Facebook finds three more errors in the way it measures ad-effectiveness

mercredi 16 novembre 2016

Facebook has discovered three more errors in its measurements of ad-effectiveness, as part of an investigation into how it inflated the average number of times video ads were viewed,

The social network's inquiry into the metrics marketers use to try to compare it to other venues for adverts — from TV to Twitter — came after it angered advertisers when it announced the mistake around video ads in September.

More from the Financial Times:
Facebook and the manufacture of consent
Face up to responsibility, Facebook and friends
Facebook and Google to restrict ads on fake news sites

Facebook has now found it also overestimated the time spent reading articles, miscalculated how many times the same viewer returned to view a business's page and underestimated how many videos were viewed to the end.

Carolyn Everson, vice-president of global marketing solutions, said Facebook had approached the probe with the "same rigour and discipline" as when it quickly changed its business model to make advertising work on mobile phones.

"I think disclosing the bugs and calculation errors we found, some of which are real edge cases, is a real indication of our commitment to transparency," she said.

Facebook engineers tested more than 220 metrics to discover the three errors.

On pages, when measuring how many people saw a page in a week or a month, it had simply multiplied the daily figure, without acknowledging that some people would have visited more than one time in the period. The real "reach" figure was 33 per cent lower on average over a week and 55 per cent lower over a month. The error, which has existed since May, will be fixed in the next few weeks.

In Facebook's Instant Articles sections, where it partners with publishers to host news stories, making them load more quickly and splitting revenue from the adverts within them, it wrongly reported the amount of time spent by about 7 to 8 per cent.

Finally, on videos, it discovered a second error, to do with how views are recorded on different mobile devices. This time, the real figures are more favourable. Videos watched to completion will now go up by about 35 per cent on average.

Facebook is also trying to soothe the concerns of advertisers by launching a measurement council, to hear their views on how to judge the effectiveness of ads. In addition, it is expanding its third party verification efforts with partners such as analytics firm ComScore, and even commissioning the measurement service Nielsen with creating metrics for its new livestreaming product Facebook Live.

"We are committed to this level of transparency and ongoing communication, while we all recognise we are continuing to evolve — we didn't have live video last year," Ms Everson said.

Follow CNBC International on Twitter and Facebook.

Let's block ads! (Why?)

Facebook finds three more errors in the way it measures ad-effectiveness

0 commentaires:

Enregistrer un commentaire