Nielsen is in the midst of a battle to count TV audiences of the future, but the company just disclosed it’s still having problems counting them in the present.

The media-measurement giant informed TV networks and other clients Wednesday that it has, since September of 2020, undercounted so-called “out-of-home” audiences — the people watching programs in offices, bars, hotels and other places — for national TV programming. Nielsen cited a “software issue” for the problem, and vowed to release corrected estimates starting in mid-January of next year. But the process, Nielsen said, is likely to take longer to remedy. Audience data for some of the months won’t be available until February, and release of information about out of home viewership between September and December of last year is still “TBD,” Nielsen said in its memo.

“As part of routine testing and quality controls, we recently identified an error that caused an understatement of reported out-of-home audiences for our National TV service. While there is no impact to most telecasts, and no impact to local television, we did find some variances for events that tend to yield larger out-of-home audiences, such as live sporting events,” Nielsen said in a statement. “The error has been corrected and Nielsen will be reissuing data from January 2021 to present in order to provide the industry with the most complete data.”

The gaffe comes after Nielsen has been under intense scrutiny for months. The Media Rating Council, an industry body that verifies audience measurement processes, removed its accreditation of Nielsen’s national and local ratings services in September after determining the company underestimated traditional TV audiences during the pandemic. Since that time, a number of big media companies, including NBCUniversal, WarnerMedia, ViacomCBS and Univision, have unveiled new efforts to build audience-measurement alternatives. ViacomCBS has started working with the large media-buying agencies owned by Dentsu, and expects to start testing transactions using other services early next year.

The error was derided by a close advocate for the TV networks. “Today’s announcement by Nielsen of more systemic errors that have further undercounted TV ratings for the last fifteen months is more of the worst possible news at the worst possible time, both for Nielsen and the buyers and sellers that use Nielsen data as trading currency,” said Sean Cunningham, CEO of the VAB, an industry organization that represents TV networks to advertisers. he called Nielsen’s mistake a “stunning omission” that adds to the “unprecedented scale of Nielsen’s 2021 undercounting and the depths of discovered defects in their core competency on TV measurement.”

Nielsen’s out-of-home counts were viewed as something of a panacea for the TV industr, when the measurement company agreed to start using them in the fall of 2020. For years, the TV networks have insisted they woo valuable audiences who watch news programming in corporate offices, sports events in bars and primetime programming in hotels and friends’ apartments, but never got credit for doing so. Nielsen executives said in 2019 that the new technology could result in an 11% uptick in audience for sports events and a 7% increase in audiences for news programming. Such viewership has become more important to traditional media companies as TV audiences migrate to streaming video hubs and social media video clips.

But the out-of-home service has also served as a bone of contention. Nielsen tried to delay implementing it in 2020, citing the effects of the coronavirus pandemic on consumers’ ability to watch TV in restaurants or doctors’ offices. Nielsen ended up following its original plan after generating intense pushback from the TV sector.

The pandemic appears to have little to do with the current glitch. Nielsen originally built the out-of-home service for use in its measures of local TV audiences, the ones who consume programming that is aired specifically for audiences in a particular region, according to a person familiar with the matter. But that product did not include homes that used broadband-delivered video only, which represents a sizable component of national TV viewership. As a result, this person said, Nielsen’s national out-of-home tabulations fail to include a significant chunk of people who ought to be counted.

The disclosure could affect perceptions of Nielsen as it vies to make headway in a competitive new era. The networks have started to put together new audience-measurement infrastructure that utilizes companies like Comscore, VideoAmp and OpenAP. Advertisers say they still want measures that have the backing of the Media Rating Council, but the networks appear to be betting that a significant portion of the audiences of the future will show up in venues like Hulu, HBOMax, Peacock or Paramount Plus. Viewership data from those services is owned by the media companies themselves — and advertisers have proven willing to embrace similar proof of audience from digital giants like Facebook or Google.

Even so, Nielsen is trying to push forward. On Tuesday, it said it had enlisted ten different media agencies and media companies — Disney and Interpublic Group’s Magna among them — to start testing its new NielsenOne measurement services, which is supposed to launch late next year and examine audiences across TV screens and new-tech ones as well. The company has also announced a plan to overhaul its commercial ratings, which have been the bedrock of negotiations between the TV networks and Madison Avenue since 2007. Under a new plan, Nielsen would track viewership of individual commercials, no matter where they appear, rather than the average audience for commercial breaks, which has been the norm.