What is invalid traffic?

Every time an ad loads, someone gets paid. The advertiser pays for an impression, a click or an install, and that money flows back through a chain of companies to the website or app that showed the ad. The whole system rests on one quiet assumption: a real person was on the other end. When that assumption is false, the industry calls the activity invalid traffic, usually shortened to IVT.

Imagine a shopping mall that charges shops rent based on how many people walk past their windows. Some of the "shoppers" counted by the door sensor are cleaners, security guards and delivery staff (honest, but not customers). Others are cardboard cut-outs someone wheeled past the sensor on purpose, because the mall owner gets a bonus per visitor. Both groups make the count wrong. Only the second group is fraud.

That analogy captures the single most important idea in this track: invalid is not the same as fraudulent. IVT is the umbrella. It includes innocent robots such as search-engine crawlers and uptime monitors, which nobody should pay for but which mean no harm. It also includes deliberate ad fraud: traffic manufactured to steal advertising money.

Where the definition comes from

The formal definitions come from the Media Rating Council (MRC), the US body that audits and accredits ad measurement. Its MRC IVT guidelines (first published in 2015 and updated in June 2020) define IVT as traffic that does not meet ad-serving quality or completeness criteria, or otherwise does not represent legitimate traffic that should be counted. The MRC then splits IVT into two categories, GIVT and SIVT, which the next lesson covers in depth. Other bodies (TAG, the IAB, and measurement firms worldwide, from Europe to India and Japan) use the same two buckets, so the vocabulary travels.

The four families of invalid traffic

Why anyone bothers to fake traffic

Because in programmatic advertising, machines buy from machines in milliseconds, and most checks happen after the money has already moved. A fraudster does not need to fool a human buyer. They only need to produce a bid request that looks like a real person on a real site, win the auction, and have the resulting impression counted. At a CPM of a few dollars, a botnet producing millions of impressions a day becomes a very profitable business with little physical risk.

The incentives also leak into honest businesses. A publisher paid per impression gains from inflated traffic, and a middleman paid on volume gains from more volume. Nobody has to be a criminal for invalid traffic to slip through. They just have to stop asking questions. That is why measurement of traffic quality is done by independent third parties rather than by the people selling the ads.

IVT looks different in every channel

On the open web, invalid traffic often means bots loading pages full of display ads. In mobile apps it frequently means fake installs or stolen credit for real ones. On connected TV it can mean servers pretending to be millions of living rooms. In search and social, it usually shows up as invalid clicks that platforms filter and credit back. The mechanisms change, but the question stays the same everywhere, from a US retailer's CTV campaign to an Indian fintech app's install drive or a German publisher's display inventory: was a real person genuinely there?

A simple worked example shows why the rate matters. A brand spends $100,000 at a $5 CPM and buys 20 million impressions. If 10% are invalid, $10,000 bought nothing. Worse, if the campaign's software then shifts budget toward the sources with the "best" engagement, and those happen to be bots, next month's waste can be larger.

Who gets hurt

• Advertisers pay for audiences that do not exist, and then their optimisation software learns from fake behaviour and sends even more money to the fake sources.
• Honest publishers compete for the same budgets against fake inventory that is cheap to produce, which pushes prices down for everyone.
• Ordinary people often unknowingly host the fraud: their laptops, phones and TV boxes are infected and used as bots, wasting their bandwidth and exposing them to other crime.
• The whole ecosystem loses trust, which drives money into walled gardens and away from the open web.

The words you will need

Throughout this track you will see an IVT rate (the share of measured traffic flagged as invalid), traffic quality (the broader question of whether traffic is human, attentive and genuine), and detection methods that work either before an ad is bought or after it is served. Keep the mall in mind: our job is to count the real shoppers, ignore the cleaners politely, and catch the people wheeling cardboard cut-outs past the door.