S7 Airlines segmented their audience into five groups based on their progress through the sales funnel. They created separate retargeting lists and experimental ad campaigns targeted to each group in Yandex.Direct. For each group, they adjusted bids and tailored ad copy to be most relevant. This allowed them to personalize messaging and see a 48% decrease in CPA and 73% increase in sales, showing the most promising groups were those farther in the sales funnel who had viewed search results or started the reservation process previously.
2. S7 Airlines (part of “Siberia” airline)
S7 brand launched
2005
“Siberia” airline founded
1992
Russian passenger airline
10.6 M
passengers in 2015
3. Period:
Tools used:
July 2016
Segment audience according to sales funnel and
determine most promising groups
› Search ads – Yandex.Direct
› Yandex.Metrica
Maximize sales
Goals
4. Putting things in context
4
An individual approach to the
client is the foundation of
effective communication.
S7 wanted to distinguish
groups of promising clients
and convey a relevant
message to each of them. To
achieve this, they segmented
users according to the sales
funnel and applied separate
bid adjustments in search.
www.s7.ru
5. Audience segmentation using the sales funnel
5
S7 divided their audience into five segments
according based on how far they had
proceeded in the sales funnel:
› “First timers,” who had never visited the
site before
› “Bouncers,” who had been on the site
but did not view the search results
› Users who viewed search results by
flight but didn’t make a reservation
› Users who started but did not complete
the reservation process
› “Loyal” users who successfully
completed a reservation
6. Solution
Set up goals in
Yandex.Metrica
based on sales funnel
Step 1 Step 2 Step 3 Step 4 Step 5
Create retargeting
lists in Yandex.Direct
Divide traffic
betweennew and
repeat visitors
Set bid adjustments
for each segment
Show repeat users
discounted offers
10. Creating retargeting lists in Yandex.Direct
10
Then they created a retargeting list
for each level of the sales funnel.
For example, the retargeting list
“Step 2: viewed search results”
has two rules: users who
completed the goal “search
results” WITHOUT completing
“data entry.”
Retargeting lists must be unique.
Their conditions must not overlap,
so that no user should fall into
more than one, and should be
relevant over a period of 30 days.
direct.yandex.ru
12. 12
Dividing traffic between new and repeat visitors
direct.yandex.ru
S7 made four copies of each
campaign in the experiment —
one for each step of the sales
funnel, with its own retargeting
list.
The keywords and campaign
settings were the same for the
four copies. The bid adjustments
and ad texts were different,
reflecting the user group each
campaign focused on.
This setupallowed S7 to
personalize their messages,
making their ads relevant for each
group.
To the right you can see a
breakdown of an original
campaign and its four copies
(experimental campaigns with
their own retargeting lists).
Flight_MOS—
from Russia_to
Russia—MOS
Flight_MOS—from Russia_to Russia—MOS + retargeting_viewed search results
Flight_MOS—from Russia_to Russia—MOS + retargeting_started reservation process
Flight_ MOS—from Russia_to Russia—MOS + retargeting_reservation complete
Flight_MOS—from Russia_to Russia—MOS + retargeting_visited site
Original campaign 4 copies — experimental campaigns
14. yandex.ru yandex.ru
Showing repeat users discounted offers
14
Users that were absolute “first timers” saw the ads from the
original campaign.
Previous visitors to the site who had gone farther down the sales
funnel, however, saw ads from the experimental campaigns. These
ads contained additional information about a discount promo-code.
The link led to a separate landing page for the “It’s Time to Fly”
sale, offering a 2% discount off any ticket.
www.s7.ru
16. direct.yandex.ru direct.yandex.ru
Setting bid adjustments for each segment
16
To correctly divide their traffic, S7 applied a -100% bid
adjustment in the original campaign for users who had already
visited the site, guaranteeing that their ads only appeared to
users who were absolute “first timers.”
To prevent new users from seeing the experimental ads, they set
bids 4-6 times lower than in the original campaign for “first timers.”
This guaranteedthat the experimental campaigns would lose in
auctions for users who hadn’t been to the site. Conversely, they
applied a +1100% bid adjustment for return users, making sure
that only the intended audience would see the promotions intended
for return visitors.
18. metrika.yandex.ru metrika.yandex.ru
Quality control results: creating Yandex.Metrica reports
To check the quality of your traffic segmentation, you can create a report in
Yandex.Metrica on campaigns aimed at “first timers.”
And a report on campaigns focusing on return clients.
18
In the “Yandex.Direct summary” report, they clicked on “Visits in which” → “Sources” → “Last source” → “Yandex.Direct” → “Campaigns”
and listed all the experimental campaigns.
19. Quality control: “first timer” campaigns
19
The “percentage of new
users' sessions” metric in
the report shows that new
users made up roughly 93%
of those whose last source
were the experimental
campaigns aimed at “first
timers.”
metrika.yandex.ru
20. metrika.yandex.ru
Quality control: creating a report in Yandex.Metrica
20
The “percentage of new
users’ sessions” metric in the
report showed that less than
8% were new, and the other
92% had returned to the S7
site from Yandex search.
22. Results: determining the most promising groups
22
The conversion rate for users
that made it to the search
results was 6.7 times higher
than first-timers. Users who
had made a reservation before
were 7.7 times more likely to
book tickets than new users.
The CPA for these users were
85% and 87% lower
respectively.
Most reservations (42%) were
made by new users. 33% come
from users who started
reservation process, and 14%
were repeat orders.
Normalized data from “New users” presented as a percentile difference, and absolute values of the relative sizes.
42%
33%
14%
6%5%
First timers
Started reservation process
Completed reservation
Viewed search results
Bouncers
0.92
0.15
0.4
0.13
1
0
0,2
0,4
0,6
0,8
1
1,2
CPA
1.09
6.72
2.64
7.74
1
0
2
4
6
8
10
CR
Transactions
+572%
+674%
–85% –87%
23. The approach we’ve described — separate promo landing
pages with targeted messages for each segment — improved
the campaigns’ performance and optimized communications
through increased personalization.
23
Alexander Nazarov
Performance Group Head, Performics (Publicis Media)
24. S7 Airlines
24
Sales from search campaigns CPA
- 48%+ 73%
Period
July 2016 — current
Tools used
› Yandex.Direct
› Yandex.Metrica
Objectives
1. Segment audience according to
sales funnel and determine most
promising groups
2. Maximize sales
Airline
Before
After
25. Case prepared in cooperation
with
www.performics.com
Performics (Publicis Media)