For Advertisers- Mobile Attribution- A quick guide - By Sarang Kinjavdekar

If you are an Advertiser currently running App Marketing campaigns and finding it difficult to measure or if you are looking out for some more details on Mobile Attribution, some general issues that you might face and some questions that you need to ask to your Current/Potential Attribution Partner and some to yourself as well, then keep reading. I have not detailed out the different models, I have written a post earlier about it and if you want to know how each model works, click here to read that post.

Note- Before we begin, this is not a guide on how to choose the best Mobile attribution partner, I personally have used Adjust from last 6 years and I am very satisfied with it. However, this is not a post that talsk about any specific partner, but it is more on educating about the Mobile app attribution in general. Below content is based on my experience, I hope this will help you to define your attribution strategy

Defining Attribution

Attribution- Specifically talking about the Mobile, attribution refers to measurement of user events which  are the results of marketing activity. Or it can be defined as A method of assigning credit to a particular marketing interaction, brand touch point or channel in order to quantify the contribution   that are made by them towards a desired business goal (like Turnover, Profit, Customer retention, sales volume, revenue etc)

Examples- App Installs, product purchased(IAP), Level completion(in gaming), app launch etc

Traditional online advertising attribution techniques- Uses  (1) Cookies (2)Pixels(or tags) (3) Appended URL

Mobile attribution techniques/Universal Mobile App Attribution- (1) Google Install referrer (2) Identifier matching (3) Fingerprint Matching (4)Open  URL with click id (many more are used like location tracking etc but these 4 are the major ones used by mobile tracking companies like Adjust, Appsflyer, MAT, Apsalar etc)

Mobile attribution is very complex if compared to the traditional attribution. For this presentation,  we will focus on Mobile attribution. First we will go through why attribution is needed through a sample conversion in the coming paragraphs/pictures (Major Players: Adjust, Apsalar, Appsflyer, Kochava, MAT/Tune etc)

There are several methods for attribution but it depends on 4 factors- Not all these can be tracked together through a single method 

1.Platform/App store- iOS/Apple iTunes  App Store, Android/Google Play Store, Amazon Appstore for Android, etc.

2.Engagement type- Click through vs View Through

3.Conversion type- Installs ve events

4. Flow- Web to app or App to App

Attribution benefits

1.Accurately measure mobile campaign performance

2.Accurately measure revenue and ROI

3.Improve personalization efforts as you get the holistic view of the customer journey and touchpoints

4.Make powerful optimizations

5.Compare multiple networks and see which is performing and which one does not

6.Manage budget efficiently and reduce media wastage

7.Correctly attribute credit to the source and manage payouts without efforts

Overall App Journey

The overall app marketing process can be split into 5 major areas.

1.New user acquisitions- Attract new users + Download the app

2.Installation- This is the stage where user opens it for the first time

3.Engagement- User opens the app, engage with it. Can be organic (self driven) or promoted (ads)

4.Remarketing- Attract existing users, engage and drive more traction

5.Re-engagement- Completion of an event in the app- deep links usage

Below is a picture that covers the above in easy language (click to enlarge)

Click Based Attribution

Install attribution is based on pre-determined timeframes which is also known as lookback windows. Below are some of the windows used in attribution

1.7- day standard- Network X serve the ads to user A. User A clicks on it within 7 day window, it will get the credit for that install (assuming that no other ad network is running ads simultaneously)

2.24 hour device fingerprinting attribution- Lets say if the device id or referrer is not available, the tracking is done using the device fingerprinting method. It can only record accurate data in very short window and hence the window is 24 hours

3.Facebook, Twitter and Google- Facebook- Fixed 28 days, Google- Fixed 30 days (non configurable window) while Twitter offers configurable windows of 1/7/14/30/90 days

4.Configurable window- Gives networks and advertisers great flexibility- networks need more window while advertisers can do an apple to apple comparison when running different ad networks under the same window. Its best used when we are running time bound campaigns

View through Attribution

Some agreements allows giving attribution credits even when ad is viewed and not clicked. Generally click is more powerful and it wins always. FB and Twitter charges on 24 hour window.

Click Vs. Impression attributions?Winner?

Let's say , we are running campaigns with 2 different Ad networks. Network A has the click through window of 7 days and view through window of 1 day while Network B has only click through window of 15 days, no view through.

Here network B wins because it has the last click within the lookback window and also click wins over view even if it is within the window

General rules of attribution and priority

1.Clicks takes precedence over impressions

2.Deterministic attribution 1:1 (unique identifiers, device IDs) takes precedence over non-deterministic non 1:1 (fingerprinting)

3.Clicks measured with a click date time take precedence over clicks measured without a click date time

4.When no deterministic identifiers are found then at last device fingerprinting method is used

Multi Touch Attribution- Assisted Installs and non Window conversions

In the last paragraph, we saw how we can credit an install to a network by using different methods. But that is not enough, we also need to understand how different ad networks or partners influenced the user before he finally downloaded the app. It also gives us an idea of the partner performance at individual level as well as to determine their reach

Below are some of the important terms:

Download- When user go to the app store and downloads the app into his device

Install- When the user opens the app for the first time

Attributed installs- Partner with last click and within the respective attribution window

Assisted installs- Prior to last click but within the attribution window

Non-window conversions- Partners with click within 30 days but outside of the attribution window

What are some common issues that we face normally in the Mobile attribution scenario?

1.       For mobile attribution- no model or technique is universally accepted yet (unlike the traditional web attribution which is perfect)

2.       Not all models can measure everything single handedly like Platform (iOS/Android etc.), Engagement type(Click vs view), Conversion type(Install vs events) and Flow(app to app vs web to app)

3.       None of the model is 100% accurate

4.       Every Mobile attribution partner has multiple methods of attribution so selection of vendors becomes very difficult- they use either deterministic or   probabilistic identifiers- so by using any one, so it  not possible to get the clear picture. Even if both are used still its not 100% accurate

5.       There are multiple devices that people use to engage with brands, and without a universal method to track campaigns across all of these devices, the efficacy of mobile campaigns can be under-reported or devalued.

6.       Users move from online to offline, so getting the credit for mobile in discovery phase becomes very difficult

7.       Cross device attribution model is still not universally developed and accepted

8.       Industry lacks a scalable device agnostic model that can tell how the upper funnel mobile touch points attributed to the lower funnel sales and revenues

9.       A typical user journey is becoming far complex and fragmented

10.   Cross device and cross channel campaigns possess a serious challenge for attribution

11.   The attribution is still based on last click

12.   Poor quality of identifiers leading to inaccurate attributions

13.   The technology is constantly evolving and there are new attribution providers constantly coming up

14.   Data science talent needed- to apply increasingly sophisticated models.

15.   The attribution problem actually falls into 3 buckets (a) Cross channel attribution (Search, social, email etc.)(b) Cross platform (Phone, desktop etc.) and Offline attribution (footfalls, phone calls etc.)

16.   App Marketing  technology and tracking needs skilled resources and time. For instance- deeplinking needs skilled labour

 17. Tracking and then attributing results across channels in a cookie-less platform has been a challenge.

18.   There are 2 environments- Mobile web and Mobile app- each works in silos- lacks common tracking and varying degrees of functionality

Some Questions you can ask your Potential/Current Attribution partner and some Questions you might ask yourself as well

1.       You cannot run the campaign on Mobile web if you want deterministic approach and looking for device ids. It is only available for in-app inventory and also the ad network should agree to provide it. Not every Ad networks agree or have the capability to provide it

2.       Before selecting the attribution partner, make sure that they support both- mobile web and mobile app

3.       Also check that the attribution partners can pass unique variables (like order id etc.) through conversion tag so you can validate the purchase

4.       Be very careful while selecting the parameters like de-duplication and reattribution window. This should be in-line with the overall app journey and time taken for overall conversion

5.       Very short reattribution window is not advisable; it should ideally be 7 days

6.       Be careful while you look at the attribution for Facebook (28 days) and Google (30 days), they might not be in-line with your attribution default window. Facebook even charges for impressions based conversions that happens within 24 hours

7.       Multiple vendors generally claim for the same app events or install events if the parameters are not set properly

8.       Share the revenue tokens and purchase tokens details with all the vendors so that they have the clarity on the revenue that they are generating. It’s one unique token for each and created one time. This will give your vendors the total purchase and total revenue generated and helps them in optimizing the campaign

9.       Always have the post-back passed back to the networks, this will help in optimization. For affiliates, always get publisher id, ask them to append URL

10.   Use S2S and callbacks to fetch any data into the CRM and Business Intelligence systems

11.   Make sure that all the vendors work on same attribution window. If one network has a look back period of 24 hours and another has a lookback period of 7 days, it’s likely that the latter network is taking credit for more installs

12.   Leadership should also be involved and should share the same vision, if not it will not be successful

13.   Frauds clicks from devices, channels and platforms causes serious issues in the overall attribution

14.   Work with a vendor who offers  purchase verification, fraud prevention, feature like cohort analysis, customizable attribution modeling, uninstall analysis, deeplinking, re-targeting, re-engagement, audience segmentation, de-duplication, tracking purchase in different currency  etc. to name a few

15.   You don’t have to create separate trackers for each country when you work with any ad networks. Country data can be obtained by just searching the country name. Just have a single unique tracker for each network/partner 

16.   Keep the view through attribution to  zero or in case if you still want to track it, it should not be beyond 24 hours – but again if you are doing purely Branding campaigns , then you may have to re-look into this and make it longer- again depends on your business and customer /buying life-cyclye and purchase habits.

17.   Make use of all the segments and track each network upto 4 levels-Network, campaign, Ad-group and creative (I used Adjust before)