gears

Device Detection Selection Criteria

Mike

8/28/2014 3:13 PM

Device Detection Analytics News Opinion

Ensure your customer's device doesn't get in the way of your brand experience

Overview

Integrating Device Detection into your web site is more urgent than ever. 30% of web traffic in the United States now originates from mobile devices. Some sources suggest almost half of all paid searches are now carried out on mobile devices. No organisation can afford to offer a negative mobile or tablet experience.

Responsive Web Design (RWD) complemented with server side Device Detection techniques, often termed Responsive Server Side (RESS) or Adaptive Design, provides a fast, high performance solution to many of the challenges associated with supporting multi-screen. Device Detection also significantly enhances the insight available from tools which support customer engagement and business intelligence.

Some Device Detection suppliers make bold claims about the market-leading nature of their products. In reality many suppliers making such claims don’t in fact provide the highest accuracy, fastest performance, most extensive device coverage or easiest setup, often charging a high price for unnecessary professional services and support.

Yet if you choose wisely, you can have a device detection solution that not only integrates seamlessly and easily into your current platforms, but is future-proofed to grow with your business and web needs, that enables you to gain insights into your customers’ behaviours, and make it easier for you to deliver great user experiences for your customers on any device.

What Should I Look For?

What businesses need from their device detection solution?

  • Fast and Accurate Data - ensuring the right result every time
  • Device, OS and Browser Combination Intelligence - always supporting the latest devices
  • Enhanced Analytics Tools - driving bottom line revenue improvement
  • Easy to use Enterprise-Class Software - reducing operational complexity, risk and cost

In short a data insight engine that detects, responds, analyses and ultimately improves your users’ experiences, today and for as long as you need it.

We’ve put together a 13 point check List in this guide to help you make the right choice for your business. If you’d like to find out more from us at 51Degrees, we’re happy to help.

Fast And Accurate Data

There are three parts that make up fast and accurate data: Quality Data Sources, Data Accuracy, and Response Speeds.

1. Data Sources

The quality and authenticity of data is key. Good solutions use authentic data sources by combining real handset data, vendor information and analysis of real web activity. There are suppliers who source their device information from community web sites like GSMArena or PhoneArena, these sites will never cover all the new devices available when you consider that 250 to 350 new ones are released each week, many from far eastern manufacturers.

The best suppliers will have both people and analytical tools dedicated to monitoring real web activity and identifying emerging devices, as well as direct relationships with manufacturers. They’ll make their device data available on their web sites for anyone to quickly browse and validate. The device detection solution also needs to know when a device isn’t a device! A lot of web traffic comes from search engines, applications, and crawlers and a good device detection solution needs to accurately identify these as well.

2. Data Accuracy

Your supplier ought to be able to explain how the accuracy of their solution is measured and maintained. The best measure is where a known sample set of device finger prints (known as User-Agent strings) are compared to the results of the detection process. The sample needs to be large enough to be representative of the traffic on your web site. For example, 51Degrees has classified over 9 million device finger prints or User-Agents. Ultimately there’s no substitute for your own tests and any solution should be tested and evaluated against your existing traffic prior to deployment. The solution should also provide you an indicator as to the accuracy or certainty of the detection result and explain what this means for your decision making.

Data maintenance should be a consideration as part of data accuracy, check your suppliers’ processes for keeping up to date with the rapid growth in device combinations - both up-to- date data and fast accurate algorithms are important.

3. Response Speeds

The whole point of device detection is to improve your users’ experience. Therefore the process of device detection needs to be so fast, (sub millisecond - around 0.10 ms) that it has a virtually unmeasurable impact on your web site response time.

Spending Time In The Clouds

Device detection based on cloud computing is typically much slower than solutions that you host adjacent to your Web site. This diagram outlines the additional steps needed when using a cloud-based service.

220 millisecond delay in server response due to cloud service

A respectable device detection solution should be sub millisecond and cloud services do not provide this level of performance. Figure 1 shows the additional delay introduced by a web site making a request over the Internet to an external cloud service. Cloud services should be avoided for any web site where speed of response is critical.

Where cloud services can’t be avoided a solution which provides multiple geographic locations and automatically routes detection requests to the nearest location will reduce some of the delay.

Figure 2, shows a device detection solution that does not use a cloud based service has better performance because device detection is performed within the web site environment.

Sub millisecond server response with on-premise deployment of device detector

As well as performance issues, cloud-based solutions may also create data privacy and availability issues. By buying device detection services that run through a cloud, you may have to accept whatever privacy policy the cloud provider sets along with availability and performance SLA’s which reduce your overall service levels.

The location of the primary and backup data storage must be known to comply with EU data protection laws, which differ from US data protection laws.

Free cloud based device detection solutions may appear attractive financially but compromise user experience.

Checklist: Fast and Accurate Data
  1. Data Source: Do you use authentic data sources e.g. device information from manufacturers?
  2. Data Accuracy: How do you measure and maintain data accuracy?
  3. Response Speeds: How fast are your response speeds without caching, (e.g. less than 0.10ms)?
  4. Cloud: If you’re considering cloud device detection do you understand the performance privacy and availability implications?

Device, OS And Browser Combination Intelligence

The next area to consider is the number of device combinations - physical hardware, operating system and browser permutations (not just physical hardware). This is important in making sure your communications are tailored to all aspects of the individual users’ technology and not just one component.

Scope and scale is what you need from your supplier. 51Degrees has more than 150,000 combinations available, and this number is growing! 51Degrees expects device combinations to increase to at least 500,000 over the next 3 years. It’s vital to ensure that your supplier’s algorithm’s performance does not degrade as device combinations increase.

51Degrees performance and device combinations graph

Older algorithms, created when operating system, browser and hardware were intrinsically linked in the last decade, may struggle to deliver consistent performance and accuracy as device combinations grow, leaving you with the unexpected decision to replace a solution that’s slow and unresponsive.

Unclassified Devices

Some device, operating system and browser vendors do not always publicise their products to partners in advance of launch. In such situations even the best suppliers will need a few days after launch to classify the device and complete the necessary tests. A modern detection algorithm needs to handle these scenarios by finding the most appropriate previous version and indicating the accuracy of the result provided. For example; the previous device to the Samsung Galaxy 5 is the Samsung Galaxy 4. Until the Galaxy 5 has been classified the detection algorithm should return the Galaxy 4 and the correct operating system and browser information along with an indicator to show a previous version was returned. Importantly it should not return random results or default generic information. Accurate detection of prior versions and devices is particularly important to organisations that do not update their device data on a daily basis.

Checklist: Device Data
  1. Device Combinations: How many devices, operating systems and browsers do you cover today, including spiders, web crawlers and robots?
  2. Data Maintenance: How do you maintain your device data?
  3. Future Proofing: How do you manage the rapid rise in device combinations, and keep performance and accuracy standards high?
  4. Unclassified Device Handling: How are new, as yet unclassified devices, operating systems and browsers detected? Does the solution provide an accuracy indicator when a precise result is not available?

Enhanced Existing Analytics

Why introduce a separate analytics platform when your business already uses leading tools like Google Analytics, Omniture, Coremetrics, Webtrends or others?

51Degrees device detection adds an extra dimension to existing analytics systems

You need to ensure your device detection solution easily integrates with your existing tools to provide a single view of the customer and does not lose key data during the aggregation process.

Physical screen size is one of the most important factors to consider when optimising a web site for multi-screen. But surprisingly, many web site analysis solutions lose this information when they aggregate data by mobile, tablet or desktop. This is becoming a more significant factor as the definitions of various device types blur.

What’s needed is enhanced analytics that can provide a more rounded three dimensional view to complement your existing analytics data.

51Degrees Enhanced Analytics solutions provides this missing information and other key insights into the device such as its retail price, manufacturer and how quickly your web pages display on your users device.

Checklist: Enhanced Analytics Tools
  1. Integration: How do you integrate with your existing analytics systems?
  2. Insights: Can your service offer device insights like retail price, device age and page response times?

Easy To Use Enterprise-Class Software

There are a number of hygiene factors you and your IT department would expect from any enterprise-class software supplier. Here are some we think are important:

  • Proven software - ideally deployed by major brands, in high volume environments and used by millions of web sites.
  • Open source licensing permitting commercial use.
  • Maintained by a professional organisation with support agreements and easy access to core developers
  • Complements both leading Analytics platforms and Responsive Design techniques
  • Easy Integration and implementation

Total Cost Of Ownership

Make sure you understand the total cost of ownership. Some providers may charge for expensive professional services to integrate and maintain their solution alongside your existing systems.

  • Integration with existing systems - Good Integration with Content Management Systems (CMS), web platforms and key software programs is essential. Many CMS and web platforms include device detection out of the box, or there’s easy to follow guides and plug-ins to add device detection. If the solution you’re considering doesn’t have this, it may be time to think about your supplier again. For example, all Microsoft Web Tech solutions include hooks for 51Degrees Device Detection.
  • Standard open source licencing - As well as your IT department, your Legal department are likely to have a few expectations too. Consider the legal implications of using open source projects with complex commercial use clauses, or proprietary software where you don’t have access to the source code. What are the escrow arrangements? Look for open source which allows commercial use.
  • Flexible pricing - It’s also worth saying that quality Enterprise Class software shouldn’t mean major Enterprise pricing or ongoing support costs. Look for flexibility in pricing models that suit your business needs.
  • Always try before you buy - Many vendors will offer a FREE trial period so you and your web/IT teams can test the software for yourselves.
Checklist: Easy to Use Enterprise-Class Software
  1. Proven: Is your software proven, used by millions, including the world’s leading brands?
  2. Integration: Will your software integrate with our existing systems?
  3. Free Trial: Can I get a free trial using my data - Can I install the software for free?

Full checklist

  1. Data Source: Do you use authentic data sources e.g. device information from manufacturers?
  2. Data Accuracy: How do you measure and maintain data accuracy?
  3. Response Speeds: How fast are your response speeds without caching, (e.g. less than 0.10ms)?
  4. Cloud: If you’re considering cloud device detection do you understand the performance privacy and availability implications?
  5. Device Combinations: How many devices, operating systems and browsers do you cover today, including spiders, web crawlers and robots?
  6. Data Maintenance: How do you maintain your device data?
  7. Future Proofing: How do you manage the rapid rise in device combinations, and keep performance and accuracy standards high?
  8. Unclassified Device Handling: How are new, as yet unclassified devices, operating systems and browsers detected? Does the solution provide an accuracy indicator when a precise result is not available?
  9. Integration: How do you integrate with your existing analytics systems?
  10. Insights: Can your service offer device insights like retail price, device age and page response times?
  11. Proven: Is your software proven, used by millions, including the world’s leading brands?
  12. Integration: Will your software integrate with our existing systems?
  13. Free Trial: Can I get a free trial using my data - Can I install the software for free?

Further Reading: