A response to a login alert can stop an impostor or inconvenience a real customer. The number of alerts alone cannot tell a business which of those happened.
Akamai’s October 29, 2026 research blog illustrates the distinction. It says 26.6% of login attempts in a 30 day sample from its Account Protector product were classified as high or critical risk. The blog’s text does not specify the sample’s dates or number of attempts.
That is a share of attempts receiving a risk classification. It is not a count of confirmed account takeovers, affected customers or money lost. Nor is the vendor sample an estimate of fraud across India.
The Breakdown
The product assesses requests using signals such as earlier devices and locations, patterns of behaviour, source reputation and bot detections. Where an individual profile is unavailable, it can draw on population profiles.
The score feeds a response chosen by the organisation. Akamai’s technical documentation describes additional authentication or correlation with other fraud information as possible steps. Scoring a request, taking an action and establishing an incident are different stages.
The cost of getting the label wrong
Akamai’s documentation explicitly recognises false positives, where legitimate traffic is classified as malicious, and false negatives, where malicious traffic is treated as legitimate. It provides a way for customers to report those errors.
That creates two questions for a business reviewing its controls. How much genuine abuse is being detected? And how often are legitimate users being interrupted? A large flagged share does not answer either question on its own.
The same caution applies when comparing periods or vendors. A change in the traffic being sampled, the scoring definitions or the action policy can change what the dashboard is counting. Consistent definitions matter before a percentage becomes a performance target.
The Crunch
For an Indian business running customer accounts, the useful review keeps flagged requests, confirmed incidents and the effect on genuine customers separate. The research offers a reason to examine that chain, not a ready made fraud rate for the business.
A security score is valuable when it helps decide what to check next. Treating it as the final outcome obscures both missed abuse and unnecessary customer friction.
Leave your thoughts