The Impossibilities of Scale
Google & Facebook fix the wrong problems. A new ad network for advertisers has a ton of traffic, but at what cost? And three fun new tools for eComm advertisers.
Thanks for subscribing to Cache - a weekly summary of the most important stories, trends, & commentary in eCommerce marketing, curated by Rob Bettis.
Last week’s issue had a 37% open rate. Share this issue with a Salvation Army bell ringer.
Let’s talk about scale.
Servers, downloads, square footage, and production - they all scale.
Amazon order pickers, customer support, and hate speech moderation - they don’t scale.
Humans don’t scale.
They have different paradigms, value systems, and expectations. And unlike commodities, they are, at times, completely unpredictable.
If the 80/20 rule holds, then 80% of these cases are ok. But 20% are not. And the 1% (or 5%) of the cases that are complete outliers are a total disaster. And at scale, that 1% can represent hundreds or thousands of disasters.
This week we will hit on two stories about scale. Google and Facebook are both trying to solve human problems at scale. And both will potentially (read: likely) result in thousands of disasters for advertisers.
We will also unpack a study that shows how customer sentiment is changing away from personalization in the name of data privacy.
These are two conflicting headlines. Tech wants to feed machine learning more data to scale solutions to human problems. And humans want service, but not at the expense of privacy.
Overspending mysterious keywords. Last week I saw one of my client’s budget was overspending. Digging into the offending campaign, I found keywords that I never saw before. There was no record of these keywords even being added (in the change history) but here they were, overspending by 50% of my target budget.
A Beta that “changes everything”. Upset and confused, I called Google’s support for an answer. After a long back and forth, they gave me a url to access my agency’s “Auto Applied Recommendations Control Center” – a Beta that seems to give Google permission to automatically make changes to 33 different parts of an account.
Whenever I onboard a new client, I check to make sure auto-apply ad suggestions is disabled. It seems that practice has excluded my clients from this...sketchy...program.
Check to see if Google is screwing you over here.
Google is going to start showing new Recommendations for switching to broad match in Google Ads.
“If you’re using Smart Bidding, we’ll identify existing keywords that are likely to improve performance if you switch them to broad match,” Google said in a blog post Monday.
For example, says Google, the broad match keyword “women’s hats” could match to “winter headwear for women” or “women’s accessories.”
“By pairing [a broad match] keyword with Smart Bidding, you can use auction-time signals to set the right bid for each of these queries. This means that you no longer need to anticipate and manage every potential search.”
Leaving keyword to query matching up to the algorithms has been Google’s strategy for some time now as we’ve seen close variants continue to expand.
I have certainly seen accounts that were the benefactor of this strategy.
I have also seen accounts where this strategy was needlessly burning money. And although that might be the exception - a small majority of total accounts - to the individuals footing the bill, it is the only account that matters.
Facebook Updates Report Moderation Systems to Ensure Worst Case Content is Dealt With First | Social Media Today
"In the past, [Facebook's] moderators reviewed posts more or less chronologically, dealing with them in the order they were reported. Now, Facebook says it wants to make sure the most important posts are seen first, and is using machine learning to help. In the future, an amalgam of various machine learning algorithms will be used to sort this queue, prioritizing posts based on three criteria: their virality, their severity, and the likelihood they’re breaking the rules."
The process will ensure that Facebook's team of human moderators are being guided towards the worst-case reports first, optimizing their workload and limiting the spread of such, based on automated detection.
While this is likely a good step for preventing the worst Facebook has to offer, for us marketers, I suspect this means our manual ad reviews will be even longer.
While admittedly late to the programmatic game, TrafficJunky’s director of sales Graham Collie said the company was doing fine because it had exclusive access to ad inventory on Pornhub, which is one of the most popular porn sites in the U.S.
“But obviously that’s not a sustainable long-term solution, especially once we start dealing with advertisers who are much more sophisticated than the typical advertisers who will come through adult sites,” said Collie.
While the DSP is up and running, audience segments and lookalikes are still being built out. The TrafficJunky DSP can plug into other exchanges and buy inventory on non-MindGeek sites. However, Collie said the initial focus will be on monetizing MindGeek’s properties, which the company said in June sees 1.35 billion ad impressions per day in the U.S. per data.
📈 Reporting & Revenue
In its S-1, Affirm also shed more light on its financials. It revealed both increased revenue and net loss so far in 2020. Specifically, the company’s revenue surged 93 percent to $509.5 million in its fiscal year ended June 30, 2020, compared with $264 million in its previous fiscal year. At the same time, its net loss declined slightly – to $112.6 million for the fiscal year ended June 30, 2020, compared to $120.4 million in its last fiscal year.
Notably, the S-1 reveals that a significant portion of Affirm’s revenue comes from Peloton, which means that the company needs to be careful not to put all its eggs in one basket.
Specifically, Affirm states Peloton that was its top merchant partner, making up about 28% of its total revenue for the fiscal year ended June 30, 2020, and 30% of its total revenue for the three months ended September 30, 2020.
“The loss of Peloton as a merchant partner, or the loss of any other significant merchant relationships, would materially and adversely affect our business, results of operations, financial condition, and future prospects,” it wrote.
Those figures are even sharper in the U.S., with more than half (57%) of consumers spurning personalization to preserve their privacy. As marketers adjust their strategies to account for policy changes from major platforms like Apple and Google and stricter data-privacy regulations, the research makes clear that they need to establish greater trust and transparency if they want to keep consumers engaged.
A recent survey by consulting firm McKinsey & Co. found that about three out of four people have tried a new shopping method due to the coronavirus and that more than half of all consumers intend to continue using curbside pickup and grocery-delivery services after the pandemic is over. Nearly 70% of consumers surveyed intend to continue buying online for store pickup.
Amazon is blocking FedEx Ground for certain Seller Fulfilled Prime (SFP) orders during the holiday shopping season. SFP are items that display the Prime badge indicating 2-day delivery, but are fulfilled by sellers rather than by Amazon.
🛠 Tips & Tools
In theory, a native version of the notoriously resource-hungry browser might run more efficiently on Apple’s Arm-based computers. In our reviews of the MacBook Air, MacBook Pro, and Mac mini equipped with the new M1 chip, though, we found that the version of Chrome built for Intel chips already runs well on Apple’s new Macs, so hopefully this native version runs even better.
You’ll be able to pick which version of Chrome to download from the browser’s website. The Verge’s Dieter Bohn saw this prompt on his MacBook Air review unit with an M1 chip when he visited the site:
For those of you who have recently upgraded your laptop.
View the FB/IG ads from the top 1.000 Shopify Stores (that run online ads) 🦄
50,000+ Facebook & Instagram ads available for you below - for FREE.
Named "At-Risk Meeting Notifier," this new feature is a service that runs on Zoom's backend servers and works by continuously scanning public posts on social media and other public sites for Zoom meeting links.
When At-Risk Meeting Notifier finds a Zoom meeting URL, it automatically sends an email to the conference organizers with a warning that other people may be able to access their room and possibly disrupt their meeting.
🤷🏻♂️ Just For Fun
I stumbled upon this app Tuesday, and it blew my mind.
Polycam uses the LiDAR sensor on new Apple devices to enable you to quickly create color 3D scans of objects and spaces. Polycam's colorization engine is optimized to generate the highest quality color 3D scans in the least amount of time. Single room captures take only seconds to process, and all computation happens securely on your device -- no internet connection is required.
Watch it in action in this video:
Questions, comments, inquiries? I’d love to hear from you! Email firstname.lastname@example.org.