The Beginning of Multi-Touch Attribution in Mobile Advertising is Here!

Becky Doles

When we first launched MobileAppTracking, the world of mobile advertising was still small and as such our clients worked with just a handful of advertising partners. Some worked with 1 or 2 advertising partners while others maybe worked with 15 to 20. As clients embraced the new-found mobile attribution technology – like MobileAppTracking – that allowed them to more efficiently and effectively scale their advertising campaigns, our clients were enabled to and some quickly did begin working with over 300 different advertising partners simultaneously. Alongside this growth, over the last two years the number and diversity of ad networks, DSP, exchanges, etc. has skyrocketed as well. In such a complex mobile advertising world, it’s imperative that we continue to drive our industry forward with both innovation and transparency.

The beginning of multi-touch attribution in mobile advertising is here!

Defining The Install

In the world of desktop web, you’re probably already familiar with Google Analytics launching multi-touch funnel analysis in 2013. With this launch, Google Analytics moved beyond the conversion – a single event – into multi-touch attribution by defining three types of interactions:

  • Last Interaction – the referral that immediately precedes the conversion.

  • Assist Interaction – any referral that is on the conversion path, but is not the last interaction.

  • First Interaction – the first referral on the conversion path; it’s a kind of assist interaction.

Well, that’s great for desktop web advertising, but what about the realm of mobile? We’ve taken our experience, our clients’ experience and our industry knowledge to move beyond attributed installs by providing insight on multi-touch attribution. All of this experience and hard-won knowledge has led us here at MobileAppTracking to categorize installs as:

  • Attributed Installs – Installs where the publisher has the last click and that last click happened within publisher’s attribution window.

  • Organic Installs – Installs that are not a result of any publisher campaign (i.e. happen organically) and therefore are not attributed to any publisher.

  • Installs = Attributed Installs + Organic Installs

Our Attributed Installs measure the number of last click conversions while Organic Installs are conversions that we don’t know who’s responsible for helping a new user install a mobile app.

Identifying Additional Install Interactions

With mobile advertising becoming increasingly complex, there are more and more potential scenarios where more than one advertising partner interacted with a user prior to them installing your mobile app for the first time. Google defines such interactions as an “assist”, and we measure assists for installs in a similar manner.

  • Install Assists – Publisher(s) with click prior to last click, within publisher’s attribution window.

If an advertising partner is part of a conversion path for an install of a mobile app but not responsible for the last click, we measure these as “Install Assists” – i.e. an assist to the install. Since there can be multiple advertising partners involved in the conversion path, double-counting can occur. Double counting would occur when a single user sees ads from three different advertising partners. We credit the advertising partner with the last click with the Attributed Install. The two other advertising partners both would be credited each with an Install Assist. So there would be one Attributed Install and two Install Assists for the single mobile app install.

We think of the publisher that was responsible for the last click and the other publishers that assisted with the install as contributing to the install – both types help drive mobile app installs. Thus, we can think of Attributed Installs plus Install Assists as Publisher Install Contributions.

Publisher Install Contributions = Attributed Installs + Install Assists

Publisher Install Contributions provide a more holistic metric on a publisher’s impact in generating mobile app installs. You the advertiser can communicate with your advertising partners (publishers) on their total impact using this metric since it’s a metric you’d both have access to.

Screen Shot 2014-01-02 at 2.20.18 PM

In the example above, Ad Partners 1 and 2 both are responsible for most of their install contributions as providing the last click. However, Ad Partner 3 has four times more Installs Assists as Attributed Installs. Thus, Ad Partner 3 actually interacted with more new users that installed the mobile app than any other advertising partner.

Note that there were only 400 mobile app installs by new users in the above. There were an additional 460 Install Assists.  Don’t make the mistake of thinking that Publisher Install Contributions is the total number of mobile app installs by new users since it’s not. Also, if we only allowed one assist per attributed install, then Install Assists would never be greater than Attributed Installs. However, we credit as many assists per Attributed Install as advertising partners involved. Thus, Install Assists can be many times greater than Attributed Installs.

The Importance of Analyzing Additional Install Interactions

Analyzing Install Assists can provide insight on:

  • Uniqueness of advertising partner’s traffic sources

  • Whether an advertising partner is missing out on credit for installs due to last-click attribution

  • Duplication / overlap between advertising partners

  • Focus of advertising partners’ campaigns on awareness vs. direct response

  • Full impact of advertising partners’ campaigns

The level of Install Assists provides insight on how unique the advertising partner’s traffic source is. If the advertising partner connects with a completely unique audience of users, then they should have no assists for installs as they would be no other advertising partner showing the same user an ad. If the advertising partner has a high level of Install Assists, then they don’t have unique traffic sources as the users they connect with also interacted with your other advertising partners’ ads.

Prevent Crediting Duplicate Installs

Without implementing a solution like MobileAppTracking, all of your advertising partners would take responsibility for an install regardless if another one of your advertising partners actually had the last click. This is because they have no way of knowing about the campaigns from your other advertising partners. At the end of the day, you want to pay the appropriate amount per install and can’t afford to pay your entire CPI multiple times for only one unique new user.

Once implementing an attribution technology like MobileAppTracking, an advertising partner’s install rate (conversion rate) can decrease since we’ll only give one advertising partner credit for the install based on the last click. The Publisher Install Contributions would be the actual count of installs the advertising partner would take credit for if you didn’t have our attribution technology implemented.

If an advertising partner has a high number of assists, this indicates they have a high level of overlap with your other advertising partners. You may be used to hearing this referred to as “duplicate installs” – where more than one advertising partner would take credit for a single install if a technology like ours was not implemented. If you’re paying on a performance basis, thousands of duplicate installs can cause seemingly ROI positive advertising campaigns to actually be extremely negative when analyzing them together.

By only attributing an install to one advertising partner and giving assists to all other advertising partners involved in the conversion path, the issue of duplicate installs is eliminated while still providing insight into all the new users they interacted with that in the end installed the mobile app.

Evaluate Different Campaign Goals: Awareness vs. Direct Response

The ratio of Attributed Installs compared to Install Assists indicates whether the advertising partners’ campaigns focuses on awareness vs. direct response. Advertising partners who focus on awareness via banner, video ads, etc. can be highly effective in encouraging users to install and engage with your app at a later point in time. When an advertising partner is great at awareness – first touch interactions – they will typically have a low ratio of Attributed Installs to Install Assists.

An advertising partner great at direct response would look like this:

100 Attributed Installs / 10 Install Assists = 1000%

An advertising partner great at awareness would look like this since other advertising partners better at direct response would be getting attributed with the install by having the last click:

100 Attributed Installs / 400 Install Assists = 25%

Therefore, if you have advertising partners who have varying purposes – drive awareness vs. drive installs – comparing Attributed Installs to Install Assists can give you insight into how well each is achieving their goal.

How the Attribution Window Affects Install Rates and Interactions

While measuring Attributed Installs and Install Assists is certainly awesome on its own, it’s limited to the advertising partner’s attribution window. The attribution window is the maximum time between their click and the install that they will take credit for the install.

If the advertising partner’s attribution window is seven days, and they interacted with a new user with an ad five days before the install, then the click is within their attribution window and will be used to attribute the install.

If the advertising partner’s attribution window is seven days, and they interacted with a new user with an ad ten days before the install, then the click is outside their attribution window and would not have been used to attribute the install.

What complicates this even more is that the attribution window between two advertising partners is likely not the same. You may agree to be billed on installs occurring within 3 days of clicks, others 7 days, 14 days, 21 days, 28 days, etc. So how can you actually compare one advertising partner’s performance to another accurately? Apples to apples…

Non-Windowed Install Contributions: Why You Should Care

With the release of our multi-touch attribution, we now measure installs and assists happening outside the advertising partner’s attribution window as:

  • Non-windowed Install Contributions – Publisher(s) with clicks within 30-days but outside of publisher attribution window.

By measuring these potentially attributed installs and assists as non-windowed contributions, you can accurately compare apples to apples using our maximum attribution length of 30 days. Think about Non-windowed Install Contributions as additional attributed installs or assists if their attribution window was longer. We reference this metric as “contributions” since they could have had an attributed install or assist so they helped with these installs. They are contributions since they did help, but we don’t share this data with them and you don’t owe them for this impact since it’s outside their attribution window.

Non-windowed Install Contributions can be used to analyze all potential installs where advertising partner was involved within 30 days. This can be done by adding up all our new metrics to calculate the new value we like to call 30 Day Install Contributions.

30 Day Install Contributions = Attributed Installs + Install Assists + Non-windowed Install Contributions

Let’s compare Drawbridge who has an attribution window of 14 days to Facebook who has an attribution window of 28 days. If we just compared Attributed Installs or even included Install Assists, it would seem like Facebook was part of more conversion paths for new users installing a mobile app. By using analyzing them using our max attribution window of 30 days, Drawbridge would actually have impacted more new users to install the app than Facebook.

Screen Shot 2014-01-02 at 3.42.28 PM

As the above scenario makes clear, without Non-windowed Install Contributions you cannot directly compare an advertising partner’s entire performance against another accurately.

You’re probably already thinking you may want to compare two advertising partners performance with shorter windows (i.e. length of time between click and install) instead of our max length of 30 days, and, not to worry, this is something we’re certainly thinking about already. Instead of comparing Facebook vs. Drawbridge at 30 days, why not at 1 day or 7 days?

By using the advertising partner’s standard attribution window or something custom you’ve agreed to as part of your insertion order, your costs can be minimized legally instead of attributing everything using a max length like 30 days. Measuring the non-windowed interactions gives you insight you need to then compare them.

While our alpha partner interface does provide publishers with access to Attributed Installs and Install Assists, they are not provided access to Non-windowed Install Contributions since these are outside of their attribution window. Non-windowed Install Contributions are for you the advertiser to be able to compare one publisher to another accurately.

We hope that this first release of our multi-touch attribution will empower you to build more epic advertising campaigns than ever before and make 2014 the year that mobile app advertising and attribution takes the world by storm.

Becky Doles

Becky is the Senior Content Marketing Manager at TUNE. Before TUNE, she led a variety of marketing and communications projects at San Francisco startups. Becky received her bachelor's degree in English from Wake Forest University. After living nearly a decade in San Francisco and Seattle, she has returned to her home of Charleston, SC, where you can find her enjoying the sun and salt water with her family.

5 responses to “The Beginning of Multi-Touch Attribution in Mobile Advertising is Here!”

  1. Ambrish Bajaj says:

    Hi Lucas, thanks for this interesting article and useful feature. How do you think advertiser should compensate publisher or ad network for install assists in addition to CPI paid out for attributed installs? Also how do I evaluate whether an install assist is good quality or aided in an install later.

    • Lucas says:

      Hi Ambrish. Right now is a learning period for the advertiser and our industry on how valuable assists are. There is no standard on this. I don’t believe we should try to put a $ on an Install Assist yet.

      The same question on how we can evaluate whether an Install Assist is good quality can be asked of the Attributed Install. Did the publisher with the Install Assist actually do all the work while the last-click publisher just swooped in and got attributed with the install? These are certainly questions we’re planning on solving as we make attribution more transparent.

  2. Kate Thompson says:

    Lucas, how does last click attribution work if the user clicks AND downloads the app, but does not open it? Case in point, user clicks on an ad game from Pub1 and then downloads but does not open. User then clicks on same ad game from Pub2. App store will show the “Open” button instead of “Install” button since it’s already been downloaded. After the user opens, who would get the attributed install in this case? Thanks in advance.

  3. So Young says:

    Hi Lucas, thank you for this useful article! by the way, i can’t find some images in this articles which makes me wonder what it is. Please check the images. =)

Leave a Reply