Your SlideShare is downloading. ×
ALPS WG Update - IAB Ad Ops Summit, Fall 2009
Upcoming SlideShare
Loading in...5
×

Thanks for flagging this SlideShare!

Oops! An error has occurred.

×

Introducing the official SlideShare app

Stunning, full-screen experience for iPhone and Android

Text the download link to your phone

Standard text messaging rates apply

ALPS WG Update - IAB Ad Ops Summit, Fall 2009

384
views

Published on

Working group status update at the Interactive Advertising Bureau's 2009 Ad Operations Summit in New York, Nov 16, 2009. …

Working group status update at the Interactive Advertising Bureau's 2009 Ad Operations Summit in New York, Nov 16, 2009.
The Ad Load Performance Scoring (ALPS) working group, with membership from AOL, Yahoo, Microsoft and Google, is developing a method for 'scoring' the load performance of ads, and incorporates best-practices compliance as well as measured load speed.

Published in: Technology

0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total Views
384
On Slideshare
0
From Embeds
0
Number of Embeds
1
Actions
Shares
0
Downloads
1
Comments
0
Likes
0
Embeds 0
No embeds

Report content
Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
No notes for slide
  • VRTA is a performance analysis tool for web site profiling and optimization The current version of the tool—without the ad scoring—is available as a free download from MSDN Browser tool used to fetch ads has been modified to flag unique requests It does so by injecting ICMP Ping messages to single NavigateComplete2 events Prototype will allow for further study by the Working Group
  • Scoring should be browser independent as well, but further study is required Could score across different browsers to account for possible differences in behavior between browsers The leading causes of slow Ads involve… Geographic latency between the end user and the servers The amount of ad content … and the number of network round trips needed to fetch this content The way a particular browser downloads and processes the Ad selection and content requests Other factors include… The way the Ad has been implemented See the best practices… The design of the publisher’s page See the best practices… The quality of the user’s network connection Packet loss, congestion, routing, DNS lookup times The capabilities of the end user’s computer Processor speed, bandwidth, memory The time spend in the servers of the Ad Network selecting an Ad
  • Benefits at each level Ad Network, Publisher, Advertiser Ad Agency Ad Agency, Ad Network, Publisher Ad Agency, Ad Network, Publisher Ad Network, Publisher Ad Network, Publisher
  • Will be conducting additional review with Microsoft Research Small-scale study will compare predicted Ad download times with results under simulated WAN conditions Larger-scale study will compare predictions with results of Ad download times measured in production environments Discuss opportunity with IAB Research Council to tie key Business metrics with the Ad Load Performance Score
  • Transcript

    • 1. Ad Load Performance Scoring Working Group Update Eric Goldsmith Tony Ralph, Pramod Khincha, Bryant Mason, Mark Masterson, Sameer Ajmani 16-Nov-2009
    • 2. Motivation
      • Does speed matter?
      • Causal connection between page load-time and user behavior
        • Public data from Google, Microsoft, Facebook, AOL, Amazon, etc.
        • Non-public data from even more ;-)
      • When pages load faster, users:
        • Browse more
        • Search more
        • View more
        • Buy more
        • DO MORE
      p
    • 3. Motivation cont’d
      • Faster Pages Users Do More See More Ads
      • Faster Ads Faster Pages … See More Ads
      • How to make ads faster?
        • Qualitative
          • Follow Best Practices
            • www.iab.net/media/file/IAB_Ad_Load_Perfomance_BP_FINAL.pdf
          • Audit Compliance (tool)
          • Best practices important, but don't guarantee fast ads
        • Quantitative
          • Measure load time (tool)
          • Identify / resolve performance issues
      p
    • 4. Ad Load Performance Scoring ALPS
      • ALPS Working Group goals
        • Establish standards for measuring ad load performance
          • Test individual ads in isolation
          • Qualitative Measure - compliance with Best Practices
          • Quantitative Measure - score ads to predict actual ad download time
        • Provide common language for ad speed, to help specify
          • Agency / Vendor delivery performance
          • Publisher speed expectations / requirements
        • Extend existing Ad Specs to incorporate load performance
          • http://www.iab.net/iab_products_and_industry_services/508676/508767/Rich_Media
        • Allow independent testing
          • Hosted service provided by IAB
          • Open to all members
      p
    • 5. Qualitative - Best Practices Scoring
      • YSlow
        • Tool used internally at Y! and by many publishers to improve web page performance
        • Measures best practices compliance and generates easy to understand letter grades [A-F]
        • Help in creating better user experiences by improving quality earlier in the cycle
        • Studies have shown that following these rules can improve performance by 25% - 50%
      • YSlow for Ads
        • Develop specialized rule set to score against Ad Load Best Practices
        • Rules, thresholds & scoring details will be published
          • Can be implemented in any tool
    • 6. Best Practices Highlights
      • How to speed it up?
        • Reduce content!! Rule of thumb – No more than 4 components a page
        • Not Found Error – One of your ad components is missing which renders the ad useless
        • Too many serving domains per ad – DNS Lookups are expensive – restrict to 2 domains per ad
        • Avoid redirects – Not optimal to use redirects for counting.
        • Cache aggressively -
      • Easy Optimizations – Slim it down
        • Compress as much as possible – Easy setting on serving side
        • Reduce cookie size
        • Use a CDN when you can
        • Minify JS/CSS
        • Avoid duplicate JS/CSS -
        • Optimize Images
    • 7. Best Practices cont’d
      • Advanced Techniques – Best user experience
        • Avoid nesting IFRAMEs
        • Do not scale images in HTML
        • In-lining vs. caching of JS/CSS
        • Counting beacons no more than 1
      • Ad Ruleset Demo using AdTool
    • 8. Quantitative - Load Performance Scoring
      • Microsoft is a developing a prototype to profile and score Ad download performance
        • Tool is based on Visual Round Trip Analyzer
        • The “score” is the count of network packets in the critical path of the Ad response
        • The algorithm quantifies Ad performance independent of end user’s network connectivity and geographic location
      • Prototype can predict real-world Ad download times
        • Given the score and the location of a target user, the tool can estimate actual Ad download times by accounting for various underlying network interactions
        • The network modeling code was developed by Microsoft Research and is based on published studies
      p p
    • 9. Load Performance Scoring cont’d
      • The scoring algorithm is platform independent
        • Analysis is performed on captures of network traffic
        • Scoring could be built into existing Ad preview tools and Creative Acceptance processes
      • The resulting score is also independent of the Ad Network
        • Scoring can be done before the Ad has been submitted and published
      • This approach complements existing measurements conducted by Publishers and in Ad Networks—but will flag performance issues earlier in the Ad lifecycle
      p p
    • 10. Load Performance Scoring Example Ad Load Performance Score Result Total number of packets 8 Predicted Download Times Time in ms Domestic end users 270 Nearby international end users 798 Remote international end users 1589 Ad Load Performance Score Result Total number of packets 24 Predicted Download Times Time in ms Domestic end users 598 Nearby international end users 1770 Remote international end users 3527
    • 11. The View from the Summit
      • Expected benefits of the ALPScore ™
        • Correlate Ad download times with Business performance metrics, such as CSAT, abandonment, and CTR
        • Highlight opportunities for optimization of Ad download times during content development
        • Provide comparison of download performance of new Ads with similar Ad products
        • Produce estimates of Ad download times for new markets
        • Act as an additional standard for Creative Acceptance
        • Provide an additional criteria for selecting Ads at runtime
      p p
    • 12. Next Steps
      • Finish development of features needed for both scoring tools
      • Run a small-scale study in a controlled lab environment
      • Refine scoring based on feedback and test results
      • Perform a larger-scale study with Ads from multiple vendors and Ad Networks
      • Conduct research to correlate Business metrics, CSAT, and AdSat with the Ad Load Performance Score
      p p
    • 13. Questions? p p