Herd immunity to infectious disease has been an important, if sometimes controversial, achievement in public health. When Western explorers first had contact with previously isolated populations in ‘new territories’, in the Americas, Africa, and Australia, those populations were often devastated by infections such as measles, which had become relatively well-tolerated in Europe. We now face similar problems as childhood immunisation programmes falter, and our herd immunity falls.

Operating systems currently lack any adaptive ‘resistance’ to malware in the way that humans have immune systems, but herd immunity may still be a useful concept for mass security. When many users are still running old and vulnerable versions of an operating system, maintaining its security is a lost cause from the outset. Fixing vulnerabilities in the most recent releases does little to address the population problem, and the operating system becomes the platform of choice for malware development.

Microsoft’s drive to drag Windows users to the latest release of Windows 10 is therefore understandable, if dreadfully executed. It looked as if Apple might have been setting off down that same painful road, but thankfully it has stopped short of trying to make users upgrade.

The result so far for macOS Sierra appear very favourable. Traffic-based estimates from GoSquared suggest that, less than four weeks after its release, macOS 10.12 is being used on over 20% of Macs running OS X / macOS. El Capitan accounts for another 40% or more, and around 30% are still on Yosemite.

Figures for iOS appear even better, although it has only been available for a week longer than Sierra. David Smith’s analyses suggest that more than two-thirds of all iOS users are already running iOS 10, and 25% are still using iOS 9.

Windows seems to be struggling by comparison. Viewed across all desktop operating systems, NetMarketshare reports that nearly 60% of all desktop users are still running Windows 7, a further 14% Windows XP, 15% Windows 8 or 8.1, and only 0.13% are running Windows 10. If that last figure is anything near an accurate reflection of operating system use, Microsoft should be deeply concerned.

Android is even more worryingly fragmented. Google’s own figures show 4.4 (KitKat) is most popular, with less than 30% of users, 5.1 (Lollipop) on 22%, 6.0 (Marshmallow) on 18.7%, and 5.0 on 13%. That leaves over 15% running Jelly Bean (4.1-4.3). Bidouille offers several interesting visualisations of what has happened with Android versions over time. Given that Android 7 (Nougat) was released a month before Sierra, those figures are grim, and are reflected in the menagerie of malware available.

There are many complex factors which determine whether users upgrade their operating system, but some of the most obvious are also the most important. Whilst Microsoft may feel that threatening to charge for an upgrade, and forcing updates on those who have upgraded, are effective, in many cases those strategies have caused greater problems.

It is far better to offer users as many carrots as you can muster: most importantly, users must see an upgrade as having worthwhile benefits, and a minimum of costs and disadvantages. For the great majority of those still using Yosemite and El Capitan, the only real issues in upgrading are with third-party support for peripherals, and updating third-party apps. Those are both self-fulfilling to a degree too: as more users upgrade, so the incentive/demand on third-party vendors grows, strengthening their business imperative to release updates which secure compatibility with the new version of the operating system.

The idea that there is a single ‘critical mass’ beyond which third-parties will inevitably support a new version of an operating system is oversimple. Each developer has to assess the cost, in terms of the development required to make their products fully compatible, against the benefit, in terms of new product sales.

Investing heavily in supporting legacy products is only justifiable where the vendor considers that doing so will retain customer loyalty. Those still using a peripheral which was discontinued more than five years ago are hardly going to be willing to pay to update its software support, and extending the working life of legacy products can limit sales of newer products. Where competition is fierce, the vendor may value customer loyalty sufficiently to make updates essential. And very popular legacy products which do not benefit from updated software support can prevent significant numbers of users from upgrading their operating system.

Predicting uptake of an operating system upgrade, such as macOS Sierra, is thus incredibly complex, and probably beyond meaningful modelling. Whatever Apple’s expectations, releasing Sierra was a shot in the dark. So far it seems to be on target.