Unlock Personalization Use-cases With Detailed Device Tracking
Discover how detailed device recognition can significantly enhance the accuracy and scope of the device & browser data collected in your session_start for enhanced analytics and personalization
About the Author:
Co-founder of Datacop, agency that fulfils marketing operation roles in large eCommerce companies such as OluKai, Melin, Roark, Visual Comfort and Company, Dedoles and others.
The potential ways customers browse online - changes and develops. When the internet was first “born”, everyone accessed websites via a Desktop. Today, web-enabled screens come in practically every shape and size and are profilerating into cars, onto our wrists and (into our glasses?). As these changes develop, it is important for data analytics team to be able to track this information with much more detail.
The problem is, that most session_start events typically only track high level device categories like “Phone / Tablet / Desktop”. At Datacop, we have found a solution to this problem.
In this article we will go over the additional attributes that are possible to add into your session_start event thanks to “Detailed Device Tracking”. We will discuss the results of our internal device testing to see how accurate it is and how this data can be used in practice by e-commerce companies, particularly electronics brands.
“User-agent” is difficult to interpet
Since the early days of the internet, browsers would add the “user-agent” into their HTTP requests. The user-agent is a string of data passed on in the request that enables the server to know which browser/device is requesting the website and so the server knows which version of the site to serve to the browser/device.
However, there was never a process of standardisation in the industry regarding what should be passed on in the user-agent by browser/device combinations. As a result - device manufacturers (Apple, Samsung, etc.) and browser developers (Google, Mozilla, etc.) would “identify” their devices and browsers each in their own way. Furthermore, competitive pressures among browser developers has lead to at times browsers ‘impersonating’ other browsers in their user-agent to avoid providing their users a poorer browsing experience. Over time the user-agent has become difficult to interpret and it is not easy to “fingerprint” the type of device, brand, make of the device.
Therefore out of box session / page_visit event tracking by CDPs like Bloomreach Engagement, Segment, mParticle suffer from 2 limitations when tracking browser and device attributes of visits. Out of box they are only able to track high-level and sometimes inaccurate information about the devices and browsers.
Limitation 1: device attribute values in Bloomreach out of box are limited to a handful high-level values such as: “iPhone”, “Android”, “Other (Desktop/Laptop)”, “iPad” or “Blackberry”
Limitation 2: less prominent browsers like Edge (Microsoft), Yandex are not recognised by out of box tracking and are instead mis-identified as one of the major browser players like Google Chrome or Safari.
To go beyond these out of box limitations, we have found a solution for this at Datacop. We recommend implementing a detailed device detection repository in your Bloomreach Engagement / mParticle / Segment SDK.
Detailed Devices Detection
We recommend to track the following 5x additional event paramaters to your key front-end events such as session_start / page_visit / view_item / cart_update / etc. :
advertised_browser: returns a more accurate browser value
complete_device_name: returns the brand name and model of the device such as “Apple iPhone 14 Pro Max / Samsung Galaxy S23 Ultra”
form_factor: returns the device category of - such as “Desktop / Tablet / Smartphone / Smart-TV / Robot / Car Screen / Other non-Mobile”. Compared to the standard out of box device recognition, this recognizes new device types.
is_app_view: returns true / false, depending on if the page was visted via a Mobile App Browser, such as if you open a page via the Instagram / Facebook App.
is_robot: returns true / false, depending of if the session was generated by a bot such as an online crawler or other programmable agent
Testing & Results
At Datacop we did thorough testing of the detailed device detection repository to evaluate how it performs in practice.
According to Statcounter, in October 2024 Apple devices represented approx. 28% of mobile sessions. As you can see on the screenshot below, detailed fingerprinting is capable of identifying the specific makeof the iPhone from the 46 potential Apple iPhone models, or iPad from the 38 potential Apple iPad models.
However, we have noted in our testing that if the user has an adblocker installed on their iPhone, the detailed device detection will not identify the model of the iPhone, just that it is an iPhone.
According to Statcounter, in October 2024 Android brands such as Samsung, Xiaomi represented approx. 22% and 10.5% of mobile sessions respectively. As you can see on the screenshot below, with detailed fingerpriting it is possible to identify the specific popular Android devices by brands like Samsung S23 Ultra, Xiaomi Redmi Note 11 Pro and LG G6.
However, we have noted in our testing that not all Android devices are recognisable even with the detailed device detection. Some Android devices are simply identified as “Generic Android 6.0”. Since there are more than 24,000 different Android devices it is reasonable to have such “catch-all” values for the long-tail of the potential specific Android device brands and models.
Our testing revelaed that the detailed device fingerpriting is able to more precisely identify browsers. As mentioned in the introduction, browsers sometimes have incentives to “pretend” to be a different browser. As a result, compared to the standard browser recognition there are browsers such as Edge, Yandex, Brave the detailed device recognition is able to recognize, but are mis-identified in the standard device recognition. Note such instances on in the screenshot below.
If you click open a webpage on a mobile device within the interface of a mobile application such as Instagram or Facebook; you may have noticed that it can open the session within an “App Browser”. With detailed device recognition it is possible to distinguish such an App Browser visits. Standard device tracking does not distinguish these sessions from others.
Can you imagine the possibilities for your personalisation and analytics capabilities if you could track the above mentioned information reliably into your Bloomreach Engagement?
Use-cases
So our testing has revelead that the detailed device tracking can enable your business to track more accurate and detailed device and browser details especially regarding mobile devices. So how can this information be used in practice?
1. Personalizing based on Device Model
Detailed device detection enables businesses to craft personalized shopping experiences and marketing campaigns based on the user’s specific device. These use-caes are particularly relevant to eletronics e-commerce brands that sell or service mobile devices and e-shops that offer accessories for devices.
Device-Personalized Shopping Experiences: Knowing the exact device model allows e-commerce companies to recommend products uniquely suited to the user's needs. For instance, users browsing with an "Apple iPhone 14" can be shown recommended accessories specifically for that model. On category pages, the e-shop may pre-filter on the user’s device model or prompt them to do so. Additionally, certain other devices such as smart watches, laptops and earphones that are compatible or recommended to use with “Apple iPhone 14”.
This kind of personalisation can reduce a lot of friction for users and can potentially significantly improve conversion rates to add to cart and conversion rates to purchase on your web.
The personalisation of the user experience using the detailed device can go beyond the web and can be used in retention campaigns. Consider the following examples:
Automated cross-sell campaign for user’s device: send a follow up email / sms to user after the session that highlights best selling servicess & accessories for that device, such as device covers, device glasses, insurance, etc. Automated up-sell campaign for user’s device: send a follow up email / sms to user after the session that highlights accompanying compatible products with the user’s device such as laptops, smartwatches, earphones, etc.
Device Upgrade Campaigns: This is specifically relevant for e-shops that sell mobile devices. Identify users with older devices and target them with upgrade incentives. For example, targetted banners on the web could highlight iPhone trade-in programs for users browsing on an "iPhone 11," showcasing the trade-in value and promoting the latest models. Additionally, automatically generated comparison tables between the user’s current device and that brand’s latest model can be compared to see what features and capabilities the user is currently missing out on.
2. Upgraded Analytics on Device / Browser
Enhanced detection provides a more granular view of how websites perform across different devices and browsers. Essentially in any analysis where the device or browser attribute of the session was concerned is upgraded:
Conversion Rate Insights: Analyze whether conversion rates vary significantly by detailed device or browser. Is there a device model that is experiencing significantly below or above average conversion rates in their funnel to purchase? If users on "Microsoft Edge" have a lower purchase conversion rate than users on "Google Chrome," it may indicate a compatibility issue or a suboptimal user experience.
Consider comparing the CR% to purchase, Average Order Value, Average Unit Price and Customer Lifetime Value of customers with different device models. Do users with more expensive mobile devices have a higher AOV and a higher CLV? If so, it would make sense to experiment with providing more premium opportunities for users with more expensive devices and perhaps more value-deals for users with less expensive devices.
Anomaly Interpretation: Whenever there is a positive or negative anomaly (a sudden drop in traffic for example, or a surge in sales), among other things - it is always worth examining the change split up by devices to better understand what happened. Having more granular device data as well as more accurate browser readings would greatly help with interpreting changes in the e-shop.
Customer Understanding: For e-shops that sell mobile devices, electronic devices or products related to mobile devices - data on the specific models and brands used among your customers can be a treasure chest of insights that can be useful in many analyses.
3. Cleaner Data & Metrics
Detailed device detection also helps improve the quality of data used in analytics and decision-making by accurately filtering out irrelevant or misleading sessions. Since it is capable of identifying robot sessions more accurately, it provides another layer of “defense” in maintaining data quality.
Unidentified robot sessions and crawlers can skew key metrics like traffic, conversion rates, and funnels. By identifying and excluding robot-generated sessions, managers at e-commerce businesses can ensure cleaner and more actionable data.
Implementing Detailed Device Recognition in Your Bloomreach Project
If you’d like to find out how to implemented detailed device recognition in your project and discuss how to drive business results with this feature book a free 30 minute meeting with us at the button below:
If you found this article useful
f you enjoyed it, don’t miss out on future articles—subscribe for free and have our latest content delivered straight to your inbox.