Intro
How Apple Intelligence can win
Apple Intelligence made waves when launched in June but interest has remained cooly skeptical since. Now that it’s officially live in OS 18.1, popular sentiment is it’s still “half-baked". It's not looking like an immediate hit.
However, Apple's potential is still huge. As Ben Thompson repeatedly points out, Apple’s ownership of end-user experience puts them in a position to effectively commoditize the winners of the large language model (LLM) wars. Because they have such a powerful grip on users, effectively, the AI race is theirs to lose. Sam Altman underscores this in his view where future AI “serve(s) as autonomous personal assistants who carry out specific tasks on our behalf like coordinating medical care on your behalf.” This reads like a super Siri.
So the question is, how would Apple Intelligence actually win the hearts and minds of users? It will come down to two main hurdles, achieved by the two parties that service end users:
It doesn’t need to be perfect, but Apple Intelligence will need to do some things much better than the alternative. The best example here is today’s most popular LLM, ChatGPT. It's far from perfect, unreliable, and will often generate nonsense, but it stands alone in the ability to provide aggregated, summarized information quickly. Much like ChatGPT, users will overlook Apple Intelligence flaws for narrower stand-out capabilities.
App developers and brands will need to buy in. Most app developers haven’t yet realized that the ultimate success (or failure) of Apple Intelligence largely rests in their hands. Initially, Apple Intelligence functionality is limited to helping with core features of the phone. The true value will be unlocked-- much like the earliest versions of the iPhone-- when it enables connection to third-party apps. Truly ground-breaking use cases will be achieved by the actions and content surfaced to it by app developers.
With these two foundational shifts, Apple Intelligence could become our defacto connection to artificial intelligence, both as an on-device small-language model and as the connective tissue to the continually advancing large language models in the cloud. If it hits, it will create a gold rush of opportunity for apps to fundamentally rewrite how users interact with apps.. Until then, we'll be stuck making custom emojis.
Industry Buzz
The case for augmented reality is bullish: but how far away is it?
Well-read consensus is centering on augmented reality as the clear “winning” outcome for extended reality (xR). The argument –- underlined by rave reviews of Meta’s Orion prototype –- is that real-world experience will always trump (and simplify) interaction when compared to pass-through video, used by today’s VR platforms.
But how big is this market today? Surprisingly, big. The consumer market is already generating $40B this year which is almost half of the mobile gaming market and reaches an astounding 52% of the market. This number is slightly misleading as it includes social media filters and “try before you buy” e-commerce functions; separated AR & VR hardware represents only 4% market penetration or 34 million cumulative devices – so it can hardly be considered mainstream. That said, Ray-Ban Meta glasses have quietly become the top-selling product in stores.
That’s not to undersell the bear case, which is easy to see when perusing the graveyard of “challenged” devices. The infamous Magic Leap, founded in 2010 raised $4.5 billion dollars only to recently become a technology licensing firm. Even the best recognized device on market-- Apple’s Vision Pro—was reported to have cost $20B to develop, and is whispered to be ceasing production by end of year. In other words, having sold around fewer than half a million devices, Apple has lost an eye-watering $36,500 per device.
So what’s holding back mainstream adoption?
- The hardware: In summary, today’s technology is too clunky, too expensive, achieves poor battery life, and isn’t ready for prime time. For details, read Hugo Barra’s, or Mathew Ball’s very in-depth analysis.
- The software: Despite 34% of game developers building for the Quest Store, the common perception – especially for the Apple Vision Pro—is developers haven’t flocked to the platforms, leaving native content sparse and repurposed 2D apps mediocre.
- The killer use-case: as the result of both hardware and software shortcomings, the real problem is today’s xR experience isn’t demonstratively better than alternatives. It’s not more useful than a phone, not more functional that a PC. Put another way, it'll "need to become something people want before they can afford it"(and for the record I don't count a virtual monitor as a killer use case).
So where does that leave us? Meta’s ability to produce Orion -- at today’s best estimate- is two to five years out. Looking at the iPhone’s path it took another one or two years for software and consumers to catch up, bringing us to three-seven years from mainstream- which actually aligns with Hugo Barra’s timetable. The better half of a decade may seem like an eternity in technology, but if AR represents the next compute platform, it’ll be the beginning of decades of huge growth.
Privacy & Security
In search of incrementality
One of the most notable outcomes in the last four years of privacy-centric data clampdowns is the advent and release of ML-driven advertising products released by the large advertising platforms. While these data munching ad-targeting systems are proving hyper-effective, they function much like black boxes, where reporting is aggregated and highly obfuscated. Advantage Shopping Campaigns and Performance Max may work, but the advertiser will never unravel why. And this can create a problem. As Eric Seufert points out:
“the platform needs only to achieve the budget-level performance standard established by the advertiser — not to maximize ROAS for every targeting segment and placement it experiments with.” paywall
Put another way: The ad platform won’t stop while it’s ahead. It’s incentivized to waste money, while still hitting targets. And advertisers can't do anything about it. This is what's behind Walgreen’s media chief calling these metrics "a lie”.
The same increasing privacy restrictions have fundamentally changed the state of measurement.. Marketers are struggling to eke out more insights from ever-decreasing sources of data. At Branch we’ve been busy developing new, sophisticated products to help increase measurement signal in a privacy-centric manner.
Early testing on this new technology shows very positive results – over 18% increase in attributed installs on IOS – but while it will help you win back some of the loss in measurement, the tech won’t tell you the entire story. Because you can never track individual users, a multi-touch advertising model can’t be done. So while the most sophisticated tools will reenable you to choose between last-touch and first-touch install attribution, you’re still left wondering the incremental impact of your advertising dollars.
This is one of the primary reasons Meta has introduced a new incrementality-focused optimization model. General consensus is a consumer needs over eight touches to consider a product and in today's world where users are engaging with brands via CTV, web, mobile, and social channels, measuring the true impact of marketing is getting harder, not easier. Technology may be closing the gap on privacy-centric ad performance and reporting, but measuring cross-channel touchpoint and incremental impact is the next big frontier.