 1-2 | Weeks 1-3 | Weeks 1- 4 |
Exposed, % (95% CI) | 0.4 (0.32-0.53) | 0.6 (0.48-0.72) | 0.8 (0.64-0.93) | 0.9 (0.75-1.07) |
Unexposed, % (95% CI) | 0.1 (0.08-0.20) | 0.2 (0.15-0.30) | 0.3 (0.23-0.41) | 0.4 (0.30-0.50) |
|
4.672 | 5.012 | 5.294 | 5.339 |
|
<.001 |
Figure 3 summarizes the proportion of exposed and unexposed panelists who visited any of the non–Tips 2012 cessation sites, including NRT sites, general cessation information sites, and state-specific cessation sites. The proportion of exposed panelists visiting other cessation websites increased from 0.2% in Week 1 to 0.3% in Week 4 after initial ad exposure. Unexposed panelists also visited other cessation websites but at lower rates, ranging from 0.0% in Week 1 to 0.2% at Week 4 (see
Shows the change in visits to cessation-related websites over the course of the campaign by digital ad exposure.
Visits to other cessation sites (national, state, NRT).
|
Week 1 | Weeks 1-2 | Weeks 1-3 | Weeks 1- 4 |
Exposed, % (95% CI) | 0.2 (0.09-0.22) | 0.2 (0.14-0.30) | 0.3 (0.19-0.36) | 0.3 (0.20-0.41) |
Unexposed, % (95% CI) | 0.0 (0.00-0.08) | 0.1 (0.03-0.13) | 0.1 (0.07-0.20) | 0.2 (0.10-0.26) |
|
3.135 | 3.053 | 2.560 | 1.973 |
|
.001 | .001 | .005 | .019 |
Figure 4 summarizes the proportion of exposed and unexposed panelists who conducted searches on any of the campaign-related terms. The proportion of exposed panelists searching for cessation-related information increased from 0.2% in Week 1 to 0.7% in Week 4 after initial ad exposure. Unexposed panelists also conducted searches but at slightly lower rates from 0.2% in Week 1 to 0.5% at Week 4 (see
Shows the change in CDC Tips 2012 campaign-related searches over the course of the campaign by digital ad exposure.
Search for any cessation terms.
|
Week 1 | Weeks 1-2 | Weeks 1-3 | Weeks 1- 4 |
Exposed, % (95% CI) | 0.2 (0.04-0.36) | 0.4 (0.20-0.55) | 0.6 (0.32-0.81) | 0.7 (0.40-0.94) |
Unexposed, % (95% CI) | 0.2 (0.03-0.28) | 0.3 (0.12-0.39) | 0.4 (0.19-0.60) | 0.5 (0.32-0.75) |
|
0.377 | 1.033 | 1.037 | 0.773 |
|
.414 | .066 | .032 | .122 |
In summary, exposure to Tips 2012 digital display and search ads influenced visits to the Tips 2012 campaign-related websites, with visits occurring even up to 4 weeks after initial ad exposure. The total proportion of exposed panelists who visited the Tips 2012 website was higher than the total CTRs for digital display ads over the entire campaign period (0.9% vs 0.1%), suggesting that CTRs alone may underestimate campaign reach. These results are consistent with findings from a previous study that used a similar methodology to examine the influence of digital display ad exposure on visits to Florida state tobacco cessation website and social media pages [
Interestingly, exposure to Tips 2012 digital ads influenced other information-seeking behavior online. The increased visits to other cessation sites (eg, NRT sites, quitnet) among those exposed to the Tips 2012 ads suggest that the campaign had an added benefit of driving traffic to other cessation sites. Therefore, state programs could leverage the timing of a national campaign like Tips 2012 and supplement it with additional local ad buys to drive target audiences to seek cessation resources. Those exposed to the Tips 2012 digital ads were not consistently more likely to search for campaign-related cessation information than those unexposed. This may be because both groups were exposed to messages from other channels (eg, Tips 2012 television/radio ads) that may have influenced similar patterns of information-seeking behavior online. In this study, we were not able to control for exposure to campaign messages from other media channels. However, since consumers are increasingly using multiple media devices simultaneously (eg, nearly 40% of Americans use their tablets or smartphones while watching television [
We saw an increase in the proportion of panelists visiting the Tips 2012 campaign site over time, but we cannot be sure whether this was due to a latency effect (ie, panelists remembering the campaign ad and visiting the site later) or increased exposure to multiple Tips 2012 ads during the post ad exposure 4-week follow-up period. In this study, we examined only first exposure to Tips 2012 digital display ad, but future studies should assess whether there is a dose-response relationship between the amount of digital advertising exposure and information-seeking behaviors online. Future studies should also examine the relative effectiveness of different types of digital ads (ie, display ads vs video ads vs mobile ads) on information-seeking behavior as we were able to examine exposure only to display and sponsored link search ads in this study.
While we found statistically significant differences in website visits and campaign-related searches between the exposed and unexposed groups, overall, the magnitude of the visits and searches and the difference between the groups were small. It is challenging to put these findings in context given the paucity of research on the effects of digital ad campaigns. Further research is needed to build the evidence base for digital media campaign effects.
This study has several strengths. First, Web behavior data were collected unobtrusively, and ad exposure was measured regardless of whether the ad was clicked or not. Prior studies [
This study also has limitations. First, we were unable to determine whether the increased visits to the campaign website were due to latency effects, increased level of digital ad exposure, or exposure to campaign content from other media channels. Future studies should examine the level and timing of ad exposure across media platforms to better understand dose-response relationships and cross-media effects. Second, we examined the influence only of display and search ad exposure, so we cannot determine whether these results would also translate to video or mobile ad exposure. Video ads may be more effective than display ads because advertisers can deliver more engaging and longer content in video formats and place these ads on sites like YouTube, which generate substantial traffic. Third, comScore’s panel is a convenience sample, and although estimates are weighted to the online population, results may not generalize to the US adult population. Fourth, we were unable to examine how smokers specifically responded to campaign ads because information on panelists’ smoking behavior was not available for this study. It is possible that panelists who were exposed to the digital ads were more likely to be smokers interested in quitting and therefore engaged in more information-seeking behavior online than the unexposed panelists. In this study, panelists were matched on demographic characteristics to isolate the influence of digital ad exposure, but future studies should investigate the impact of digital ad campaigns on specific subgroups. The audience that responds to digital ads is likely to be demographically and behaviorally different from the audience that responds to television ads. Therefore, understanding who is being reached can help campaign planners optimize media purchases across channels to reach target audiences most effectively. Finally, due to the confidential nature of proprietary data collected from comScore, we were unable to obtain detailed information about their methods (eg, specifics of data mining procedure) that may be needed to replicate studies of similar scope in the future. This is a common limitation when using data from digital analytics companies like comScore. For this reason, the national Media Rating Council conducts detailed audits of media industry companies to ensure that audience measurement services are valid, reliable, and ethical [
The results of this study show that exposure to digital display and search ads is associated with confirmed visits to the campaign website up to several weeks after initial ad exposure regardless of whether the ad was clicked or not. Results also suggest that these ads may cue audiences to seek other cessation-related websites. Web behavior data from online panels are useful for examining exposure and behavioral responses to digital campaign ads because they provide a more comprehensive assessment of campaign impact than relying on ad impressions and CTRs alone. Future studies should examine the optimal dose needed to achieve information-seeking behaviors, the relative impact of different types of digital ads, cross-platform influences and synergies, and impact on specific subgroups like smokers. Digital advertising is a potentially powerful tool for motivating audience’s information seeking around behaviors that are targeted in campaign messages. Researchers and practitioners have an opportunity to harness the vast volume of digital data to provide a more evidence-based approach to designing and evaluating digital media campaigns and to help inform best practices.
Centers for Disease Control and Prevention
click-through rate
National Cancer Institute
nicotine replacement therapy
Uniform Resource Locator
This work was funded by the Centers for Disease Control and Prevention. The findings and conclusions in this manuscript are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention or RTI International.
None declared.