Beyond Time Spent and Engagement

You only have so much attention. When you consume media, roughly 27% to 30% of your attention can be directly monetized, and there’s perhaps a tolerance for another 15% that can be wrung out with product placement. As a result, Time Spent is an attractive metric for those who create and monetize attention.

For example, if you can attract 1,000,000 hours of attention, then you can monetize 300,000 to 450,000 hours of it. In theory, the amount you can charge for your attention depends on the value advertisers place on the audiences’ attention. And that depends on who they think you are, how susceptible you are, how much you spend in the relevant category, how causal the purchase decision is in the context, and perhaps distressingly, questions of prestige and peer status. There’s a lot of hidden knowledge, a lot of badly out of date information, and quite a few poorly documented heuristics in all of this. As a result, there are a lot of different proxy measures for attention. It affects decisions.

People in their twenties and thirties consume a lot and are forming habits. People in their forties and fifties aren’t spending as much and have either formed habits or had their habits formed for them. If I’m selling beer, aside from the product quality attributes, how many times do I need to air my spot on the three shows 18-22 year olds watch, which events do I need to sponsor, and where I can insert myself on Youtube and the tok? How many hours do I need to produce in order to move a single point in market share? How much do I have to spend to get a single power-drinker? How many Gross Ratings Points do I need to move a thousand hectolitres of beer?

Time spent is a pretty useful metric. In 2022, there are 2.5 million Canadians right in that sweetspot of youth and susceptibility, consuming roughly 34 hours of commercial media a week — so that’s 85,000,000 hours of attention, of which, as much as 35,000,000 hours are available for sweet monetization.

Doesn’t sound too nice when I express it this way, does it?

Mass digitization and attention aggregators (Google, YouTube, Meta et al, Snap, Twitter, Tok) have played around with the Time Spent metric. Meta infamously so. YouTube created incentives for content creators to produce 10 minute videos. TikTok is unreasonably effective at attention hacking. So, who do you trust to manage your attention?

Meta made it ridiculously easy for anyone to identify and target particularly susceptible individuals. I’m not talking about just Cambridge Analytica [1]. I’m talking about the fact that Facebook made it so easy that even Jared Kushner could figure out how to target very susceptible segments of the population with hyperfocused messaging. If Jared could figure it out, then perhaps as many as half of all Americans 14-80 could figure it out. I don’t feel good writing about it, so I won’t. Apple, in theory, put a stop to it on their platform and advertisers are left with the Android/Google ecosystem for that kind of targeting, to varying degrees of sadness and gladness. I’m not convinced that it isn’t still happening on Apple devices. It’s just that the tools aren’t as accessible to the masses.

Engagement is a malleable cloud of concepts, and can be effectively structured against any dependent variable. Legacy media didn’t quite get there, frequently falling back to Time Spent metrics to understand digital. Those in mainstream digital analytics who understood how Time Spent was calculated rarely pursued it as a dependent variable. Product analytics pushed the Engagement paradigm to its inevitable conclusion. How often do you engage with the platform by logging in, amplifying something that somebody else has created, creating something yourself, editing something that you said, doomscrolling, and swiping? One of the most nefarious use of engagement came from the casino/lootbox app world. They’re looking for signals of addiction, its precursors, and, indicators that somebody was falling out of it.

A key fork in the decision tree around engagement was whether you would accept that there was a dependent variable in the first place. If you could, the discussion shifted to the criterion of which independent variables should be included. If you couldn’t or just wouldn’t, the discussion was either finished or it degenerated into edge case salience. Those could be fun but never generalized enough to be useful.

So much of this language is focused on how attention as a product is created and monetized. The audience, to the extent that anybody thinks of them, is assumed. Globalization, digitization, and vertical integration have combined to create the media, both mass, conspiratorial, and individual, that we have come to consume.

Who pays?

The audience pays for everything.

They pay for the content with their attention to ads, and then again when they pay the markup in the price of the things they buy. Nearly 350 billion dollars were spent in advertising the US last year. Who ultimately paid for that? American consumers and taxpayers paid for that! The consumer and taxpayer makes it all go.

The audience also decides which content gets created and which content does not through their attention. They vote not only with their wallets, but through their eyeballs, ears, and hands. The reason why there is so much hammy content on LinkedIn is because there is a large audience on LinkedIn that likes ham. The reason why Dancing With The Stars exists is because there is a large enough audience that likes Dancing With The Stars. The reason why broadcasts audiences like predictable broadcast television stories (the guest star always did it!) and why subscription audiences like risky content is because that’s what they like. If people stop paying attention, then new content like it is no longer created.

So, Time Spent and Engagement are proxies for paying attention, which in turn, combined with psycho-demographics, establishes a proxy for clickthrough, purchases, repurchase, loyalty, retention, market share, margin and shareholder value. If you’re just starting out, you can think of this as a funnel, if you’re a little bit beyond and understand reinforcement and hidden effects, there is no limit to how you can model it out. Easy, breezy, beautiful, you can cover it all, girl.

What’s Beyond?

If you had a positive reaction to the idea that the audience pays for everything, then what’s beyond are constructive models about the audience. Your mind is cleared for that kind of work.

If you had a negative reaction to the idea that the audience pays for everything, then there probably isn’t much beyond, is there? The main way that the business behind the platforms think about the audience is that the advertiser pays the bills, so the advertiser calls the shots. This perspective can permeate all the way through to what content creators create. Attention to this tension, between what the creators want to create and what the advertisers will tolerate, flares periodically. It’s nothing new.

But what if we looked at it from the audiences perspective? What would happen then?

Many reading this space will scoff and retort, “Who cares?”, and I hear you. Paying with attention is different than paying with money, and we’re living in an era of reductio ad aurem. Fair enough. Focusing on predictive modelling is valuable in that context, and the world is filled with susceptibility to targeted messaging. Everybody shouting wake up sheeple is sheeple. Lambs to the slaughter.

But what if you looked at it from your perspective as a member of the audience?

How do you want to be treated?

Is it in your strategic interest to be addicted to a media platform? Is it in your strategic interest to be depressed, to suffer by comparing yourself to others all the time, to represent yourself to the world in a way that is unauthentic and hollow? Is it in your strategic interest to be manipulated into enragement, induced to want things that are in your interest, or, to be mislead?

A Thought Experiment

Let’s say that you founded a corporation with the sole purpose of developing the technology that is responsible for selecting the media that you consume.

How would you direct that corporation?

Would you hire a person and give them instructions on what to sort of media to select? Would you develop an algorithm and train it on what to select? In either instance, how would you supply the training data? What is your optimization objective? What signals from the environment would be valuable in making the decision about what to bring to your attention, and what not to bring to your attention?

How would you optimize your experience of media if you were directing the work?

Would you want to be angry all the time? Would you want to be happy all the time — with the illusion that everything is great? How much do you value the time you spend in a happy state? In a concerned state? How much do you value knowing what’s going on?

What’s your real return for the time spend consuming media?

Your answer to these questions are going to be different from others. Some people optimize their experience for the consumption of celebrity. Likewise, others optimize their attention for outcomes they can’t do anything about even if they were informed about them. Many consume financial news. Others, financial entertainment. Others still, financial celebrity entertainment. Sports fans consume sports news, sports entertainment, and sports celebrity entertainment alike. The same goes for people who consume movies and those that consume film. Comic books, romance novels, poetry, linear television, the varieties of NCIS, CSI, Law and Order, FBI and all the departments in Chicago, their mysteries, and some, even, consume comedy. Allegedly journalism still produces substantive investigative pieces in between copying and pasting press releases and rage farming. The mix and the match of your slice of the 85,000,000 hours a week is going to vary.

And who pays for the process of selecting which media to consume and which to ignore?

You do. It’s always been you. Since the beginning. It’s always been so.

What would be the likely result of you directing what you saw?

The Illusion of Control

What leapt out for me in doing the thought experiment was just how little autonomy I have on what I consume.

First, at the most basic level, content creators ultimately decide my consumptive experiences. At the time of writing, content creators are mostly human. They make human decisions about their human craft, and pretty much shape what is created. Some of what I see is generated by DALLE and other applications of stable diffusion. One could argue that narrow machine is a tool in much the same way as a paintbrush, a pen or a chisel.

Centralized creative control has been eroding for several decades, thanks to the co-evolution of the integrated circuit, the HTTP protocol, and perhaps compounded by ceaseless vertical integration with globalized corporate consolidation. We’re creating more media than ever. And creation is linked to belief. Belief has always been subjective and in many ways a social virtual good. What happens when people choose to believe in deep fakes even after discovering that they are deep fakes? We have some evidence from the field already. Isn’t this already happening with the consumption of participatory conspiracy experiences? Isn’t the process of the creation of the Qanon conspiracy megaplex really a co-created, mass experience? [2]. In this way, your beliefs are shaped by what content creators, and increasingly yourself, choose to create, potentially independent of any objective truth and the updating of any priors.

Second, there is bias in the process of selection. Content creators and regulators talk about the concept of discoverability. The desperate belief in Canadian circles is that only if Canadian content was discoverable by audiences, then audiences would inevitably consume them. That idea isn’t exclusive to Canadians. It causes much hand wringing by tortured artistic communities in every corner of the world. (If only we could jam our art down their throats much like advertisers cram their messages!) Disturbingly, the selection of what you pay attention to, and what you don’t, are tied up in other forms of bias. Identity effects exist and are insidious [3]. So, let’s say that you direct your corporation to group content by genre, and Canadian content is a genre. If your latent attitude towards CanCon is that it is inferior to American, French, British, Japanese, or any other country, then your propensity to discover something particularly awesome about it is much lower than if the CanCon was grouped in some other way. For instance, if you assign particular importance to a Cardassian or a Kardashian, would you want CanCon, if it was likely to be good, to be endorsed by one of them? Sometimes, the signal you emit by virtue of consuming media creates more of that media.

Third, there is the illusion of control over data emission and machine learning over those emissions. As you consume digital media, you leave signals behind. The quality of those signals depend on the quality of the instrumentation these concerns deploy. In general, Google, Netflix, Disney, Amazon, Apple, Meta, Microsoft all actively watches you watching. Advertisers, and their agents, watch what you watch, watch what you read, and, sometimes, roughly, watch what you listen. Your corporation wouldn’t be able to stop emitting data about consumption. But it could reduce the value assigned to that data. Would it be in your interest to reduce the value of the data you emit?

Possibly.

If it becomes in the commercial interest of a firm like Google, Meta, or Bell Media to omit information sources in response to your media consumption habits, it stands to reason that they would make a business decision to omit that information source. Publishers omit information all the time: they just don’t cover something. Editors do too: they just bury it on page ten. Aggregators can dial things up and down as well. Perhaps then, it would be in your interest to confuse information about you? How might your corporation achieve this?

Maybe you direct the development of technology that is continuously scraping information sources? It loads and saves, in PDF format, not just hundreds of news articles that it thinks that you might want to read, but thousands of other articles that it’s confident that you have no interest in. Maybe it simultaneously streams a show on Netflix, Apple, and Disney, while you consume cat clips on YouTube? Maybe the corporation erects a wall anonymity around your profile by making it more difficult for these firms to build a profile? Maybe the corporation procures multiple credit cards, on different accounts, using different constructed identities, so that you can’t be profiled through the credit rating agencies and payment processors?

Such lengths one would have to go through to optimize for control!

It’s through a paradigm centered on your optimization objective, your dependent variable, that we can get to which metrics are beyond Time Spent and Engagement.

What’s Beyond Time Spent and Engagement

What you see depends on where you sit. What content creators, aggregator platforms, advertisers, and marketers see is Time Spent and Engagement. But what do you see from where you sit?

Hypothetically, if you were able to defragment the platforms you consume and centralize those decisions from your point of view, what would you optimize for? Would you emit signals that you’re a discount shopper while consuming prestige film? Would you want to scoop up every bit of drama from the Real Housewife world while blotting out the Cardassians entirely and let the ad aggregators know that you’re planning a trip to Brasil? What experience do you want?

After all, it’s your attention.

[1] To this day it’s still distasteful to discuss the model they used to predict so much about so many using so little.

[2] This paragraph is especially dense with callbacks to May and September 2022.

[3] Taylor, S.J., Muchnik, L., Kumar, M. et al. (2022) Identity effects in social media. Nat Hum Behav.