About client and tasks
The CarMoney audience is car owners who want to take out a loan secured by a vehicle. The advantage of the fintech service of car loans is that it is easier and faster to get money here than in a bank: fewer documents are required, and the speed of decision-making is higher.
The service already had search campaigns and advertising in YAN and KMS, but the results needed to be improved. For example, campaigns for ad networks did not convert well, so in general, ads brought bids at a cost that did not suit the client.
Our task was to develop and implement a strategy aimed at increasing the number of loans issued within a given issuance cost.
Period of work: February – May 2019.
Phase 1: Optimization of ongoing campaigns, growth within base inventory
At the start, we redesigned contextual campaigns. In search advertising, Yandex.Direct and Google Ads segmented key phrases: target queries, credits, brand queries. For each segment, we made a division by city and prepared relevant ads for each segment of the audience.
We relaunched direct demand campaigns (loan secured by PTS, etc.) in YAN and KMS: we divided campaigns into segments (for example, according to the structure of key phrases), for each of them we wrote unified ads. Based on CPA indicators and the number of micro-conversions, we filtered sites every day, and gradually reached the set indicators.
At the same time, we launched targeted campaigns in myTarget: multi-format for target interests, static retargeting for site visitors, contextual targeting, look-alike for site visitors.
This stage took about 1.5 months. As a result, we managed to increase the volume of applications by 19% and reduce the cost of applications by 8% in comparison with the same period last year.
Stage two: developing a growth strategy
It was clear that within 2-3 months the search performance ceiling would be reached: ads for targeted and credit-related queries in 92% of cases were located at the top of search results. The volumes of impressions in the regions of the Russian Federation were practically exhausted, and the Moscow region and St. Petersburg did not provide an acceptable CPA on large volumes.
Impressions at the top of the page,%
Working with near-topic queries in search and networks did not bring the desired CPA result. The data available on social networks did not allow us to single out a significant number of people according to our interests.
We decided to try programmatic advertising and DMP. Google Display & Video 360 met our tasks best of all: a large amount of data, built-in optimizer, integration with any DMP, independent campaigning.
There are other ways to interact with a new audience, for example, an out-of-the-box branded display campaign, TV advertising. But for CarMoney, these tools were not suitable: branded media and TV campaigns provide a large reach (and we need a narrow audience) and are too expensive.
The effective capacity of the DV 360 is sufficient but finite. We realized that after a while we would run into the same thing as in the context and social networks. Therefore, we decided to expand the existing segments with external data, for this we immediately integrated the pixel of the DMP platform Aidata into the site. It collected information about the interests of users who came to the site and made a conversion.
To get a complete picture of the effectiveness of the programmatic campaign, we used AdRiver tracking. With its help, post-view data was transmitted to Google Analytics. In addition, we linked data on expenses from advertising systems and a funnel from a CRM client using our own reporting automation system “Artics. Figures”.
The main advantages of programmatic placement are flexibility in management and optimization for specific events after pixels are installed on the site.
There are many Programmatic platforms on the market, the largest one is from Google – Display & Video 360. It allows you to buy display ads, plan and implement campaigns in various ad networks.
In this case, the following features of Display & Video 360 came in handy:
- a large amount of data that is regularly updated;
- an internal optimizer, which is important for performance campaigns;
- there is a possibility of integration with any DMP;
- full independent campaign management, maximum transparency.
DV 360 has all the tools you need to work with programmatic campaigns of various difficulty levels: setting conversion tags and an optimization strategy for them, Brand Safety settings and filters, a multifunctional reporting module, audience constructor, audience portrait analysis, using your own advertiser’s data, and others. …
The DV 360 platform can be accessed directly through Google or through official resellers. It is important to keep in mind that for direct purchase from Google, you need to purchase a certain volume. There are no minimum budgets for launching campaigns in DV 360; first of all, you need to build on the goals and objectives of the campaign.
Any programmatic tool, and in particular DV 360, can solve a variety of tasks: from getting inexpensive traffic to increasing the amount of traffic and increasing the conversion of performance channels with tracking all user taps until the final conversion (action on the site, target lead from CRM, etc. .d.).
At the same time, you should not focus solely on direct conversions in programmatic campaigns. They will be, but CPA will be significantly higher compared to classic performance channels. Retargeting campaigns will be an exception, since we work with our own audience in them.
Programmatic is always part of the overall marketing strategy, and should be considered in conjunction with all other tools used. Non-targeting programmatic placements should be assessed by post-view indicators and related data: the growth of new users from contextual campaigns and organics, the level of user engagement in conversion chains
Phase Three: Launch DV 360, DMP Integration and Data Collection
We launched a programmatic campaign with customized post-view analytics and data transfer to Google Analytics.
The main audience groups that we used for targeting:
- interests (Affinity & In-Market): loans and borrowings – standard interest groups from Google;
- custom audience segment based on the affinity of visitors to sites of competitors and peers;
- look-alike by converters – based on the audience collected by the DV 360 pixel.
To make the communication in messages on the landing page and with the call-center specialists relevant, we decided to find out who our audience is. To do this, we clustered the CarMoney audience and saw that there are segments of users with similar interests: repairs, vacations, health, trading, refuseniks of banks. Assuming that clients take loans for these needs, we built LAL and used the same categories of interests as targeting.
For the test placement, we chose regions with historically good indicators in terms of conversion volumes, but with narrow coverage – Krasnodar, Nizhny Novgorod, Chelyabinsk.
In March, we achieved maximum coverage in these regions (with budget growth, lead growth stagnated), so we scaled campaigns to all regions of the client’s presence. This allowed us to significantly expand our reach and collect a large pool of statistics to assess the results of the increase in conversions due to media campaigns.
For each of the selected target audiences, we have created separate creatives in order to better meet the needs of target audiences. At the same time, each idea was implemented in two versions: the most consistent with the corporate identity of CarMoney (with a white background and a photo) and non-standard (graphics with a minimum of branding). Here are examples of messages and announcements:
we went to the refuseniks in the bank with the communication – “Refused in the bank? Come to CarMoney! “;
to those who are interested in repairs – “It’s time to complete the repair!”;
to those who are interested in trading – “Don’t miss the trend! We will lend a shoulder. “
And separately for all segments, we added a standard creative with a low rate and benefits.
In Display & Video 360, the cost of conversions was high at the start, but the platform gradually learned and over time began to attract better quality users. The cost of conversion in the third month has reached indicators similar to contextual advertising.
Stage four: working with DMP data
In parallel with the campaigns for the segments available in the site interface, we were creating custom segments.
DMP machine learning algorithms processed the results of promotion in DV 360 and formed homogeneous audience groups: family people, those who are interested in investment services and business, travel, health, repairs.
Segments from Aidata overlapped with those interests for which we targeted in DV 360. However, thanks to external data, we were able to expand and get new segments (about 15% of users additionally), superimpose them to highlight the most target audience within the segment due to additional parameters.
How campaigns were optimized
The primary parameters for evaluating and optimizing campaigns were micro-conversions on the site: click on the “Submit request” buttons, start filling out an application, and others. This allowed us to collect a larger amount of data; the correlation of micro-conversions to leads was quite high. For example, the conversion from interaction with the calculator to the lead reached 14.3%. Based on the data obtained (after a two-week collection of statistics), we estimated the cost of micro-conversions and, based on the results of comparison with the reference segments, made a decision on segment scaling.
To understand how leads converted to loans, we focused on the average conversions in the funnel and estimated the predicted conversion to the target lead and the issue. In parallel, in order to collect more accurate statistics, we began the process of integrating post-view parameters into the client’s CRM.
The difficulties we faced
The main problem in optimizing these campaigns was the time lag of the “income” of post-view conversions for campaigns and the required long period (two weeks) for collecting statistics. During the period of statistics collection, the CPL could jump sharply, which could lead to incorrect conclusions when assessing effectiveness.
To work out the situation, we chose the following algorithm of actions: for two weeks, while statistics were being collected, we made minimal changes to the campaigns, then we looked at the conversions that came and made the final decision on segment optimization.
DV 360 not only gave us direct conversions, but also increased the share of conversions that are made in other channels. To track this, we created a report in which we passed the lead parameter (“after viewing the banner”), and the flow from channel to channel was seen in the report by sources. The total growth in the number of SERPs since the launch of DV 360 was 14%, the number of leads from contextual advertising due to programmatic advertising has grown by about a third.
We stayed within the admissible KPIs for CPA and the client’s economy, received more returns than we would have received with an increase in expenses in a context where the return on each ruble invested began to decline critically. In addition, we have seen an increase in offline and mobile search results – in those channels where conversions are not attributed to our project.