ABSTRACT
There are competing explanations for why longer advertisements are remembered better, such as more time to memorize, add branding and claims, tell stories, and get attention, with some acknowledgment of diminishing returns. The authors compared seven-, 15-, 30-, and 60-second versions of the same commercials for their brand recall, advertisement liking, and brand attitude, with additional biometric measures. Seven-second advertisements were almost as effective (measured by unaided recall) as 15-second advertisements and 60 percent as effective as 30-second advertisements, a finding that confirms and extends a diminishing-returns explanation for recall and other measures of effectiveness, such as advertisement liking. With six-second commercials appearing on television and online, short advertisements can provide an efficient option if used with the knowledge of what they can and cannot deliver.
MANAGEMENT SLANT
Short commercials can deliver effectiveness efficiently, because commercial length has diminishing returns.
Most of the effectiveness of long commercials is delivered in the first five seconds.
Short commercials can achieve rates of brand recall nearly comparable to those of longer advertisements, but longer commercials allow more time to tell a story and take viewers on an emotional journey, which increases brand recall and advertisement liking.
Short commercials potentially increase clutter, which reduces recall and advertisement liking.
Short advertisements seem a sensible tactic to increase continuity of a campaign in conjunction with longer copy, at least for established brands.
INTRODUCTION
Video advertisements are getting shorter, on television and online. It has been a long journey. In the 1950s and 1960s, television advertisements were typically 60 seconds, shrinking to 30 seconds in the 1970s, with 15-second commercials appearing in the late 1980s (Elliott, 2005). Now, video advertisements can be six seconds long (Crupi, 2017) or skipped after five seconds.
Previous studies have compared 15-second advertisements with 30-second advertisements (Mord and Gilson, 1985; Newstead and Romaniuk, 2010; Patzer, 1991; Singh and Cole, 1993; Stanton and Burke, 1998) and 30-second advertisements with longer advertisements—60 seconds, 90 seconds, and up to 240 seconds (Newell and Henderson, 1998). The current study investigated the effectiveness of shorter seven-second advertisements, half the length of 15-second advertisements. By using multiple advertisement lengths (seven, 15, 30, and 60 seconds), the current study tested whether the relationship between advertisement length and effectiveness is linear or nonlinear (with diminishing returns).
The other purpose was to further document moderators of short-advertisement performance. Advertisement effectiveness varies enormously with creative execution, for example (Hartnett, Kennedy, Sharp, and Greenacre, 2016). The current research investigated the effects of early branding, a creative-execution tactic that eliminated the disadvantage of shorter length in a prior field study (Newstead and Romaniuk, 2010). Given that younger viewers, such as millennials, may be more responsive to shorter advertisements (Lang, Schwartz, and Mayell, 2015), the effect of advertisement length for younger versus older viewers also was explored. Finally, the research investigated the impact of clutter, given that shorter advertisements allow media networks to show more advertisements, which increases advertising clutter.
Advertising effectiveness can be investigated in a range of ways. Ultimately, advertisers are interested in behavior change (e.g., sales). Data that link advertising to sales are rare, however, with known confounds (Taylor, Kennedy, and Sharp, 2009). For this reason, many advertisers employ intermediate measures of effectiveness (Rossiter and Percy, 2017).
This research used three intermediate measures of effectiveness: unaided brand recall, advertisement liking, and brand attitude. Additional measures of brand awareness (aided brand recall and brand recognition) tested clutter, which prior research has suggested affects only unaided recall (Brown and Rothschild, 1993). Second-by-second responses to advertisements, which are difficult for viewers to self-report, were investigated with promising biometric measures (Bellman, Nenycz-Thiel, Kennedy, Larguinat, et al., 2017; Kennedy and Northover, 2016; Varan, Lang, Barwise, Weber, et al., 2015). Given the lack of literature on biometrics and advertisement length, the biometrics were used only as potential explanations for effects.
LITERATURE REVIEW AND HYPOTHESIS DEVELOPMENT
Advertisement Duration and Memory
At least three explanations have been proposed for the finding that longer advertisements more likely will be remembered than shorter advertisements (Martín-Santana, Reinares-Lara, and Reinares Lara, 2016; Mord and Gilson, 1985; Patzer, 1991; Pieters and Bijmolt, 1997; Rethans, Swasy, and Marks, 1986; Singh and Cole, 1993; Singh and Rothschild, 1983). The first, the “total time hypothesis,” proposes “a simple linear relationship between time available to commit something to memory and the amount retained” (Baddeley, 1976, quoted in Singh and Cole, 1993, p. 92). Longer exposure length provides more time for encoding, storage, retrieval, and rehearsal (Anderson, Bothell, Lebiere, and Matessa, 1998; Lang, 2000). If recall is a straight-line function of exposure time, then there should be no upper limit to the effect of exposure time on advertising effectiveness. A linear relationship also implies that there is nothing special about the first few seconds of exposure to an advertisement.
A second explanation is that more time in the advertisement allows more opportunities for branding and for making brand claims (Singh and Cole, 1993). When short and longer advertisements have the same amount of branding, they have equally correct brand identification (Newstead and Romaniuk, 2010; Singh and Cole, 1993). The third explanation is that longer advertisements provide more time for viewers' attention to return to the screen and so more likely will gain attention and therefore be recalled later (Martín-Santana et al., 2016; Rossiter and Percy, 2017). These last two explanations also suggest that measures of recall should be a linear function of advertisement length.
An alternative theory proposes that exposure time has a nonlinear diminishing-returns effect on memory. According to this theory, people develop templates for storing incoming information. The complexity and availability of a template depend on the person's expertise. This theory was developed to explain why the effect of exposure time on memory (for positions on a chess board) was different for expert and novice chess players (Gobet and Simon, 2000).
The probability of remembering increases rapidly at first as incoming information, such as a brand mention, is stored in available slots in the template. With longer exposure time, there are fewer available slots to fill, so time has diminishing returns and an implied upper limit for recall. This diminishing-returns model has been applied successfully to static web-display advertisements (Goldstein, McAfee, and Suri, 2011). In the current study, the authors tested whether a diminishing-returns model of exposure time also applies to dynamic television advertisements.
Pricing of advertisements suggests that the industry has an implicit theory that the relationship between advertisement length and advertisement effectiveness is nonlinear. There is a rule of thumb that 15-second advertisements are 80 percent as effective as 30-second advertisements (Newstead and Romaniuk, 2010; Patzer, 1991), with their cost about 80 percent (between 60 percent and 90 percent) of a 30-second advertisement (Singh and Cole, 1993). A nonlinear relationship between length and effectiveness matches the pattern that advertising has diminishing returns from repetition and spending (Simon and Arndt, 1980; Taylor et al., 2009). Logic suggests that the pattern should hold for advertisement length. Just as the first exposure is the most effective when there are repeated exposures, for any single exposure the first seconds should be the most effective (assuming the copy is good and linked to the brand).
The current study tests the likely diminishing-returns function as it applies to memory for television advertisements (See Figure 1). If 15 seconds of exposure is about 80 percent as effective as 30 seconds, measured by advertisement recall (Newstead and Romaniuk, 2010; Patzer, 1991; Singh and Cole, 1993), then half the effectiveness of that 30 seconds of exposure is delivered by the advertisement's first five seconds. This explains why 15-second advertisements are priced at more than 50 percent and closer to 80 percent of the price of normal 30-second advertisements (Patzer, 1991).
A model of the nonlinear effect of advertisement length could be used to estimate the performance of shorter advertisements. If five-second advertisements deliver half the effectiveness of 30-second advertisements, for example, they could be worth half the price of a 30-second advertisement. There may be a minimum effective exposure time for video advertisements (e.g., one or two seconds). Before looking at such short lengths, however, the authors investigated video-advertisement lengths that currently can be bought.
In this study, the measure of advertising memory used was unaided brand recall:
H1: Longer advertisement length (as compared to shorter) increases unaided brand recall with diminishing returns.
Diminishing Returns for Advertisement Liking And Brand Attitude
Advertisement liking and brand attitude are affected by emotional responses as well as by brand awareness and simple learning. Two additional explanations exist for the effects of advertisement length on advertisement liking and brand attitude. The first of these is that viewers find longer television advertisements more persuasive, because longer advertisements provide more time for information and arguments (Pieters and Bijmolt, 1997). According to the second explanation, longer commercials have the opportunity to generate stronger emotional responses (Singh and Cole, 1993), and therefore increase advertisement liking (Newstead and Romaniuk, 2010), because they have a more complex story arc, with ups and downs in emotional response (Mord and Gilson, 1985; Newstead and Romaniuk, 2010; Stanton and Burke, 1998).
Like previous studies (e.g., Singh and Cole, 1993), this study tested shorter advertisements that were cut-down versions of longer advertisements. Advertisement liking and brand attitude may be underestimated or overestimated if cut-down advertisements remind viewers of the advertisements' longer versions. For these reasons, the authors measured ups and downs in emotional responses using biometrics, which potentially are unaffected by memories of the longer advertisement.
The effects of advertisement length on emotional responses may be linear, because longer advertisements offer continually increasing scope for response (Mord and Gilson, 1985). Alternatively, the difference between a linear and a nonlinear effect may be too hard to detect in the very small differences between short and long advertisements on measures of advertisement liking and brand attitude (Patzer, 1991). Whether enjoyment from longer and stronger emotional experiences adds much to effectiveness may depend on whether the advertisement uses an emotional appeal (Pham, Geuens, and De Pelsmacker, 2013; Singh and Cole, 1993) as well as on the category being promoted. Road-safety messages may nudge good behavior without being enjoyed, for example. Prior research nevertheless suggests a nonlinear relationship due to satiation (Grass and Wallace, 1969) and wear-out (Berlyne, 1970; Cacioppo and Petty, 1979; Rethans et al., 1986). Thus:
H2: Longer advertisement length (compared to shorter) increases advertisement liking, and the favorability of brand attitude, with diminishing returns.
Advertisement Duration and Age
In the 1990s, the industry was concerned that younger viewers did not like shorter advertisements (Mord and Gilson, 1985; Patzer, 1991). Now, younger viewers are watching less traditional television (Richter, 2017) and more online video (McCarthy, 2017), and they may prefer shorter advertisements because of their familiarity and also their faster pace (Lang et al., 2015). A recent facial-tracking study found that viewers younger than 29 years old were happier and more engaged than other viewers when watching advertisements less than eight seconds long (Freewheel Council for Premium Video, 2018). The current authors sampled younger, middle-aged, and older viewers, to test the following hypothesis:
H3: Shorter advertisements (compared to longer) will be more effective, as measured by unaided brand recall, advertisement liking, and brand attitude, for younger versus older viewers.
Early Branding
Shorter advertisements that use early branding can be as effective as longer advertisements for correct brand identification (Newstead and Romaniuk, 2010). Branding early and often, however, potentially hampers storytelling (Stanton and Burke, 1998). The current study tested whether the benefits, for shorter advertisements, of early branding on memory extend to advertisement liking and brand attitude:
H4: Early branding can enhance advertising effectiveness for shorter advertisements, as measured by unaided brand recall, advertisement liking, and brand attitude, to similar levels as late-branded longer advertisements.
Product Involvement
If advertising for a product that is more involving is processed more deeply (Petty, Cacioppo, and Schumann, 1983; Petty, Schumann, Richman, and Strathman, 1993), then it may not matter whether the advertisement is short or long. The current research used lab experiments in which product involvement was manipulated, to control for prior exposure and product-usage effects (Vaughan, Beal, and Romaniuk, 2016), which potentially explain why a previous field study found higher recall for low-involvement-product advertisements (Newstead and Romaniuk, 2010).
H5: High product-category involvement can enhance the effectiveness of shorter advertisements, as measured by unaided brand recall, advertisement liking, and brand attitude, to similar levels of effectiveness achieved by longer advertisements for low-involvement products.
Repetition
Many advertisers launch campaigns with a long advertisement, then use short advertisements as reminders (Newstead and Romaniuk, 2010). Short exposures (e.g., to fast-forwarded advertisements) can be effective as reminder messages (Bellman, Schweda, and Varan, 2010; Gilmore and Secunda, 1993; Neale, Bellman, Treleaven-Hassard, Robinson, et al., 2013). Using shorter advertisements for repeat exposures potentially minimizes negative wear-out effects (Singh and Cole, 1993). This suggests that repeat exposures are more effective than a single exposure, even when the repeat exposure is supplied by a short advertisement:
H6: Repeat exposure will enhance advertising effectiveness, as measured by unaided brand recall, advertisement liking, and brand attitude, even when the later exposure is to a shorter version of the advertisement.
Context Effects
Another expectation tested was that short advertisements work best in a related context (Sharma, 2010): on mobile devices, or as prerolls before short videos. This context-matching effect has been shown to apply to programs and advertisements with matching negative emotion (Kamins, Marks, and Skinner, 1991) or similarly informational content (Bellman, Wooley, and Varan, 2016). A match-up effect on short advertisements also could derive from positive-affect transfer if the viewer is enjoying the related mobile or short-program context (Pham et al., 2013; Wang and Lang, 2012).
H7: Shorter advertisements (compared to longer) will be more effective, as measured by unaided brand recall, advertisement liking, and brand attitude, (a) on a mobile device or (b) in the context of shorter program content.
Clutter
Another reason for pricing 15-second advertisements at a higher cost per second than 30-second advertisements is to deter advertisers from buying them (Patzer, 1991). Shorter advertisements can increase advertisement clutter, measured by the number of advertisements in advertising breaks. The effects of clutter are summarized by an empirical generalization: As clutter goes up, the percentage of advertisements remembered goes down (Hammer, Riebe, and Kennedy, 2009).
In a field study, unaided brand recall declined from 50 percent when there were four advertisements in a break to 20 percent when there were 17 advertisements per break (Pieters and Bijmolt, 1997). Additionally, a higher number of shorter advertisements may influence perceptions that breaks are longer than usual (Mord and Gilson, 1985; Patzer, 1991) and more annoying (Potter, 2009), which reduces advertisement liking (Mord and Gilson, 1985) and, potentially, program liking.
H8: Clutter will reduce the effectiveness, as measured by brand recall, advertisement liking, and brand attitude, of shorter advertisements.
METHOD
The hypotheses were tested in three related lab experiments testing the same advertisements (See Table 1). The shorter advertisements were cut-down versions of the original long version (See Appendix 1). Testing shorter versions of the same advertisement controls for many potential differences between advertisements for different brands and products (Mord and Gilson, 1985; Newstead and Romaniuk, 2010; Patzer, 1991; Singh and Cole, 1993). To average over differences in creative execution, including early versus late branding (Newstead and Romaniuk, 2010), the researchers used 11 test advertisements across the three studies. Late branding first appeared in the last third of the advertisement. High product involvement was defined by high purchase risk (Newstead and Romaniuk, 2010; Rossiter and Percy, 2017), even for regular buyers of products, such as automobiles or insurance; others were low risk, such as beverages.
The data were collected in Austin, TX, and Chicago, from 722 consumers. Participants sat in individual NeuroQube® stations and watched the content on a large computer screen or (for Study 2 only) a smartphone fixed to a stand.
Dependent Variables
These studies' effectiveness measures (unaided brand recall, advertisement liking, and brand attitude) were validated in previous studies (e.g., Bergkvist and Rossiter, 2007). Their external validity will be lower, however, because they were collected after forced exposure. Unaided brand recall was measured in the post-test questionnaire, after questions about program liking but before questions about the advertisements (See Appendix 2).
An open-ended question asked participants to “Please list the names of the brands and products (e.g., Secret deodorant) that you remember seeing or hearing advertisements for in today's session.” Correct brand recall (misspellings allowed) was coded as 1, otherwise 0. Aided recall (Studies 2 and 3) was assessed next; the product categories were used as cues (correct brand = 1, otherwise 0). To measure brand recognition (Studies 2 and 3), the researchers showed participants a list of the test brands and their nearest competitors (i.e., the guess rate was 50 percent) and asked them to check all the brands that were advertised during the program (correct recognition = 1, otherwise 0). Advertisement liking and brand attitude then were collected, and a manipulation check of product category involvement was performed.
The authors acknowledge that interval-scale measures of advertising effectiveness, such as advertisement liking and brand attitude, do not have a meaningful zero; people still have an attitude at the neutral midpoint. For these measures, 15/30 ratios do not apply.
Biometric Measures
To explore the impact of advertisement length on emotional response, the researchers used biometric measures of emotional intensity (arousal) and positive valence (smiling), measured second by second. Electrodes attached to the nondominant hand measured arousal by skin-conductance level, which increases with sweating (Potter and Bolls, 2012). A high-definition camera recorded smiling activity on the participant's face, which was coded by computer as a probability (percentage) of smiling.
Because of the number of peaks and troughs in the skin-conductance and smiling traces, the authors compared the arc length of the traces: Longer traces have more ups and downs. In a validation of this measure, skin-conductance traces for higher rating sitcoms had longer arc length, which indicates that these programs took their audiences on a more varied and enjoyable emotional journey (Juckel, Bellman, and Varan, 2016). Given the lack of literature on biometrics and advertisement length specifically, the biometrics have been used only as potential explanations for the effects.
STUDY 1
Method
Study 1 tested the first five hypotheses, which investigated the effects of advertisement length, branding, and product involvement. In Study 1, normal 30-second advertisements were compared with short 15-second advertisements and long 60-second advertisements (See Table 1). Each participant saw all three advertisement lengths, and each advertisement length was represented by three different brands (i.e., participants saw nine test advertisements in total). Participants saw each test advertisement just once, in a normal level of clutter, with five advertisements per break. The participants saw these advertising breaks in the context of a one-hour television program.
The test advertisements were rotated evenly by nine presentation-order variations, and participants randomly were assigned to watch one of four program genres. Brand recall, advertisement liking, and brand attitude were averaged across the three brands for each advertisement length (e.g., the three 15-second advertisements). The researchers compared these averages using repeated-measures analysis of variance. To investigate age effects (H3), the researchers classified participants into two groups: younger (29 years or younger) and older (30 years or older). Program liking was used as a covariate to control for genre effects.
Results
Hypothesis 1. H1 proposed that longer advertisement length (as compared to shorter) increases unaided brand recall with diminishing returns. In support of H1, advertisement length increased unaided brand recall: F(2, 171) = 19.45, p < .001, ηp2 = .18; with controls for covariates, F(2, 131) = 24.50, p < .001, ηp2 = .27 (See Table 2). A normal 30-second advertisement (M = 15 percent) was more effective than a 15-second advertisement (M = 9 percent, p = .003) but less effective than a 60-second advertisement (M = 23 percent, p < .001; 15 versus 60, p < .001).
Advertisement length had diminishing returns (See Figure 1). Fifteen-second advertisements, half the length of a 30-second advertisement, delivered not half but 80 percent of a 30-second advertisement's unaided recall. Longer 60-second advertisements delivered not twice but only 153 percent of a 30-second advertisement's effectiveness.
Hypothesis 2. H2 proposed that longer advertisement length (compared to shorter) increases advertisement liking, and the favorability of brand attitude, with diminishing returns. This hypothesis was supported partially. There was a significant effect of advertisement length on advertisement liking: F(2, 171) = 19.85, p < .001, ηp2 = .19; with controls for covariates, F(2, 131) = 17.76, p < .001, ηp2 = .21. The effect of advertisement length on brand attitude was not significant, however: F(2, 171) < 1; with controls for covariates, F(2, 131) = 1.44, p = .24, ηp2 = .02.
Doubling advertisement length from 15 seconds (M = 4.6) to 30 seconds (M = 4.8) significantly increased advertisement liking by .2 of a scale point (p = .001). Doubling advertisement length again from 30 seconds to 60 seconds (M = 5.0) also significantly increased advertisement liking (p = .013), but because the amount of increase was the same (.2), the rate of increase was lower and therefore consistent with diminishing returns (.2 over 15 seconds = .01 points per second; .2 over 30 seconds = .007 points per second). The only (marginally) significant difference in brand attitude was between 15-second advertisements (M = 4.6) and 60-second advertisements (M = 4.8, p = .093; 15 versus 30 seconds [M = 4.7], p = .34; 30 versus 60 seconds, p = .53).
The biometric evidence supports a story-telling or emotional-response explanation for why advertisement length improves advertisement liking. There were no effects of advertisement length on average smiling activity, F(2, 80) < 1, p = .39, ηp2 = .02, or average skin conductance, F(2, 108) = 1.57, p = .21, ηp2 = .03. Smiling and skin conductance showed greater variability (up-and-down movement) during longer advertisements, however (See Figure 2).
With advertisement duration normalized to 15 seconds, a perfectly flat line (with no ups or downs) would have an arc length of 15 units. Up-and-down movement in emotional responses increases arc length. Advertisement length had significant effects on arc length for smiling, F(2, 80) = 404.73, p < .001, ηp2 = .91, and skin conductance, F(2, 108) = 679.24, p < .001, ηp2 = .93. The increase in arc length from 15-second advertisements (M = 15.1) to 30-second advertisements (M = 15.5, p < .001) was larger than the increase from 30-second advertisements to 60-second advertisements (M = 15.8, p < .001), which is consistent with diminishing returns. The same diminishing-returns effect was seen for skin-conductance arc length (15 seconds [M = 15.1] versus 30 seconds [M = 15.6] versus 60 seconds [M = 15.8], all differences p < .001).
Hypothesis 3. H3 proposed that shorter advertisements (compared to longer) are more effective in terms of unaided brand recall, advertisement liking, and brand attitude for younger versus older viewers. This hypothesis was supported partially by a marginally significant Age × Advertisement Length interaction on unaided brand recall, F(2, 109) = 2.96, p = .06, ηp2 = .05. Age, however, had no moderating effect on advertisement liking, F(2, 109) = 1.14, p = .32, ηp2 = .02, or brand attitude, F(2, 109) < 1, p = .66, ηp2 = .01.
For younger participants, 15-second advertisements had unaided brand recall (M = 6 percent) almost equal to 30-second advertisements (M = 13 percent, p = .10) but still less than 60-second advertisements (M = 28 percent, p < .001; 30 versus 60, p = .019). For older participants, 15-second advertisements (M = 10 percent) were less effective than both 30-second advertisements (M = 16 percent, p = .004) and 60-second advertisements (M = 22 percent, p < .001; 30 versus 60, p = .052). In Study 1, age differences were not significant for specific advertisement lengths.
Hypothesis 4. H4 proposed that early branding can enhance brand recall, advertisement liking, and brand attitude for shorter advertisements to the same level as late-branded longer advertisements. This hypothesis was not supported by Study 1. Early-branded 15-second advertisements (M = 6 percent) had significantly lower unaided brand recall than late-branded 30-second advertisements (M = 22 percent), F(1, 53) = 16.29, p < .001, ηp2 = .24. Early branding made no difference to advertisement liking or brand attitude; both were F(1, 53) < 1.
Hypothesis 5. H5 proposed that high product-category involvement can enhance the effectiveness (unaided brand recall, advertisement liking, and brand attitude) of shorter advertisements to levels of effectiveness similar to those achieved by longer advertisements for low-involvement products. This hypothesis was not supported by Study 1. First, a manipulation check confirmed that participants had greater involvement with products coded a priori as high involvement (M = 4.8) than with products coded as low involvement (M = 4.3), t(172) = −4.24, p < .001, d = .3.
Second, the Involvement × Advertisement Length interaction was not significant for any of the dependent variables:
unaided brand recall, F(4, 129) = 1.06, p = .38, ηp2 = .03;
advertisement liking, F(4, 129) < 1; and
brand attitude, F(4, 129) = 1.67, p = .16, ηp2 = .05.
Nevertheless, for high-involvement products, such as Allstate or Volvo, there was only one significant difference in unaided brand recall (15 [M = 13 percent] versus 60 [M = 22 percent], p = .03), F(2, 131) = 2.75, p = .07, ηp2 = .04. There were multiple significant differences among 15-, 30-, and 60-second advertisements for low-involvement products, such as milk and bottled water—for example, Evian, F(2, 131) = 12.68, p < .001, ηp2 = .16. Similarly, there was only one significant difference in advertisement liking (15 [M = 4.6] versus 60 [M = 4.9], p = .002) for high-involvement products, F(2, 131) = 4.96, p = .008, ηp2 = .07, whereas there were multiple significant differences for low-involvement products.
The authors explored the possibility that other characteristics of the advertisements improved the performance of their shorter versions, beyond product involvement and early branding, the two factors manipulated by the study's design. The results for each brand were compared separately. Aspects of the Allstate, GoDaddy, Volvo, and Walmart brands and their advertisements, such as brand equity or the specific celebrity used, might have helped their shorter advertisements achieve unaided brand recall equal to their longer advertisements. Volvo's shorter advertisement also performed equally well on advertisement liking and brand attitude. Apart from Volvo, only Evian, a brand with significant differences in recall, had no significant differences on advertisement liking or brand attitude.
STUDY 2
Method
Study 2 built on the results of Study 1 by comparing short 15-second advertisements with very short seven-second advertisements. Study 2 also tested H6 and H7, which investigated repeat exposure and viewing context. Participants saw test advertisements for seven brands, and for five of these the advertisements were repeat exposures (See Table 1). In the :30 + :07 repeat-exposure condition, for example, the first exposure was a 30-second advertisement, and the repeat exposure was a seven-second advertisement. The test brands were rotated across the exposure conditions by seven presentation-order variations.
To investigate the effects of viewing context, in Study 2 the authors randomly assigned participants to one of three groups. One group viewed a one-hour drama on a large television screen (as in Study 1). The other groups viewed a small smartphone screen, watching
the same one-hour program or
short videos with prerolls (as on YouTube).
Results
Hypothesis 6. H6 proposed that repeat exposure enhances advertising effectiveness (measured by unaided brand recall, advertisement liking, and brand attitude), even when the later exposure is to a shorter version of the advertisement. This hypothesis was supported by Study 2. There were significant effects of exposure condition on unaided brand recall, F(6, 208) = 32.60, p < .001, ηp2 = .48, and advertisement liking, F(6, 208) = 9.45, p < .001, ηp2 = .21, but not on brand attitude, F(6, 208) < 1.
There was no significant difference in unaided recall between single exposures (15 [M = 17 percent] versus 7 [M = 12 percent], p = .32; See Table 2). Fifteen-second advertisements (M = 4.6) were liked more than seven-second advertisements, however (M = 4.4, p = .016). Very short seven-second advertisements were as effective as short 15-second advertisements as reminders of longer 30-second advertisements (30 + 7 [M = 56 percent] versus 30 + 15 [M = 57 percent], p = .98).
Repeated seven-second advertisements (M = 40 percent) more likely were recalled than single exposures to seven-second advertisements (M = 12 percent, p < .001). Like longer exposure, repeat exposure increased advertisement liking. Every type of repeat exposure (e.g., 30 + 7 [M = 4.9], 15 + 7 [M = 4.7]) was significantly more effective (p < .05) than a single seven-second exposure (M = 4.4), with one exception. Repeated seven-second advertisements (M = 4.4) were not liked more than single exposures to seven-second advertisements (p = .50).
Hypothesis 7. H7 proposed that shorter advertisements (compared to longer) are more effective (unaided brand recall, advertisement liking, and brand attitude) on a mobile device or in the context of shorter program content. This hypothesis was not supported by Study 2. The Viewing Condition × Exposure Condition interaction was not significant for
unaided brand recall, F(12, 416) < 1;
advertisement liking, F(12, 416) = 1.45, ηp2 = .04; or
brand attitude, F(12, 416) = 1.62, p = .08, ηp2 = .04.
There was a significant main effect of viewing condition on unaided brand recall, F(2, 213) = 10.63, p < .001, ηp2 = .09, and a marginally significant main effect on advertisement liking, F(2, 213) = 1.39, p = .053, ηp2 = .03 (brand attitude, F < 1).
In contrast with H7a, all advertisements, including shorter advertisements, were liked more when viewed in long content on a television (M = 4.8) than when viewed in long content on a mobile device (M = 4.5, p = .015). In contrast with H7b, all advertisements, including shorter advertisements, had lower unaided brand recall when seen as prerolls to short content on a smartphone (M = 29 percent) compared with advertisements in breaks in long content, either on a television (M = 46 percent, p < .001) or on a smartphone (M = 42 percent, p = .001).
STUDY 3
Method
The main purpose of Study 3 was to test H8, which investigated the effects of increasing clutter when 15- and seven-second advertisements allow more advertisements to be shown without increasing the length of advertising breaks. Unlike Studies 1 and 2, in which participants saw a mix of shorter and longer advertisements, most participants in Study 3 saw advertisements that were all 30, 15, or seven seconds long, in advertising breaks that were each 150 seconds. The results of Study 2 suggested that 15- and seven-second advertisements were equally effective, so (for economy) Study 3 included just one short-advertisements normal-clutter control condition, with a mix of 15- and seven-second advertisements (five advertisements per break).
The researchers tested the effects of clutter, rather than length, by comparing the high-clutter (15-second) and very-high-clutter (seven-second) conditions with the results for the short-advertisements control condition. In previous research (Brown and Rothschild, 1993), clutter has affected only retrieval (measured by unaided recall) and has made no difference to memory encoding (measured by recognition) and storage (measured by aided recall; Lang, 2000), so in Study 3 the researchers collected these additional awareness measures. Presentation order, including the allocation of brands to 15- and seven-second lengths in the short-advertisements control condition, was controlled for by eight presentation-order variations.
Results
Hypothesis 8. H8 proposed that clutter reduces the effectiveness (unaided brand recall, advertisement liking, and brand attitude) of shorter advertisements. This hypothesis mostly was confirmed by Study 3. There were significant effects of clutter on unaided brand recall, F(3, 273) = 7.49, p < .001, ηp2 = .08, and, contrary to the authors' expectations, on aided brand recall, F(3, 273) = 6.49, p < .001, ηp2 = .07.
First, a test showed that the unaided recall difference between seven-second advertisements (M = 12 percent) and 15-second advertisements (M = 17 percent) in the short-advertisements control condition was not statistically significant at traditional cutoff values (Z = −1.83, p = .067, paired sign test). The clutter effect might have been stronger, however, if the authors had used a seven-second-advertisements control condition. The results for Study 3 (See Table 2) come from the high-clutter (15-second) and very-high-clutter (seven-second) conditions.
Unaided brand recall was significantly higher for the short-advertisements control condition (M = 14 percent) compared with the high-clutter condition (M = 9 percent, p = .013) and the very-high-clutter condition (M = 6 percent, p < .001). Very high clutter also reduced aided brand recall, compared with the short-advertisements control condition (M = 25 percent versus M = 34 percent, p = .006). Psychophysiological arousal, measured by skin conductance, was significantly lower for seven-second advertisements during very high-clutter breaks (M = 2.7 microsiemens [μS]), F(1, 14) = 11.09, p = .005, ηp2 = .44, and marginally lower for 15-second advertisements during high-clutter breaks (M = 2.9 μS), F(1, 14) = 3.60, p = .079, ηp2 = .20, compared with the short-advertisements control condition (M = 3.1 μS).
Regression Results. The results suggest that a regression model incorporating diminishing returns (See Figure 1) will be a better fit to the data from these three studies (the “Average” rows in Table 2) than a simple linear model. A linear model would have the following form: (1) where Y is the dependent variable (e.g., unaided brand recall) and X is the independent variable, the length of the advertisement in seconds. If one assumes no unaided brand recall after zero seconds (i.e., a = 0), a linear model with b = .004 fits these data well (R2 = 92 percent).
A diminishing-returns model fits the data even better (R2 = 98 percent). It has the following form: (2) where ln(X + 1) is the natural log of advertisement length in seconds, plus 1 in case length = zero. Again, a = 0, but b = .05. Advertisement liking and brand attitude may be nonzero after zero seconds of exposure, so if one assumes a = 3.5 (on a six-point scale) for advertisement liking, a diminishing-returns model (with b = .39) fits the data better (R2 = 99 percent) than a linear model (b = .03; R2 = 81 percent).
Similarly, if one assumes a = 4 for brand attitude, a diminishing-returns model (with b = .25) fits the data better (R2 = 96 percent) than a linear model (b = .02; R2 = 69 percent). Finally, a diminishing-returns model also fits the data for smiling arc length (a = 0, b = 4.4; R2 = 98 percent) better than a linear model (a = 0, b = .35; R2 = 79 percent), as does a diminishing-returns model of skin-conductance arc length. These regression results provide further evidence of the diminishing-returns relationship between advertisement length and advertising effectiveness.
GENERAL DISCUSSION
This research investigated whether advertisement length has a nonlinear diminishing-returns effect, as opposed to a linear effect, on intermediate measures of advertising effectiveness—unaided brand recall, advertisement liking, and brand attitude—with some investigation of effectiveness using biometrics. The results of the three studies show that short advertisements can deliver effectiveness efficiently, because advertisement length has diminishing returns (Simon and Arndt, 1980; Taylor et al., 2009). Consistent with a template theory of memory (Gobet and Simon, 2000), most of the effect of advertisement length on advertisement awareness is delivered by the first five seconds of exposure.
In the current research, seven-second advertisements were just as effective as 15-second advertisements (i.e., twice their length), with unaided brand recall as the measure of effectiveness. Seven-second advertisements delivered 60 percent of a 30-second advertisement's brand recall, even though they were less than a quarter of the length. For different reasons, such as satiation (Grass and Wallace, 1969), advertisement length also had a diminishing-returns effect on advertisement liking, although the size of the effect was smaller and harder to detect. No effect of diminishing returns was found on brand attitude, except when the results were analyzed with regression. The biometric measures used in the present research confirm that longer advertisements were better at taking viewers on an emotional journey with more ups and downs (Mord and Gilson, 1985), measured by smiling activity and skin conductance.
These studies also tested the effects of potential moderators of the diminishing-returns effect of advertisement length. Previous field research (Newstead and Romaniuk, 2010) showed that early and continuous branding can lift short advertisements' brand recall to the level achieved by late-branded advertisements twice their length. In these lab studies, however, like previous lab studies that have controlled exposure and manipulated branding (Singh and Cole, 1993), advertisement length was still more important than branding. Early branding had no effect on biometric measures of emotion or advertisement liking, which suggests that early branding, if used, does not hamper storytelling and advertisement enjoyment. For very short advertisements, such as seven-second advertisements, early branding has a smaller effect because late branding in these advertisements delays branding by less than five seconds. Another potential moderator is age, because shorter advertisements may be more effective for younger viewers (Lang et al., 2015). In the current research, there was some marginally significant evidence that shorter advertisement length makes no difference for younger viewers (younger than 30 years) for unaided brand recall (but not advertisement liking or brand attitude). This may be because younger viewers are more used to seeing shorter advertisements online (McCarthy, 2017).
There was some evidence for another moderator, high versus low involvement. Again, these results from manipulating involvement contrasted with a previous field study (Newstead and Romaniuk, 2010). Advertisement length had less of an effect on brand recall for advertisements for high-involvement products, such as automobiles and insurance, which may be explained by viewers paying more attention to these advertisements (Petty et al., 1983; Rossiter and Percy, 2017). The smaller effect of advertisement length on advertisement liking for high-involvement products may be because these advertisements relied on cognitive utilitarian appeals rather than advertisement liking (Petty et al., 1993; Pham et al., 2013).
The authors explored other potential explanations for their results, such as the use of a well-known celebrity, as in Volvo's use of Jean-Claude Van Damme (Agrawal and Kamakura, 1995). Future research should test the effects of other potential explanations, such as brand equity (Keller, 1993), brand attachment (Park, MacInnis, Priester, Eisingerich et al., 2010), or a range of other tactics (Hartnett et al., 2016).
The final moderator tested in this research was the viewing context for shorter advertisements. Viewers might like shorter advertisements more when they see those advertisements in a familiar context (Pham et al., 2013; Wang and Lang, 2012). In this research, however, viewing context made no difference. All advertisements, including shorter advertisements, were liked more when seen on a television, and all advertisements, including shorter advertisements, were remembered less when seen as prerolls, most likely because multiple short videos increase clutter (Anderson et al., 1998; Hammer et al., 2009).
Finally, this research tested the effects of two potential uses of shorter advertisements. Advertisers likely use short advertisements as reminders of longer advertisements (Newstead and Romaniuk, 2010). Replicating prior research (Bellman et al., 2010; Gilmore and Secunda, 1993; Neale et al., 2013), seven-second advertisements were as effective as 15-second advertisements as reminders of longer advertisements.
Another potential use of shorter advertisements is to increase clutter, by using large numbers of shorter advertisements in breaks. In a replication of previous research (Brown and Rothschild, 1993), having to remember 100 seven-second advertisements in the very-high-clutter condition made no difference to brand recognition—viewers could encode memories of these advertisements—but significantly reduced aided recall (memory storage) and unaided recall (memory retrieval; Lang, 2000). High clutter reduced unaided recall for 15-second advertisements but made no difference to brand recognition or aided recall.
Practical Implications
Given that social networks, such as Facebook, are encouraging the use of very short advertisements and that six-second advertisements also are appearing on television, research results such as these provide important insights to advertisers. This research provides models for estimating the likely effectiveness of advertisements of any length on measures of unaided brand recall, advertisement liking, and brand attitude.
The unaided brand-recall model suggests that six-second advertisements (close to the authors' seven-second advertisements) may be a good buy, even when they are half the price of a 30-second advertisement. Half of the unaided brand recall of a 30-second advertisement is delivered by its first five seconds (17 percent) of exposure. This suggests that seven- and six-second advertisements are potentially efficient alternatives to 15-second advertisements. Previous research has shown that even shorter three-second exposures (to static display advertisements) work well as reminder advertisements (e.g., Neale et al., 2013).
The results of this research also provide guidance about how to create shorter advertisements and how they can be used. Shorter advertisements potentially can overcome their exposure-time disadvantage with smart scene selection and branding. Short advertisements require simple, straight-line stories and are used best as reminders in an “always-on” reach campaign. A “burst” campaign, which may be appropriate in some conditions, such as events and launches, could use high-repetition seven- or six-second advertisements to create awareness, for less cost than 30- or 15-second advertisements. These results suggest that the effectiveness of short advertisements is not confined to younger viewers, to mobile devices, or to short-video contexts such as YouTube.
For brands that can afford to shoot and air good longer advertisements, such as 30- or 60-second advertisements, which more likely will be noticed (Rossiter and Percy, 2017), longer advertisements are better at generating cut-through, measured by unaided brand recall in this research. Longer advertisements also were liked more, with improved brand attitude.
Short advertisements allow television networks to increase clutter by showing more advertisements. The current research shows that short advertisements can be effective, but clutter should be minimized. Alternatively, short advertisements allow advertising breaks to be shorter, and networks are experimenting with one-minute advertising breaks.
Limitations
These results need replication
with different sets of advertisements;
for different products;
in different environments (e.g., social media);
in the field as well as in the lab; and
with different measures of advertising effectiveness, especially behavioral (e.g., advertisement avoidance) or sales-effectiveness measures.
In a laboratory, with forced viewing, participants paid more attention to the advertisements than they would have in natural viewing situations, which lowers the external validity of the dependent variables. Future experiments should use more ecologically valid conditions, such as the presence of distracting personal devices, and continue to explore biometrics, given the promising initial results presented here.
Unfortunately, this research did not test six-second advertisements, which did not exist when these experiments were designed. Future research should compare six-second advertisements with longer advertisements, including the 10- and 20-second advertisements used in the United Kingdom. Future research should pretest cut-down versions of longer advertisements to ensure their comparability to the original in all aspects except length, or at least systematically document how they vary.
A six-second advertisement can replace a five-second sponsorship bumper; and during a football game, a six-second advertisement can play during a review of a referee's decision. Six-second advertisements also are seen in social media, where scrolling past advertisements can be common. Future research should test short advertisements in these new environments.
Finally, future research should test the boundary conditions of the diminishing-returns effect of advertisement length. If the diminishing-returns effect on advertisement awareness requires pre-existing templates in memory (Gobet and Simon, 2000), then this effect may not be seen for new brands. The diminishing-returns effect on advertisement liking and brand attitude may require the satiation and wearout of emotional responses, which may be lacking for informational advertisements (Singh and Cole, 1993).
CONCLUSIONS
Short advertisements can be effective and efficient media buys, especially when used in conjunction with longer advertisements in a reach-with-continuity strategy, because advertising exposure has diminishing returns and because short advertisements have a role to play in a balanced advertising portfolio across media. Most of the work of a long exposure is done by its first few seconds. For optimal exposure, however, longer advertisements work better than shorter advertisements, because they attract more attention and allow more time to tell a story with greater emotional variability. Future research may find that longer advertisements are particularly important for getting the attention of a brand's light buyers and nonbuyers, who are critical for brand growth (Sharp, 2010). Longer advertisements also may be more important for new brands, which have not yet built extensive memory structures in the minds of their potential customers (Anderson, 1983; Kennedy, Sharp, and Hartnett, 2017).
Short advertisements potentially increase clutter in advertising breaks, which has negative effects on recall. Ideally, advertisers and networks should minimize clutter and use short advertisements to make shorter advertising breaks or to advertise outside of advertising breaks. This research needs replication and further extension but offers empirical guidance to advertisers about how, when, and where to use short advertisements.
ABOUT THE AUTHORS
Duane Varan is the chief executive officer, MediaScience, Austin, TX. He also oversees Beyond :30, a collaborative industry project exploring the changing media landscape. He has been awarded the Australian Prime Minister's University Teacher of the Year, among other accolades, and is published in the Journal of Advertising Research, Journal of Communication, and Journal of Economic Psychology.
Magda Nenycz-Thiel is a research professor at the Ehrenberg-Bass Institute for Marketing Science, University of South Australia. Magda's areas of expertise are category and industry growth, e-commerce, and neuromarketing. Her work has been published in Journal of Advertising Research, European Journal of Marketing, and Journal of Business Research.
Rachel Kennedy is a research professor, director, and one of the founders of the Ehrenberg-Bass Institute. Her research is focused on advertising and media knowledge to help grow brands. Kennedy is on a number of editorial boards, including for the Journal of Advertising Research. Her work can be found in JAR, Journal of Advertising, Journal of Business Research, and Journal of Retailing and Consumer Services, among other journals
Steven Bellman is a research professor at the Ehrenberg-Bass Institute. His research on media and advertising responses is funded by the Beyond :30 project, whose sponsors include television networks and advertisers worldwide. His research has appeared in the Journal of Marketing, Journal of the Academy of Marketing Science, and Management Science, and he is on the editorial boards of the Journal of Advertising, Journal of Advertising Research, and Journal of Interactive Marketing.
Appendix 1 Descriptions of the Test Advertisements Used in the Studies
Appendix 2 Questionnaire Measures
All measures were reported on a 1–6 scale, with higher scores indicating a greater amount of the measured variable (e.g., program liking).
PROGRAM LIKING
(Coulter, 1998; α = .92)
“I'm glad I had a chance to see this program.”
“If I knew this program was going to be on television, I would look forward to watching it.”
“I think that this is a program I would be interested in watching in the future.”
Strongly disagree—Strongly agree
PRODUCT CATEGORY INVOLVEMENT
(Mittal, 1995; α = .96)
[Product-category products]:
“Are important”—“Are unimportant” [reverse-scored]
“Are of no concern to me”—“Are of concern to me”
“Mean a lot to me”—“Mean nothing to me” [reverse-scored]
“Do not matter to me”—“Matter to me”
“Are significant”–“Are insignificant” [reverse-scored]
ADVERTISEMENT LIKING
(Bergkvist and Rossiter, 2007)
“Which of the following statements best describes your feelings about this ad?”
6 = “I liked it very much,” 5 = “I liked it,” 4 = “I liked it a little,” 3 = “I disliked it a little,” 2 = “I disliked it,” 1 = “I disliked it very much”
BRAND ATTITUDE
(Bergkvist and Rossiter, 2007)
“What is your attitude toward [brand name] as a brand of [product category]?”
Bad—Good
- Received March 11, 2018.
- Received (in revised form) November 12, 2018.
- Accepted December 19, 2018.
- Copyright© 2020 ARF. All rights reserved.
REFERENCES
ARF MEMBERS
If you are a member of the Advertising Research Foundation, you can access the content by logging in here
Log In
Pay Per Article - You may access this article (from the computer you are currently using) for 30 days for US$20.00
Regain Access - You can regain access to a recent Pay per Article purchase if your access period has not yet expired.