How LinkedIn Audience Network protects advertisers from invalid traffic

A look at the measures and tools that LinkedIn uses to ensure ads quality and transparency in its Audience Network

LinkedIn Audience Network (LAN) is a platform that helps advertisers reach LinkedIn members across our network of trusted partners. However, one of the challenges of online advertising is the presence of Invalid Traffic (IVT), which can include activity that does not originate from a genuine human user, such as bots or humans with inauthentic intent. IVT can negatively impact the performance, measurement, and ROI of advertising campaigns, so it’s essential that we take steps to measure and mitigate it. Broadly, there are 2 types of IVT: 

General Invalid Traffic (GIVT): GIVT includes automated bot traffic, spiders, crawlers, and other non-human traffic sources. Identifying and filtering out GIVT involves employing standard techniques and algorithms, such as verifying the User Agent, IP Address, and referrer URL of the traffic/ad supply.

Sophisticated Invalid Traffic (SIVT): SIVT refers to more advanced forms of fraudulent traffic that can mimic human behavior and evade traditional detection methods. This includes techniques such as ad stacking, domain spoofing, click fraud, and human behaviors with malicious or disingenuous intent. SIVT can be more challenging to detect and requires advanced algorithms and technologies to identify and mitigate effectively.

In this article, we'll explain some of LAN's methodologies for measuring and protecting advertisers from GIVT and SIVT. Additionally, we'll explain the differences advertisers see between their landing page analytics and LAN's campaign performance data in Campaign Manager.

How LAN detects and prevents IVT

LAN uses various tools to detect and prevent IVT, including:

1. We curate high quality publishers (apps, sites, and CTV platforms) for LAN. Standards for inclusion in LAN consider multiple elements such as content, traffic, and site experience. We regularly review for adherence to these standards with both manual and automated processes, removing publishers that don’t meet our standards.
2. We implement pre-bid and post-bid filtering using machine learning and real-time or near real-time signals to work to detect and prevent IVT. We use both proprietary data and third-party services from MRC-accredited partners such as DoubleVerify and HUMAN Security.
3. We collaborate with respected third-party partners to independently measure IVT.

Publisher curation

We work with our ad exchange partners to enable multiple layers of automated fraud prevention and to filter out low-quality publishers and placements for ads on LAN. LinkedIn maintains its own publisher allow list which can be accessed from the Brand Safety Hub in Campaign Manager. We consistently review and curate trusted, top-engaging publishers to add to our allow list. We review publishers and remove them from our network if they do not meet our advertising guidelines.

Pre-bid filtration of requests

Several noteworthy complexities within our IVT prevention system include Peak Queries per second of ~2 Million, multiple filters, and third-party services. The filtration process incorporates various parameters as inputs, including but not limited to: Exchange, Ad Format, Publisher, IP Address, User Agent, and certain first-party data.

IVT Filters Applied to Real Time Bidding Request (RTB): In order to understand RTB request risk, we use different industry-standard databases to evaluate the likelihood the request is IVT and drop the request if we determine it is invalid.

Post-bid filtration of impressions and clicks

Every impression and click received by LAN undergoes a series of validations. These include deduplicating multiple impressions and clicks originating from the same request and discarding late arrivals. After this, IVT filters are applied to various parameters such as IP Address and UA (User-Agent Header), along with those related to known GIVT and SIVT sources. We take many measures to invalidate these impressions, ensuring advertisers are not billed for them.

IVT filters applied to impressions and clicks

1. Ad Validator: On the supply side, publishers/exchanges seek the support of ad security or verification companies to shield users from harmful, intrusive, and deceptive ads. These companies play a pivotal role in maintaining the integrity of the advertising ecosystem by preventing the spread of malware and bad ads. LAN works to identify impressions and clicks generated by these types of ad validation technologies (such as web crawlers) through means such as including referrer URLs, IP Address, Audit Macros, etc. In such instances, we work to ensure Ad Validator’s operations continue smoothly without counting those impressions, and without billing advertisers for those impressions.
2. Industry Collaboration: We believe industry collaboration is crucial in combating invalid traffic. We work with various industry players and utilize multiple databases, including TAG Known Datacenter IP Addresses list, IAB Bot User Agent List, and proprietary databases developed internally and fetched from trusted vendors like Pixalate, HUMAN Security, and DoubleVerify.

Discrepancy with Web Analytics

Figure 2 above illustrates a scenario involving two different clicks. Click-1 is detected as IVT by LAN’s post-bid filters. Although the advertisers are not billed for this click, it still redirects to the advertiser's landing page, where web analytics tools may identify it as IVT. Conversely, click-2 passes LAN’s post-bid IVT validation, is considered a genuine click, and is billed to the advertisers before redirecting to the advertiser's landing page.

The majority of click filtration occurs from Ad Validators. When a user clicks on an ad displayed on a publisher's site, the click is directed to the advertiser's landing page through LAN. Similarly when an Ad Validators scan the landing page for malware, they simulate clicks which pass via LAN. Through our trusted partnerships  with providers of Ad Validators, we can identify and filter traffic from Ad Validators so that we don’t report or charge these clicks. However, we still redirect the HTTP request to the landing page for Ad Validators to function smoothly. This traffic may be identified as IVT by analytic tools on the landing page since they do not have access to the Ad Validator data as we do. This type of traffic often results in high bounce rates and low session times for advertisers.

Independent IVT Measurement

We understand the importance of transparency and not grading our own homework, which is why we rely on partners like DoubleVerify and HUMAN Security to independently measure post-bid IVT on our platform. We also completed our post-bid integration with HUMAN Security to flag and measure IVT for both LAN and CTV recently.

These evaluations start with our Ad Serving flow. When responding to an RTB request, we sample responses and, depending on the ad format and the chargeability of the impressions, include a partner tag in the ad creative. This tag loads on the end user's device and sends all the required signals for IVT measurement to DoubleVerify and HUMAN Security, who then independently measure the traffic.

Conclusion

LinkedIn Audience Network (LAN) has invested heavily to work to protect advertisers from invalid traffic and ensure the integrity and effectiveness of advertising campaigns. By leveraging advanced filtering techniques, high-quality publisher curation, and collaboration with industry-leading partners such as DoubleVerify and HUMAN Security, LAN detects and mitigates both GIVT and SIVT. Additionally, the transparency provided through independent IVT measurement and the detailed explanation of discrepancies with web analytics underscore LAN's commitment to delivering trusted and accurate advertising metrics. Through these robust protections, LinkedIn continues to offer advertisers a reliable and high-performing platform for reaching their target audiences.