What you need to know:

• Facebook has pulled back from culling 28-day attribution and moving to seven days.
• Its proposal, perceived as a response to incoming app tracking changes from Apple, had sparked confusion among agencies and advertisers.
• Some believe the move would have made it harder to attribute actions to spend, with advertisers missing key conversion factors as a result.
• Others, such as IAG's Willem Paling, suggest Facebook ads have little impact on longer term decision-making.
• Mutiny Managing Partner, Henry Innis, thinks current attribution tools are inherently flawed. He says there are ways to create a far sharper picture.

Fate amenable to change

Last month Facebook told businesses it would be ditching the option for 28-day attribution and moving it to seven days from October.

An update on Facebook for Business (now moved) cited ‘digital privacy initiatives’, perceived by industry to refer to Apple’s cookie and (now also delayed) ID for Advertiser changes within the IOS 14 update, which launched mid-September. Facebook’s changes were due to kick in on 12 October.

But last week it appeared Facebook had pulled back, indicating the shift, which it then referred to as a ‘small test’, would be shunted to January at the earliest.

Yesterday, it told Mi3 the change would not be going ahead at all. Per a Facebook spokesperson: “Based on feedback from advertisers, we have decided not to move forward with a small test that would change the default attribution window for some advertisers.”

But even flagging the changes has sparked broader debate about the value – or otherwise – of attribution models, and whether brand marketers obsessed with metrics might actually be mining fool’s gold.

Gunboat diplomat

The “digital privacy initiatives” originally cited by Facebook were widely perceived to be those being made by Apple and later, Google.

As part of its iOS 14 update, Apple revealed it would remove cookies after seven days on its Safari web browser. But Apple also flagged major changes to its Identifier for Advertisers (IDFA) as part of the update. IDFA is used across the ad ecosystem – including by Facebook – within targeting and attribution systems. But last month Apple decided not to make that change until next year after a lot of noise from the ad industry – especially Facebook – amid fears that 80% of targeted advertising on iPhones would be wiped out.

Digital marketers believe these changes, along with Google’s culling of third party cookies in Chrome, set for 2022, were behind the proposed shifts to Facebook’s attribution window.

“If Apple blocks anything more than seven days, then advertisers choosing 28-day windows would miss conversions from those devices,” digital and social media consultant Lucio Ribeiro told Mi3.

Ribeiro says the attribution window change, if it ever now eventuates, will likely only impact “advanced Facebook marketers” and those operating large-scale campaigns. But it would also affect automated systems which determine Facebook ad spend based on specified events.

Arbitrary

Ex-Showpo CMO and consultant Mark Baartse suggests Facebook may have lost out on revenue if marketers perceived it to be less effective as a result of changes to attribution windows.

But he says marketers should always question what they are being told, from Facebook or anyone else.

“All attribution is just an estimate and far from perfect”, he suggests. Whether 28 days or seven, “they are arbitrary numbers”, suggests Baartse.

“If someone makes a purchase after eight days, they still made the purchase. It's just reported differently. So, customer behaviour doesn't change, simply the way we report on it.”

Publicis Chief Product Officer, Jason Tonelli, says major brands have developed their own internal reporting and attribution methods, with most already operating within a seven-day window. Facebook’s mooted changes, he suggests, would not have made much difference – except maybe for big ticket, long consideration buys. But even then, Tonelli says, “most agencies can use external econometrics to close that loop anyway, so it shouldn’t be an overly dramatic issue.”

Consultant Mark Baartse agrees.

“A good number of auto purchases [for example] take longer than 28 days regardless, so the problem is there already, maybe just amplified a touch.”

Willem Paling, Director of Customer & Growth Analytics for IAG, thinks long-term campaigns might suffer if Facebook again changes its position, but says major brands are not usually using Facebook for anything other than short-term sales.

“Facebook lends itself more to the performance and short-term sales end of the funnel anyway, despite making some attempts to break into the brand space,” Paling says.

“The impact on investment is unlikely to be based on long-term campaigns, as generally marketers who have larger budgets to invest in brand lean more towards traditional above the line channels such as TV, radio, etc."

Me, I’m counting

Mutiny Managing Partner, Henry Innis, suggests any changes Facebook may or may not make are a sideshow. He says industry has been “hoodwinked” by the concept of attribution, which he says “on paper” should work - but is usually let down by ineffective and unmemorable digital advertising.

“With attribution, advertisers will argue or fight over the role they played in the sales conversion - but you could have seen that same message across 15 different sources, with no way to rank the impact they had on the purchase,” Innis says.

“Without that, you’re left with ineffective one-to-one user tracking, which is why good marketers will tell you it’s smarter to look at the overall impact on spending through regression or neural network modelling.”

Innis suggests that the on-off move by Facebook to shorten the look-back window to seven days may have actually represented a positive step. Either way, he says the issue can’t be solved by publishers or those selling ad inventory.

Marketers “obsessing” over a seven or 28-day window “tells me they are still in the wrong ballpark when it comes to this issue,” Innis says.

“We should be asking ourselves why we are talking about attribution modelling at all; it’s bad mathematics. We should be looking at moving towards ranking and contribution models.”