How Facebook is trying to recover the money it lost during the outage
indiehackers.com
indiehackers.com
As someone who spends five figures daily on FB, I saw no such drop in performance -- nor have other advertisers I've been in touch with.
Performance on FB (especially at low volumes) is highly volatile on a day-over-day basis. So, at any time, there will always be a subset of advertisers who think "CPA's are skyrocketing". And, it's easy to draw the conclusion that the pattern is platform-wide.
The "accelerated delivery" that Facebook quoted in its response only applied to the remainder of the day that the outage took place -- not future days as the article suggests.
Do you actually need to spend that much..?
Delivery for the day dropped, but performance as measured through CPA (or ROAS) remained stable.
> Do you actually need to spend that much..?
Ultimately, we're trying to maximize profit. So, we advertise until the cost per incremental conversion is greater than a customer's ltv.
But how can your engagement from clients be "the same", despite half the availability?
Does this imply that all else perfect andavailable,every day, you could have a volatile swing of up to 50 percent engagement?
If so, that seems really really silly to drop 10k+ on daily, if on half of all days you don't NEED the platform for half the day...what am I missing?
Essentially, it's 'engagement per dollar', not 'total engagement'
CPA should be affected, seems like FB is giving an artificial measurement. Reminds me of the Big Short.
If you've set a campaign-level budget, Facebook's system paces delivery to spread that budget from the start date to the end date. It automatically adjusts pacing to fluctuations in delivery, this 6 hour outage is just a drastic version of that, where there was zero delivery across the board vs. more localized fluctuations in consumer behavior that usually impact pacing for individual advertisers.
So it has to accelerate delivery for the remainder of the campaign to compensate, and since it's doing that for all ads with this type of budgeting configuration it's resulting in an auction grab for available impressions and temporarily bumping up CPMs as a result.
Which isn't necessarily a positive for advertisers, but sounds like their system doing what it normally does, and just having some unanticipated second order impacts when faced with an extreme event (compensating for global lack of delivery for 25% of an entire day).
If your budgeting/bidding/optimization configuration doesn't require a specific spend within a specific timeframe, you're not going to see this sort of acceleration. But you will see higher CPMs due to the increased auction activity from others having their pacing accelerated. And they weren't collecting FB pixel data for 6 hours either, so conversion activity will also be underreported and show higher CPAs. But CPA-based optimization configurations are still in place and doing what they normally do, and frequency caps if you use them are still limiting how many times your ad will show to the same person. So even those seeing accelerated delivery and not wanting to could adjust their ads' settings to avoid it.
> The "accelerated delivery" that Facebook quoted in its response only applied to the remainder of the day that the outage took place -- not future days as the article suggests.
It does actually impact future days, just to a far lesser extent than day-of. Pacing jacked up the day of to attempt to compensate for expected pacing that day, but for those that didn't achieve it they'll continued seeing accelerated delivery beyond that to hit the spend target by end of campaign. Plus second order effects as their optimization see-saws around until things settle back down.
Then Facebook has a 6 hour outage. Facebook can then show more ads to users per time on Facebook, to get to the $100 per day. Or it can only spend $75 that day.
I imagine some advertisers will prefer the first option, some the second. Facebook's machine learning algorithms will do something.
But I don't see the massively malicious intent here. Most likely this isn't even a human decision. The algorithms just know then have to spend $100 per day.
Fine, if there really are 50% less conversions, that's an issue, but frankly 6 hours won't really matter.
You buried the lede with things we already know for some reason.
Two things:
- It's 6 hours and 60 million dollars to them, which is absolutely nothing, but they risked harming their customers (advertisers) over it. This was Facebook's screw-up, and now they're passing it on to their advertisers. When you view advertisers as captured, you worry less about their experience and attempt to extract as much value as possible.
- It's systemic. Someone in another thread mentioned there are good people at Facebook, and they can't reconcile the behavior exposed in internal documents with those good people. This is a great example. When your performance is tied to metrics like advertiser spend, even when you had the largest outage in your recent history, you'll immediately hop on afterwards to direct your team to try and extract the most money out of advertisers possible to make up for it. These individual incentives add up to form the predatory company that we see externally.
If you were an ad customer during this period and you feel you weren't getting your expected value, ask for a credit?
If you weren't - why are you so upset? Are you just looking for some excuse to feel angry at Facebook?
Maybe? I'm not in the ads space. I don't know if there are automated systems built in with "you now only have 16 hours to display the ads, display them even when they're less relevant".
> If you were an ad customer during this period and you feel you weren't getting your expected value, ask for a credit?
Why do customers need to do this when Facebook messed up, and had to learn about this, and look into performance, from a source other than Facebook?
> If you weren't - why are you so upset? Are you just looking for some excuse to feel angry at Facebook?
Y u mad bro? hasn't worked in quite a while.
Have you bought anything from anyone ever...?
That said, it seems wrong to use accelerated delivery for anybody using the Facebook pixel for conversion optimization.
Both of these will limit the rate at which an ad spends money. If it doesn't get shown for a long period of time, both of these factors get reduced, and the ad can get selected more aggressively, particularly if you are a higher bidder than the ads that would have got chosen in your absence.
In the limited experience I have in this arena, it does improve performance. Hitting someone a second time about t+2weeks had the biggest bump, and then the next biggest another month later at t+6weeks.
I don't know if it's universally true, and the optimization may vary depending on the context and content, but pacing can definitely make a difference for some.
But attention is a limited resource. If you throw more ads at me, I'm not necessarily going to respond to more of them. The more ads there are, the more it becomes background noise.
To the individual user, they can’t tell anything different between these two types of campaigns. It just makes the advertiser feel better that their money is lasting longer.
I can’t find anything that indicates individuals are seeing more ads than usual as a result of the outage, besides the anecdote of the one agency in the article (but maybe the lower conversion rates can be explained by some other factors)
Of course, the type of data that would be really useful here (detailed ad performance metrics by the hour, across all of Facebook) is the type fb would never willingly publish.
But Facebook is actually trying to help its costumers recover the money lost during the outage.
US$ 60 million? Pff. Just ask every executive to skip one executive jet travel per week and they recover that in a month or less.
It would be rational for advertisers to pull down their spend to avoid getting hit with devalued impressions. Ironically, if enough folks did, that could cause a panic, which would make things worse for FB.
"We can also confirm that ads did not deliver during the time the systems were offline, and advertisers were not and will not be billed for ads during the outage. Ads have now resumed delivery and some advertisers may see accelerated delivery as our services recover from the outage. We encourage advertisers to review your bids and budgets to ensure they accurately reflect your marketing goals."
One way of looking at this is that Facebook is notifying its advertisers that the outage will affect how advertising is delivered in the following days, and if advertisers want something else to happen, they will have to make the changes themselves. Facebook could reasonably claim that it does not know how any given advertiser would prefer the situation to be handled.
Of course, if it is expensive or inordinately difficult to make such adjustments, then Facebook would come across as dissembling (and not for the first time.)
I suspect that it has less to do with actual profits and more to do with some wanting to keep their quarterly ad impression numbers up
Yea, that's FBs new business model :D
It is legit hard to find ads which are not a scam in some way.
I know the example mentioned is facebook, but I feel like a similar development is going on with other ad networks as well. Even on YouTube around 40-50% of all ads I get are by some guy trying to sell me on his dropshipping, forex trading, crypto or tax evasion course.
Did Veruca Salt grow up and get a job at Facebook?