A bug in Facebook’s conversion list studies tool created a buzz among the advertisers. It was discovered that the bug was causing issues by feeding false data to the advertisers.
It has been a year since the issue is there in the tool. It is reported to be a code error which resulted in undercounting the conversion lift test managed by Facebook. The tool is been used by thousands of advertisers between 15 August 2019 till August 31 2020.
The Conversion lift studies tool is free provided by Facebook used by the advertisers. It is used, for the measurement of the growth of the audience of their advertisement. The advertisers accordingly decide the spent which they want to do on their advertisements and usually spend more to boost the reach. The bug came as the real shock for the advertisers using this tool, and they felt cheated!
The bug was fixed by Facebook on 1st September 2020. Therefore, it took more than a year for Facebook to recognize the fault and fix the issue, which was shameful for such a big tech. A company having millions of user database of billions.
According to a Facebook spokesperson, they are “working with impacted advertisers,”
However, there were no alerts released by Facebook until the last week.
It is not the first time something like this happened with Facebook. A well-known case of miscalculations in video matrics was discovered in the year 2016. Also, in the year 2017, there was a case of overestimation of its reach. These issue has caused a wave of mistrust with Facebook, and the advertisers are looking for other better options.
It is what happened!
The bug was associated, with the data pipeline migration. It caused in registering impressions into Facebook’s conversion lift system.
It resulted in undercounting of the conversions, for those people who were using the impressions on the Facebook app.
Therefore, the miscalculation of conversion happened, and the measurement of sales was not accurate. It is the crucial ratio used for the calculation of the incrementality. Also, it is utilised to calculate other ratios! These ratios include conversion lift metrics.
However, to resolve the issue and compensate the advertisers, Facebook is giving away a token. A one-time coupon is offered, as compensation to the affected advertisers. The token has to be used by March 31 2021.
The tokens are distributed, according to the severity of the problem and discomfort caused to an advertiser. Thus, if Facebook feels that an advertiser is not affected by the issue, they will not get the token as compensation. Therefore, it is totally up to Facebook to decide whether the advertiser is affected or not!
However, it won’t be fair to decide without talking to the affected advertisers. The distribution of the token will be totally up to the Data science team of Facebook. Therefore, it can cause a lack of transparency in the process of token distribution.
Even though the tool is free to use for the advertisers provided by Facebook, but it is responsible for the spending made by the advertisers and directly affects the amount of money spent by them on their campaigns.
Facebook way of saying, “ Sorry” to the advertisers!
An email circulated by the company(Facebook), to the affected advertisers to provide clarity on the delay in discovering the bug stated, “due diligence to better understand the scope of the issue.”
However, most of the ad buyers are not buying the explanation provided to them for the delay!
According to an account holder, “People continued to spend money after Facebook knew there was a bug because they weren’t told.” According to the buyer, the tone of the email was “vague and corporate tone.”
Even the subject line of the email was very formal and was empathetic. The user was more interested in “‘Oh, my god, we effed up, and we’re sorry!” Although, what it said was “‘Upcoming update to Facebook conversion lift studies.” Which was not very reassuring.
While debugging the bug, the developers came across two more as quoted by Facebook “Smaller” bugs.
The first one dealt with the multi-cell studies, and the second one was responsible for the undercounting of the data for the web conversions in the statical model.
Facebook claims to develop better technology to counter such bugs more effectively in the future. However, we must all keep our fingers crossed.