Bot Traffic IndexThe bot problem, quantified

GIVT vs SIVT: The MRC Split That Decides Whether Bot Traffic Gets Refunded

Your ad report says 8 percent invalid traffic. Your vendor says 2 percent. Both numbers come from the same campaign, and the difference is not a rounding error. It is the split between the traffic a list can catch and the traffic that needs a lab.

The two tiers, stated plainly

The Media Rating Council splits invalid traffic in two, and the split governs what follows. General Invalid Traffic, or GIVT, is activity identified through routine filtration using lists or standardized parameter checks. The standard names seven areas: known invalid data-center traffic, bots, spiders, and crawlers, activity-based filtration on transaction-level campaign data, non-browser user-agent headers, pre-fetch traffic where the ad was never subsequently accessed, non-standard delivered ad sizes of 0 by 0 and 1 by 1, and sessions incapable of rendering images, such as headless browsers. Sophisticated Invalid Traffic, or SIVT, covers cases requiring advanced analytics, multi-point corroboration, and human intervention: automated browsing from emulators and hijacked consumer devices, click farms where non-robotic behavior is organized across users, and falsified measurement events spanning impression, viewability, click, location, referrer, consent string, and conversion attribution.

The MRC issued the Invalid Traffic Detection and Filtration Guidelines as final on October 27, 2015, establishing the GIVT and SIVT structure the industry still uses, and issued IVT 2.0 as final on June 25, 2020, adding data-center requirements, the decision rate, purchased-traffic disclosure, and up-front filtration guidance. Every accredited measurement organization must apply GIVT detection, which is not separately accredited because it functions as a floor. SIVT detection is optional and separately accreditable, and SIVT accreditation subsumes GIVT, because catching the harder class presumes catching the easier one.

Why your vendor's number and your number disagree

GIVT is meant to be consistent between vendors, because it leans on shared resources: the IAB and ABC International Spiders and Bots List, maintained by the Alliance for Audited Media on behalf of IAB Tech Lab, and the data-center address list run by the Trustworthy Accountability Group. On data centers the MRC requires filtration of invalid traffic from the three largest hosting entities, naming Amazon AWS, Google, and Microsoft. Two vendors measuring GIVT should land close, because they are reading the same lists.

SIVT is where the numbers diverge, because SIVT is a judgment, not a list lookup. A botnet that mimics human mouse movement and dwell time, a hijacked phone browsing in the background, a click farm with real humans clicking on schedule: these need behavioral analysis, multi-point corroboration, and often a human reviewer, and two vendors with different models will score them differently. The MRC's benchmarks reflect that gap: under 10 percent for GIVT and under 1 percent for SIVT in most display contexts, though benchmarks vary by channel. A campaign reporting 8 percent invalid traffic that is mostly GIVT has a filtering problem. A campaign reporting 8 percent that is mostly SIVT has a fraud problem, and the refund conversation is a different one. For the referral-spam version of this, see Referral Spam in GA4: The Fake Traffic in Your Reports, and for the proxy version, Residential Proxies: Why the Blocklist Stopped Working.

What to write into the buy

The contract should name the split, not just the total. Require GIVT filtration as a floor, require SIVT detection with a named vendor and a named accreditation, and require the decision rate and the purchased-traffic disclosure from IVT 2.0. And require the IVT rate per tier in reporting, not a single blended number, because a blended number hides whether the problem is a list or a lab. A vendor that will not break out GIVT and SIVT is a vendor that does not want you to see which one it is failing at.

Frequently asked questions

What is the difference between GIVT and SIVT?

GIVT is invalid traffic identified through routine, list-based filtration: known bots, data-center IPs, crawlers, headless browsers. SIVT needs advanced analytics, multi-point corroboration, and human review, because it is built to look human: hijacked devices, click farms, domain spoofing.

Why do two vendors report different invalid traffic rates?

GIVT should be consistent, because vendors read the same shared lists. SIVT diverges because it is a judgment, not a lookup, and different models score human-mimicking traffic differently.

What are the MRC benchmarks for invalid traffic?

Under 10 percent for GIVT and under 1 percent for SIVT in most display contexts, though benchmarks vary by channel. A campaign far above either number has a filtering problem or a fraud problem, and the split tells you which.

What should an ad contract say about invalid traffic?

Name the split: require GIVT filtration as a floor, SIVT detection with a named vendor and accreditation, the decision rate, and purchased-traffic disclosure. Require the IVT rate per tier, not a single blended number.