Bot Benchmark study - White Ops & DCN
Download as PPTX, PDF1 like814 views
Digital Content Next (DCN) conducted a study with 32 member companies and White Ops to analyze bot traffic. They found that DCN members had significantly lower bot traffic than industry averages, with 2.8% bots for display ads versus 23% from another study. Video ads saw an even greater reduction, with only 0.5% bots versus 11% elsewhere. However, some DCN members still experienced bot rates over 6%. The study concluded that premium publishers can maintain high human audience levels through quality control policies, direct sales, and transparency around inventory sources and metrics.
Bot Benchmark study - White Ops & DCN
- 2. Advancing the Future of Trusted Content
- 3. Digital Content Next Digital Content Next is the only trade association that exclusively serves the unique and diverse needs of high-quality digital content companies that manage trusted, direct relationships with consumers and marketers.
- 4. • Opportunity for DCN to take a proactive leadership position while the rest of the industry is swirling • Initiated study with White Ops after Board approval at April 2015 meeting • 32 participating member companies • Michael Tiffany, CEO of White Ops, is here to present a first look at the results • Goal of this meeting is to answer questions and make sure members have this intelligence in their knowledge base DCN Bot Benchmark Study
- 5. 2.8% 2.5% 11% 23% 0% 5% 10% 15% 20% 25% Display Video 89% less bot traffic in video; 75% less in standard display impressions on premium publisher sites DCN 2015 ANA 2014 Top-Line Results
- 6. DCN Perspective "We always knew DCN members had the most trusted content by consumers but now there is additional proof they should also be the most trusted experiences for advertisers. Period."
- 7. DCN Perspective “Quality content is the best filter for reaching engaged, human audiences. Based on this study we know that approximately 97% of the digital ad fraud happens elsewhere on the web.”
- 8. DCN Perspective “It’s incredibly important that all premium publishers know the work doesn't stop here. The policies and brand trust of premium publishers are something which are earned and added to each and every day. While we are in a rapidly evolving environment, this confidence and trust in the context and quality of our brands is one of our most sustainable assets. We will and must continue to lead the industry from here forward.”
- 9. Thank you to DCN Ad Fraud Committee • Dan Stubbs, Conde Nast – Co-chair • Andy Wilson, Meredith – Co-chair • Ryan Whittington, American City Business Journals • Stephen Felisan, Edmunds.com • David Payne, Gannett • Mike Kisseberth, Purch • Tammy Franklin, Scripps Networks • Jim Spanfeller, Spanfeller Media Group
- 10. Confidential. Copyright 2015 White Ops, Inc.Confidential. Copyright 2015 White Ops, Inc. 10
- 11. Confidential. Copyright 2015 White Ops, Inc. 11 Study Background and Context Executive Summary Detailed Findings and Supporting Data Conclusion and recommendations
- 12. Confidential. Copyright 2015 White Ops, Inc.Confidential. Copyright 2015 White Ops, Inc. 0.00% 1.00% 2.00% 3.00% 4.00% 5.00% 6.00% 7.00% 8.00% 0 10 20 30 40 Advanced Bot Percent DCN benchmark study overview 12 DCN Global Results 32 participating organizations 30 billion impressions 2.8% advanced bots Range: 1.6% to 6.9% 2.7% IAB Bots Range: .7% to 11.4% Average (All Bots): 5.5% Range: .7% to 11.4%
- 13. Confidential. Copyright 2015 White Ops, Inc.Confidential. Copyright 2015 White Ops, Inc. Implementation details Testing & QA: 6/22 - 7/1, 2015 Data collection: 7/1 - 8/14, 2015 • Compares individual and global DCN results with ANA 2014 study. Analysis is based on deterministic detection of bot traffic • To match ANA 2014 The Bot Baseline, we calculate results using measurable desktop (non-mobile) inventory. For total bot percentage calculations, we remove IAB registered bots and spiders. • No data or results were provided to participants or the DCN during the study 13 • White Ops Detection Tags were deployed by study participants via their ad server, directly into the pages of their website, or a combination of both • Media types were specified by participants. Where possible, White Ops deduced media type programmatically at a per-impression level • For unmeasurable desktop inventory (<10% of total impressions for DCN participants), we assume that the bot percentage is similar to measurable inventory bot percentages. Some publishers choose to sell inventory across web properties that they do not own and operate, commonly referred to as partner networks, syndication networks or audience extension networks. Such inventory falls outside the scope of the results in this study.
- 14. Confidential. Copyright 2015 White Ops, Inc.Confidential. Copyright 2015 White Ops, Inc. Definitions • Decision – deterministic identification of a particular impression, page view, or other type of online event as bot or not • Advanced bot - an invalid impression positively identified as a bot using deterministic methods • IAB registered bot – percentage of bots that are on the IAB list • Href domain – the domain where a particular ad impression, page view, video play, or other online event occurred • True domain – verified URL/domain which was loaded • Referring domain – domains that send traffic to the site • Incomplete load / non-measurable – cases where the JavaScript tag was not loaded due to factors such as browser abandonment or site configuration • False-positives – detection results that indicate bot or ad fraud impressions when the impressions are actually legitimate 14
- 15. Confidential. Copyright 2015 White Ops, Inc. 15 Study Background and Context Executive Summary Detailed Findings and Supporting Data Conclusion and recommendations
- 16. Confidential. Copyright 2015 White Ops, Inc. The DCN average bot percentage for display inventory was one quarter of the level of ANA 2014 display baseline study data, and DCN average bot percentage for video inventory was one tenth of the bot rate seen in ANA 2014 baseline study inventory. Selectively buying direct-placement digital ad inventory from this group of premium publishers, if quality is maintained consistently across inventory, could allow marketers to achieve the humanity levels of the top 19% of the ANA 2014 Bot Baseline marketers without instituting any other policy changes or buying decisions. 16 Benchmark Bot % by Media Type
- 17. Confidential. Copyright 2015 White Ops, Inc. Some premium inventory sees bot hotspots While the middle-of-the-road publisher in the DCN had a 2.8 percent sophisticated bot rate, a single publisher in the group encountered a 6.9 percent sophisticated bot rate, more than twice the average for the group. Another five publishers had sophisticated bot rates exceeding 4 percent. 17
- 18. Confidential. Copyright 2015 White Ops, Inc. 18 Study Background and Context Executive Summary Detailed Findings and Supporting Data Conclusion and recommendations
- 19. Confidential. Copyright 2015 White Ops, Inc. Sophisticated bots mimic viewability Bots are spoofing viewability in nearly three-quarters of cases. Bots that come from a host dedicated to fraud, with no real human traffic -- referred to as isolation bots -- have high viewable rates and most frequently post back detailed viewability data. 19
- 20. Confidential. Copyright 2015 White Ops, Inc. How publishers source traffic affects bot % • Large content recommenders -- such as Jungroup, Outbrain and Taboola -- refer very little sophisticated bot traffic. 20 • Traffic from reputable search engines, such as Google and Yahoo, does have large sophisticated bot rates when coming from certain top-level domains. • Smaller search engines show medium- to high- levels of bot traffic. • Traffic from social-media sites tended to have lower bot rates.
- 21. Confidential. Copyright 2015 White Ops, Inc. Ad injection affects publishers CASE STUDY: A participant’s ad injection rate was higher than their sophisticated bot rate. Ad injection to one participant was greater than 3%, while sophisticated bot traffic was less than 2%. Similar levels of injection likely affect similar publishers. 21
- 22. Confidential. Copyright 2015 White Ops, Inc. False-positives affect publishers 39 percent of publishers surveyed said they had been presented with reports of Invalid Traffic (IVT) in their data. A sample of DCN publishers and their potential monthly losses from 7% FPs in vendor reporting This graph uses the vendor’s White Ops-detected sophisticated bot % with the vendor’s self-reported CPM, volume averaged over the group. 22
- 23. Confidential. Copyright 2015 White Ops, Inc. Strategies improve inventory quality Case study: A top-volume publisher who did not source traffic and only sold inventory via direct buys had a very low bot average. 23 0 5 10 15 20 25 30 35 0 50000 100000 150000 200000 250000 300000 350000 "287551497" "308143857" "334687257" "351258897" "347426457" "331424337" "340927977" "337789737" "185745537" "344443857" "344633457" "337620417" "323058537" "301225737" "347242377" "328554777" "342191457" "341550297" "351626937" "349781097" "346942137" "337582137" "347810937" "321298137" "346838217" "348695817" "317722377" "329149977" "348187497" "345548217" "337355217" "318342417" "307300617" "342371097" "307305537" "335339457" "296921697" "330596817" "331442937" "347126937" "345571497" "351956817" "355355577" "353548617" "337700577" "353078697" "348994857" "251808297" "294549657" "185790657" "0" All campaign bot percentages through study period Decisions (Thousands) Advanced Bot Percent
- 24. Confidential. Copyright 2015 White Ops, Inc. 24 Study Background and Context Executive Summary Detailed Findings and Supporting Data Conclusion and recommendations
- 25. Confidential. Copyright 2015 White Ops, Inc.Confidential. Copyright 2015 White Ops, Inc. Conclusions While high-quality publishers have a lesser sophisticated bot problem than the advertising ecosystem in general, publishers can maintain high humanity levels through policies and strategies. Sophisticated bots will continue to better emulate humans. Increasingly, parasitical bots target cookies or implant themselves in legitimate browsers, allowing the automated traffic to don the cloak of legitimacy and appear to be an actual human visiting sites. Data-informed strategies can help publishers avoid third parties who provide traffic with high sophisticated bot rates. Direct buys to premium publishers who demonstrate the commitment to maintain low sophisticated bot rates can consistently yield human audience levels of 97% or more. Publishers can protect their marketers by agreeing to transparency in inventory quality, allowing tracking of inventory quality for validating traffic bot and humanity percentages, and by agreeing to bill based on humanity. 25
- 26. Confidential. Copyright 2015 White Ops, Inc. If sourcing traffic, monitor sourced traffic to protect marketers from increased risk Select suppliers with proven high humanity levels and a commitment to quality Insist on vendor transparency for detection methodology to improve validation For inventory with irregular humanity percentages or high CPMs, monitor for bot activity to limit bot spikes Improve page measurability to precisely manage inventory quality Monitor for ad injection and other attacks that could damage your brand Use bots’ evasive techniques to identify and expose them 26 Recommendations Publishers can win more RFPs, secure larger shares of advertiser budget and justify higher CPMs by selling on your biggest market advantage: a premium human audience.
- 27. Confidential. Copyright 2015 White Ops, Inc. Thank You! www.whiteops.com
Editor's Notes
- #3: Since April Board meeting, 11 new members added Total member companies now 67
- #13: Describe the scope of what we were able to tag for the publisher Coverage info, type media, how we got the missing data Compete numbers – unique users per month, gauge the relatively
- #14: Describe the scope of what we were able to tag for the publisher Coverage info, type media, how we got the missing data Compete numbers – unique users per month, gauge the relatively