The Only Winning Move is Not to Play

Follow us: Mastodon, TikTok, Threads, Instagram • more
Follow Tod: Mastodon, LinkedIn, TikTok, Pixelfed • more 

In This Issue:

🐦 Are Meta ad campaigns doing poorly for you this week? You’re not alone.

💻 Microsoft expands Audience Ads with AI bid strategies and new markets.

📦 Amazon urges sellers to prep early for Black Friday inventory.

📱 Elon Musk, surprisingly, reveals huge decline in X platform usage

🔑 Chatbot AI may leak hardcoded API keys.

🤖 TikTok introduces tools to label AI-generated content.

💰 Influencer marketing reaches $29 billion in 2022

📈 Publisher ad revenue sees a Q2 rebound after a Q1 slump.

🌫 Retailers adopt fog systems to deter theft and protect stores.

Meta Ads Are “Broken” This Week: Here’s a Rundown

If you’ve noticed your Meta ads have been slumping for the last few days, this may be the reason why.

DTC media buyer David Herrmann posted this afternoon that the Meta ad platform has been pretty broken since last Wednesday.

He says his campaigns still have ads stuck in review — this despite him trying the duplicate-and-try-again trick, and any new ads he’s put up since Wednesday has been performing terribly.

So, in case you’ve been seeing similar drops in performance or hikes in prices, this may be why. Some advertisers said on David’s post that they’re pulling back on Meta spend by as much as 50% while they ride it out.

Microsoft Audience Ads Get Wider Rollout

Some updates to Microsoft’s ad platform — the company today starting the rollout of two new automated bid strategies for Audience Ads.

• Maximize conversion

• Target CPA

This is launching to a small test group for now, and should be to all accounts by the end of the month.

What Are Audience Ads?

Audience ads are Microsoft’s display ads. They’re served on third-party sites as well as some of Microsoft’s own properties like MSN, Start, and Outlook. This is different than the keyword-based ads people see when they search on Bing.

Audience ads are generally considered inexpensive and good for brand awareness, but not so good for clicks and conversions. They’re more of a top-funnel placement.

More Countries

Microsoft also announced its Audience ads are now available in 58 new markets. That brings it to 187 countries in total.

Amazon Inventory Deadline Approaching for Black Friday

Time is ticking. And if you sell products through Amazon’s fulfilment program, the company is warning you have five weeks left to get your stock to them to make sure you’re ready for Black Friday.

The deadline is October 26. The company says its fulfillment center employees will spend the time after moving product around to its regional hubs to make sure things are close to expected consumer demand.

If you need more capacity, you can increase your limit with Amazon’s capacity management system.

Higher Fees Coming

Also coming in October: Higher fees for sellers using Amazon’s fulfillment services.

It’s kind of surge pricing — the extra fees are for the peak season and run from Oct. 15 to Jan. 14.

One of Amazon’s few competitors, Walmart, said it won’t charge any peak season storage fees for sellers — but there is a catch: you have to have your inventory in by Oct. 1.

Did Elon Musk Just Confirm a Massive Drop in X Usage?

Elon Musk may have just inadvertently revealed that usage on X has plummeted.

This is something independent analysts have noted for a while now. X has tried to counter it by inventing metrics like “device user seconds” — a metric nobody else in the industry uses — and saying that those DUS’s are at an all-time high.

But then, at an event yesterday, he said this:

Wait… 100 to 200 million tweets?

That’s actually not something to brag about, considering that Twitter used to regularly report 500 million posts per day. They said it as far back as 2013, and continued reporting that number in the years that followed.

In fact, even X’s new management, at one time, said they were seeing 500 million tweets:

There are factors in this number we don’t know — factors X has so far been silent on: How many bots made some of these posts? How many are retweets? For that matter, as SocialMediaToday.com noted in its coverage of this, how much of this is due to the overall industry decline in posting on social media.

But still, that’s a big drop in activity.

Also related: In the same meeting, Elon Musk told the audience that he’s considering charging people for access to the site. Not just to access Premium services — access to any of it.

APIs Are Being Revealed by AI Chatbots

There’s a reason why security experts say to be very cautious about what you put into an AI chatbot like ChatGPT or Bard.

Some people have put confidential source code into it; others have given them access to large pools of company data in the hopes that those chatbots will learn more about the company and be able to provide more accurate responses.

But a group of researchers has found that swimming around in that sea of data are API keys — and that could have a devastating effect on a company if those get out.

What are API keys?

API keys are short string of data — like 40 random characters — that let one service authenticate an account.

For instance, if you want to connect your email marketing platform up with your legacy CRM, you might use an API key. You’d generate a key from your CRM so that when your email system wants access, it provides this secret code, and the two systems can talk. This also makes it easy to revoke access — you just change or delete the key.

But keys are sometimes found in big pools of data, and if that pool is handed over to a AI engine, it’s possible somebody else might be able to ask that engine for your API keys and be given instant access to the mission-critical platforms and systems you use.

What the study found

The researchers have published a study called "Do Not Give Away My Secrets: Uncovering the Privacy Issue of Neural Code Completion Tools."

They built some software that specifically looks for the way API Keys and Access Tokens are formed — how many characters, what sequences tend to be in them, and so on. And they set their software off line-by-line, looking for strings that matched.

Their software identified 736 matching strings — 129 had keys or tokens in them.

For ethical reasons, the researchers say they didn’t check to see if keys to systems with private information in them worked, but they did test others and found at least two keys operational — both working on the Stripe payment platform.

Stripe accounts actually have two keys — a test key and a live key. The researchers found the test key, so they weren’t able to see or make any changes to actual personal data.

Still, though, a lesson for us all — be careful what you say to an AI machine. As the WOPR computer in Wargames told us all: “The only winning move is not to play.”

Meet TikToks’s New “This Is AI” Labels

TikTok today announced it will launch a new way for brands and creators to label AI-generated content — some by manual labelling, some automatically.

When the app detects AI-generated content, it will pop up a prompt in the upload flow asking users to add the label.

To users, this new label will show as a message below the username reading “Creator labeled as AI-generated.”

Competing policies

But usage of the tool is entirely voluntary. TikTok said it won’t require past videos to be manually labelled, nor will it apply any kind of penalty or strike against accounts that choose to not add any kind of disclosure label.

That seems to be at adds with its recently updated policy that gives it the right to take down realistic AI-generated images if the user hasn’t disclosed the video is fake.

The company says eventually, it plans to figure out which videos have been auto-generated and will automatically apply this label to videos itself. They’re still working on that AI-detection software.

Instagram is also working on similar tech that will label content created or edited with AI. And the EU is pressuring platforms to label AI content as well.

The OG AI

In a way, it’s all a little peculiar given that much of TikTok’s massive success is due to AI-based filters. The “Bold Glamour” filter was so accurate it was the first to be considered AI-based, rather than just a mesh applied over a face using augmented reality. At the time, TikTok refused to respond to press inquiries asking if the product was, in fact, AI-based.

To its credit, TikTok says it will rename any filter that uses AI so that it’s more clear. Though, to be realistic here, I don’t think any consumers know or care whether the tech that’s giving them bigger lips is AR or AI.

The new AI label functionality has started to roll out today, and will take a few days to get to all accounts.

Two Key Marketing Research Reports

A couple of interesting reports in the digital marketing space came out this week.

Influencer Marketing Grows Up

First, research firm PQ Media says media spend on influencer marketing went up nearly 22 last year to $29 billion and is expected to go up 17% this year.

Most of that, of course, coming from America — but, as these stats often do, as the market matures, faster growth will start coming in from other countries. The researchers say the U.S. accounted for 76% of spend in their study, but that will drop to around 68% within four years.

• The second fastest-growing market was India, followed by Japan.

• The fastest growing account size was the microinfluencer — accounts between 10,000 and 100,000 followers.

The study found that marketers using influencers are focusing less on metrics such as Likes, and are now more interested in click-throughs.

Publisher Ad Sales Comeback

Second, more signs that the digital advertising space is on a comeback.

Boostr’s Media Ad Sales Trend Report and Operative’s Benchmarking Report both show numbers on the rise for publishers — news and entertainment sites that host ads.

While that’s certainly good news, it was noted that other metrics are still lagging — notably deal volume and deal cycle length.

A Boostr executive was quoted in the report saying:

The Retail Fog of War

I guess you could call this marketing technology. It’s not online, and there’s no pixel to install.

Instead, it’s technology used by a growing number of bricks-and-mortar stores: Fog. (Like, literal fog that comes shooting out a fog machine.)

It’s called the Density Security Fogger and it’s triggered by a store employee who sees a shoplifter. Once they push the button, the entire store fills up with fog very quickly making it more or less impossible to see anything — specifically, where the exit is and how to get there.

The company selling it says it’s an antidote to an increasing number of smash-and-grab robberies.

The fog is the same kind used in concerts or theatres and takes about an hour to dissipate.

The company says it has 3,500 installations in the US so far, and a deal with a cell phone company is expected to add another 6,500.

So, next time you ask your team to install a new cloud security system, better double-check what they come back with.

Some links in this newsletter may provide affiliate revenue to us.

This newsletter is powered by Beehiiv.