Face(book)ing the Truth: Initial Lessons Learned using Facebook
Advertisements for the Chatbot-delivered Elena+ Care for
COVID-19 Intervention
Joseph Ollier
1a
, Prabhakaran Santhanam
2b
and Tobias Kowatsch
2,3 c
1
Center for Digital Health Interventions, Chair of Technology Marketing,
Department of Technology, Management, and Economics, ETH Zurich, Zurich, Switzerland
2
Center for Digital Health Interventions, Department of Technology, Management, and Economics,
ETH Zurich, Zurich, Switzerland
3
Center for Digital Health Interventions, Institute of Technology Management, St. Gallen, Switzerland
Keywords: Facebook Advertisements, Social Media, Digital Health, Chatbots, Elena+ Care for COVID-19.
Abstract: Utilizing social media platforms to recruit participants for digital health interventions is becoming
increasingly popular due to its ability to directly track advertising spend, number of app downloads and other
metrics transparently. The following paper concerns the initial tests completed on the Facebook Ad Manager
platform for the chatbot-delivered digital health intervention Elena+ Care for COVID-19. Eleven
advertisements were run in the UK and Ireland during August/September 2020, with resulting downloads,
post (i.e. advert) reactions, post shares and other advertisement engagement metrics tracked. Key findings
from our advertising campaigns highlight that: (i) static images with text function better than carousel of
images, (ii) Android users download and exhibit greater engagement behaviors than iOS users, and (iii)
middle-aged and older women have the highest number of downloads and the most engaged behaviors (i.e.
reacting to posts, sharing posts etc.). Lessons learned are discussed considering how other designers of digital
health interventions may benefit and learn from our results when trialing and running their own ad campaigns.
It is hoped that such discussions will be beneficial to other health practitioners seeking to scale-up their digital
health interventions widely and reach individuals in need.
1 INTRODUCTION
Social media platforms are becoming an increasingly
advantageous route to recruit participants for health
interventions (Arigo et al., 2018). They are
particularly helpful in digital interventions utilizing
smartphone technology whereby the tracking of
downloads from advertisements is easily facilitated
and costs per new participant measured (Platt et al.,
2016). The following paper overviews the
preliminary testing campaigns for the Elena+ Care for
COVID-19 (www.elena.plus) digital health
intervention, which offers chatbot-led digital
coaching on various facets of an individual’s lifestyle
and health promoting behaviors (e.g. sleep, mental
a
https://orcid.org/0000-0001-8603-0793
b
https://orcid.org/0000-0002-9506-4888
c
https://orcid.org/0000-0001-5939-4145
health, physical activity etc.) that may be under
increased strain during the COVID-19 pandemic
(Ollier, Joseph & Kowatsch, 2020). It is hoped that be
demonstrating early lessons learned from early
testing campaigns, individuals running future
campaigns may be able to identify relevant patterns
earlier and save funding for more cost-effective
advertisements to help individuals in need.
The paper continues then by overviewing the
rising importance of social media in recruiting for
digital health interventions, and discussing the app
used for recruitment, Elena+ Care for COVID-19.
Following this we explain the methodological
framework and continue by exploring data exported
from the Facebook Ad Manager Platform. Lastly, we
conclude with lessons learned and future
Ollier, J., Santhanam, P. and Kowatsch, T.
Face(book)ing the Truth: Initial Lessons Learned using Facebook Advertisements for the Chatbot-delivered Elena+ Care for COVID-19 Intervention.
DOI: 10.5220/0010403707810788
In Proceedings of the 14th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2021) - Volume 5: HEALTHINF, pages 781-788
ISBN: 978-989-758-490-9
Copyright
c
2021 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
781
considerations for other designers of digital health
interventions.
2 BACKGROUND
2.1 Social Media Recruitment
In scaling up digital health interventions, and after the
hard work of designing, implementing and testing a
digital product, comes the next challenge of
successful recruitment (Arigo et al., 2018; Platt et al.,
2016). This challenge is often given less thought in
planning stages, however, and less focus in academic
research. A cursory search of Google Scholar with the
term “digital health intervention”, for example,
returns a huge 2 440 000 search results, however
when simply adding the additional word
“recruitment” at the end of the search term only
154 000 search results are found, and when “social
media recruitment” is added a comparably measly
136 000 search results are found. Therefore despite
the importance of successful recruitment strategies
for health interventions, social media recruitment is a
relatively neglected area in healthcare research.
Recruitment success is vital however for all
interventions, and particularly those aimed at
population level health concerns such as obesity
(Chou et al., 2014), mental health (Sanchez et al.,
2020), alcoholism (Wozney et al., 2019) or the
current paper’s example Elena+ Care for COVID-19
(Ollier, Joseph & Kowatsch, 2020). Recruitment in
such contexts is vital to generate enough participants
to both help the public at large through cutting edge
science as well as power statistical analyses which
can analyze efficacy. Additionally, for other health
interventions which target a smaller population group
and have patient access granted via a medical
organization, following up on initial success in an
academic setting with wider recruitment will likely be
a vital part of the marketing mix in translating a
promising study to a digital start up’s new digital
product (World Health Organization, 2009).
As the proportion of spending on social media
versus traditional media has been generally rising
(Ma & Du, 2018) as marketing practice has utilized
and adapted segmenting, targeting and positioning to
better effectiveness with social media platforms
(Canhoto et al., 2013; Kotler et al., 2016) for
healthcare researchers the highly utilizable nature of
social media is becoming an increasingly attractive
recruitment route. It offers more focused targeting
and opens the door for cost-effective recruitment of
participants independent of medical organizations,
which may be particularly valuable in scaling up
digital health intervention ideas into real products
actively recruiting participants (World Health
Organization, 2009).
2.2 Elena+ Care for COVID-19
The Elena+ digital health intervention is one such
example of an emerging digital product scaling up
with the use of social media advertisements for
recruitment. Elena+ (Ollier, Joseph & Kowatsch,
2020) was developed by a group of researchers during
Spring 2020 as the COVID-19 pandemic spread
across the globe. It utilizes a chatbot embedded
smartphone application to address the collateral
damage of social distancing and lockdowns i.e. that
health promoting behaviors within individual’s
lifestyles may be under increased strain (Javed et al.,
2020). Therefore, the chatbot offers 43 coaching
sessions focusing on psychoeducational training and
activities in the fields of: COVID-19 information,
physical activity, sleep, anxiety, loneliness, mental
resources and diet and nutrition. The core recruitment
strategy for Elena+ was based upon Facebook
advertisements, of which findings from preliminary
experimentations are presented in this paper.
3 METHODOLOGY
During August-September 2020 advertisements were
created and displayed using the Facebook Ad
Manager platform for Facebook Businesses. A
variety of advertisements were tested which were
either (i) static images with text or (ii) a carousel of
images. Advertisements were run on both iOS and
Android smartphone operating systems and for users
within UK and Ireland in English only, with relevant
links to the app store included on the advertisements
and Facebook API for developers implemented so
that downloads resulting from the link click could be
tracked.
In total eleven advertisements were ran online, of
which; (i) eight were of static image with text type
(Figure 1) and three were carousel of images type
(Figure 2); (ii) seven were aimed at iOS users and four
at Android users; (iii) three were in the UK and eight
were in Ireland. Each advertisement was run for up to
one week, with a budget of circa 11 USD (10 CHF)
per day. Various metrics were collected by the
Facebook Ad Manager platform, and for the current
analyses we looked at downloads as the primary
result of interest, however, other engagement related
metrics for the advertisement (post engagement, page
Scale-IT-up 2021 - Workshop on Scaling-Up Healthcare with Conversational Agents
782
engagement, post reactions, post shares) were also
downloaded from the platform. Descriptive statistics
were calculated to outline the relative performance of
these advertisements.
Figure 1: Static Image with Text Advertisement.
Figure 2: Carousel of Images Advertisement.
4 RESULTS
All advertisements ran on Facebook between 13
th
August to 22
nd
September 2020, with a resulting 186
downloads at a cost of 463.24 CHF (approx. 507
USD). Although we specified a guideline budget of
10 CHF per day and to pay “cost per result” (i.e. by
downloads) some variation occurred due to increased
advertisement engagement (as Facebook states may
occur in their website documentation).
Advertisements typically ran for one week; however,
due to requirements of the Elena+ project (e.g.
maintenance tasks) outside the scope of this paper,
some fluctuations occurred. A summary can be seen
below of each advertisement, the total number of days
it ran, and total budget spent.
Table 1: Advertisement Summary.
Advert name Days Cost
Ireland Static 01
(
A
ndroid
)
7 70.40
Ireland Static 02
(
A
ndroid
)
7 76.22
UK Static 01
(
A
ndroid
)
7 59.37
UK Static 02 (
A
ndroid) 7 75.74
UK Static 01 (iO
S
) 7 60.38
Ireland Static 01
(
iO
S
)
3 3.98
Ireland Static 02
iO
S
)
5 38.32
Ireland Static 03
iO
S
)
5 35.96
Ireland Carousel 01 (iOS) 5 32.30
Ireland Carousel 02 (iOS) 3 2.10
Ireland Carousel 03 (iOS) 4 3.98
4.1 Elena+ App Downloads
Firstly, looking at simple comparisons between the
types of advertisements run, the static image with text
advertisement functioned much better than using a
carousel of images with regard to downloads.
Figure 3 shows that from the three carousel
advertisements run on iOS only one download
resulted. Whereas for the eight static image with text
advertisements (which were ran on both iOS and
Android) resulted in 185 downloads. When
considering the average ad return for downloads (i.e.
no. of downloads divided by no. of advertisements
used, hereafter referred to as AAR), static image with
text resulted in 23.125 users per ad, whereas carousel
ad only 0.33 users per ad.
Face(book)ing the Truth: Initial Lessons Learned using Facebook Advertisements for the Chatbot-delivered Elena+ Care for COVID-19
Intervention
783
Figure 3: Downloads by Advertisement Type.
Figure 4: Downloads by Operating System.
When making a simple comparison between
operating system of the seven advertisements aimed
at iOS users and four advertisements at Android
users, Figure 4 shows that running advertisements on
Android resulted in much higher numbers of
downloads. From the seven advertisements aimed at
iOS users, 48 downloads resulted (AAR of 6.857)
whereas for Android it was much higher (AAR of
34.5).
Figure 5: Downloads by Country.
A simple comparison by country in Figure 5
shows that the eight advertisements in Ireland
resulted in 107 downloads (AAR of 13.375) whereas
the three advertisements in the U.K. resulted in 79
downloads (AAR of 26.333). Figure 6 also shows
gender and age of downloads, whereby females aged
more than 35 years are downloading the app in greater
numbers.
Figure 6: Downloads by Age and Gender.
A summary of number of downloads for all
factors discussed above is shown below for Ireland
(Figure 7) and the U.K (Figure 8).
1
185
0
50
100
150
Ca
r
ous
e
l
S
tati
c
AdType
Elena+ Downloads
Downloads by Ad Type
138
48
0
50
100
Android
iOS
Operating System
Elena+ Downloads
Downloads by OS
107
79
0
30
60
90
Ir
e
land
U
K
Country
Elena+ Downloads
Downloads by Country
4
1
7
5
22
6
32
13
39
13
34
9
0
10
20
30
40
1
8
-2
4
2
5-34
35-44
4
5
-5
4
5
5-64
65+
Age
Elena+ Downloads
Gender
fem ale
male
Downloads by Age and Gender
Scale-IT-up 2021 - Workshop on Scaling-Up Healthcare with Conversational Agents
784
Figure 7: Downloads by Ad Type, Operating System, Age
and Gender in Ireland.
Figure 8: Downloads by Ad Type, Operating System, Age
and Gender in in the U.K.
4.2 Advertisement Engagement
Metrics
To complement the above findings a brief summary
of advertisement engagement metrics is provided in
this section overviewing how the advertisements also
affected a few selected promotional metrics. These
include: (i) post engagement, (ii) page engagement,
(iii) post reactions, and (iv) post shares.
Post engagement is defined by Facebook as “all
actions people take involved your ads” such as
“reacting to, commenting on or sharing the ad,
claiming an offer, viewing a photo or video, or
clicking on a link.”. Facebook defines this as a useful
measure to see how relevant the advertisements were
to the recipients.
Figure 9: Post engagement by Age, Gender, OS and Ad
Type.
As can be seen in Figure 9, middle-aged and older
women have the highest total summed quantity of
post engagement. Static Android advertisements
drive best post engagement, as no carousel
advertisements were run on Android, comparison is
not possible however.
Page engagement includes “interactions with your
Facebook Page and its posts” and actions such as
“liking your Page, reacting with Love’ to a post,
checking in to your location, clicking a link and
more.”. As with post engagement, Facebook notes
page engagement as a useful metric to see how
1
1
4
3
7
2
15
4
17
3
16
6
2
1
7
1
8
44
Carousel Static
Android iOS
18-24
25-3
4
3
5
-44
45-54
55-6
4
65
+
18-24
25-3
4
3
5
-44
45-54
55-6
4
65
+
0
5
10
15
0
5
10
15
Age
Results
Gender
fem ale
male
Downloads in Ireland
by Age, Gender, OS and Ad Type
2
3
2
11
4
10
7
8
4
6
2
3
1
6
2
7
1
Carousel Static
Android iOS
18
-
24
25-34
3
5-4
4
45
-5
4
5
5-6
4
65+
18
-2
4
2
5-3
4
3
5-4
4
4
5-54
55
-
64
65+
0
3
6
9
0
3
6
9
Age
Results
Gender
fem ale
male
Downloads in the UK
by Age, Gender, OS and Ad Type
41
21
44
28
38
55
79
6262
66
83
54
12
2
34
13
44
22
84
39
115
40
96
35
8
11
5
15
4
32
10
56
12
53
9
Carousel Static
Android iOS
18-
24
25-
34
35-44
45-54
55-64
65+
18-24
25-34
35-44
45-54
55-64
65+
0
30
60
90
120
0
30
60
90
120
Age
Post.engagement
Gender
fem ale
male
Post Engagement
by Age, Gender, OS and Ad Type
Face(book)ing the Truth: Initial Lessons Learned using Facebook Advertisements for the Chatbot-delivered Elena+ Care for COVID-19
Intervention
785
relevant the advertisements were to the given
audience.
Figure 10: Page engagement by Various Factors.
Figure 11: Post reactions by Various Factors.
The summed total on page engagement (Figure
10) is highest again for static Android advertisements,
and women from 35+ years again exhibit the highest
engagement.
Post reactions includes all types of reactions to an
ad, such as reacting with “like, love, haha, wow, sad
or angry.”. Facebook states that this helps
advertisements perform better, as individuals
automatically start to follow updates related to the ad
(i.e. new comments, reactions etc.) which can drive
further engagement with a business page and content.
As can be seen in Figure 11, post reactions are once
more best performing for static advertisements on
Android, and at least for Android users, the
aforementioned pattern of middle-aged and above
women scoring highest appears to be exhibited.
Figure 12: Post shares by Various Factors.
Post shares includes whether individuals share the
advertisement, for example on their or others’
timelines, in groups, on other pages etc. However, it
does not measure any subsequent engagement after
that point (for example, comments on shared posts, or
whether individuals that see the shared post navigate
to the original ad and take any actions). Similar to
previous results, Figure 12 indicates that general
middle-aged and older females are primarily sharing
the advertisements based on available results thus far.
41
21
44
28
38
55
79
6262
66
83
54
12
2
34
13
44
22
84
39
115
40
96
35
8
11
5
15
4
32
10
56
12
53
9
Carousel Static
Android iOS
18-
24
25
-
34
35-44
45-54
55-
64
65
+
18-24
25-34
35
-
44
45
-
54
55-64
65
+
0
30
60
90
120
0
30
60
90
120
A
g
e
Page.engagement
Gender
female
male
Page Engagement
by Age, Gender, OS and Ad Type
1
1 111
2
5
4
2
1
3
1
11
2
11
Carousel Static
Android iOS
1
8-2 4
25-34
3
5-4 4
45-
5
4
5
5-6 4
65
+
1
8-2 4
2
5
-
3
4
3
5-4 4
4
5
-
5
4
5
5-6 4
6
5+
0
1
2
3
4
5
0
1
2
3
4
5
Age
Post.reactions
Gender
fem ale
male
Post reactions
by Age, Gender, OS and Ad Type
1
3
1
2
1
1
Carousel Static
Android iOS
18-
24
25-
34
35-
44
45-
54
55-64
65+
18-
24
25-34
35-44
45-5
4
55-
64
65+
0
1
2
3
0
1
2
3
Age
Post.shares
Gender
female
male
Post shares
by Age, Gender, OS and Ad Type
Scale-IT-up 2021 - Workshop on Scaling-Up Healthcare with Conversational Agents
786
5 DISCUSSION
Our findings have highlighted that Android users are
more responsive to the advertisements in terms of
both downloads, the primary metric of interest, as
well as engagement metrics. This shows that Android
may be a more cost-effective platform as more results
were found directly from the advertisements
themselves, and via their higher engagement,
Android users effectively market the app further to
their contacts via their greater activity and
interactions with both the advertisements and the
Elena+ Facebook page. It is possible therefore that
Android users as a whole are less data privacy
sensitive, and thus interact more. This may be likely,
as Android phones can range from relatively cheap to
very expensive when bought firsthand and brand new,
whereas iOS models are always relatively expensive.
As income level is indicative of education (Tolley &
Olson, 1971), and higher digital literacy results in
more privacy protective behaviors (Park, 2011), it
may be due to less affluent/less educated individuals
being represented amongst Android users.
Regarding the prevalence of middle-aged aged
and above females being primary downloaders and
those exhibiting engaged behaviors, the authors feel
that this perhaps is related to either the ad or app
content. It may be simply that men and youth found
the simple advertisements we used less attractive, as
these groups, stereotypically, exhibit greater affinity
to new technology and technological devices (Olson
et al., 2011; Venkatesh & Morris, 2000). Thusly, for
targeting youth or men, perhaps better ad content
(media, text) was required. Alternatively, however, it
may simply represent that middle-aged and older
women are a particular at-risk group for the
intervention used in this study i.e. Elena+ Care for
COVID-19. Often middle-aged and older women are
responsible for care roles in the family (Dahlberg et
al., 2007) (i.e. caring for children, caring for elderly
relatives or both), and may also still be in
employment. Adding the additional strain of the
pandemic and social distancing/isolation
requirements on top of all other respective duties
could therefore disproportionally affect the group of
middle-aged women and could be an alternative
explanation as to why they downloaded/engaged
most often after being exposed to advertising.
6 CONCLUSIONS
In summary these preliminary findings from the
Facebook advertising campaigns are not to be taken
as definitive proof due to the relatively small budget
and duration of the campaigns and the fact that there
are likely many variations of success contingent on
the medical intervention being utilized and patient
population. However, it is hoped that by sharing our
findings on utilizing social media for driving
downloads of app, Elena+ Care for COVID-19, others
may benefit and that needless costs are not duplicated
by repeatedly running trial and error advertising
campaigns to find what works best, and may enable
practitioners to draw meaningful conclusions in their
own fields more speedily, saving budget for reaching
potential beneficiaries of their digital health
interventions.
REFERENCES
Arigo, D., Pagoto, S., Carter-Harris, L., Lillie, S. E., &
Nebeker, C. (2018). Using social media for health
research: Methodological and ethical considerations for
recruitment and intervention delivery. Digital Health,
4, 205520761877175.
https://doi.org/10.1177/2055207618771757
Canhoto, A. I., Clark, M., & Fennemore, P. (2013).
Emerging segmentation practices in the age of the
social customer. Journal of Strategic Marketing, 21(5),
413–428.
https://doi.org/10.1080/0965254X.2013.801609
Chou, W. S., Prestin, A., & Kunath, S. (2014). Obesity in
social media: a mixed methods analysis. Translational
Behavioral Medicine, 4(3), 314–323.
https://doi.org/10.1007/s13142-014-0256-1
Dahlberg, L., Demack, S., & Bambra, C. (2007). Age and
gender of informal carers: A population-based study in
the UK. Health and Social Care in the Community,
15(5), 439–445. https://doi.org/10.1111/j.1365-
2524.2007.00702.x
Javed, B., Sarwer, A., Soto, E. B., & Mashwani, Z.-R.
(2020). Impact of SARS-CoV-2 (Coronavirus)
Pandemic on Public Mental Health. Frontiers in Public
Health, 8, 292.
https://doi.org/10.3389/fpubh.2020.00292
Kotler, P., Kartajaya, H., & Setiawan, I. (2016). Marketing
4.0: Moving from traditional to digital. John Wiley &
Sons.
Ma, J., & Du, B. (2018). Digital advertising and company
value: Implications of reallocating advertising
expenditures. Journal of Advertising Research, 58(3),
326–337. https://doi.org/10.2501/JAR-2018-002
Ollier, Joseph & Kowatsch, T. (2020). Elena+ Care for
COVID Website. 2020. elena.plus
Face(book)ing the Truth: Initial Lessons Learned using Facebook Advertisements for the Chatbot-delivered Elena+ Care for COVID-19
Intervention
787
Olson, K. E., O’brien, M. A., Rogers, W. A., Charness, N.,
Olson, K. E., Rogers, W. A., O’brien, M. A., &
Charness, N. (2011). Diffusion of Technology:
Frequency of use for Younger and Older Adults. Ageing
Int, 36, 123–145. https://doi.org/10.1007/s12126-010-
9077-9
Park, Y. J. (2011). Digital Literacy and Privacy Behavior
Online. Communication Research, 40(2), 215–236.
https://doi.org/10.1177/0093650211418338
Platt, T., Platt, J., Thiel, D. B., & Kardia, S. L. R. (2016).
Facebook Advertising Across an Engagement
Spectrum: A Case Example for Public Health
Communication. JMIR Public Health and Surveillance,
2(1), e27. https://doi.org/10.2196/publichealth.5623
Sanchez, C., Grzenda, A., Varias, A., Widge, A. S.,
Carpenter, L. L., McDonald, W. M., Nemeroff, C. B.,
Kalin, N. H., Martin, G., Tohen, M., Filippou-Frye, M.,
Ramsey, D., Linos, E., Mangurian, C., & Rodriguez, C.
I. (2020). Social media recruitment for mental health
research: A systematic review. Comprehensive
Psychiatry, 103, 152197.
https://doi.org/10.1016/j.comppsych.2020.152197
Tolley, G. S., & Olson, E. (1971). The interdependence
between income and education. Journal of Political
Economy, 79(3), 460–480.
Venkatesh, V., & Morris, M. G. (2000). Why Don’t Men
Ever Stop to Ask for Directions? Gender, Social
Influence, and Their Role in Technology Acceptance
and Usage Behavior. MIS Quarterly, 24(1), 115–139.
https://doi.org/10.2307/3250981
World Health Organization. (2009). Practical guidance for
scaling up health service innovations.
Wozney, L., Turner, K., Rose-Davis, B., & McGrath, P. J.
(2019). Facebook ads to the rescue? Recruiting a hard
to reach population into an Internet-based behavioral
health intervention trial. Internet Interventions, 17,
100246. https://doi.org/10.1016/j.invent.2019.100246
Scale-IT-up 2021 - Workshop on Scaling-Up Healthcare with Conversational Agents
788