It's claimed that "what gets measured gets managed". Often, the thing measured is a proxy for the performance of a complex system. And complex systems are inherently hazardous. Not only do they fail; they produce both positive and negative emergent effects. It helps to have counter-indicators to mitigate the negative effects and the bias managing a measure introduces. It also helps to have a fundamental understanding of what it means to be so data-driven. Recognising the essence of measuring helps too; measurement implies a value, and many of the most valuable things cannot be represented numerically.

Nevertheless, the idea is sound; measures get managed. So what needs to be measured to improve curation? For individual curators (as opposed to multi-agent curation at scale), there are three metrics:

• Time-to-save (TTSa): how long it takes to capture something from a source in a store

• Time-to-share (TTSh): how long it takes to send something from a store to a target

• Time-to-search (TTSe): how long it takes to surface saved and shared things

Time-to-save (TTSa)

You're in a queue, browsing the web on your smartphone. You come across an interesting article but don't have time to read it now. You decide to save it to your "read later" app. So you hit the share button in the browser's toolbar, select the option for your app and navigate through the dialogue, perhaps adding a tag and a quick comment before hitting save and seeing a confirmation.

This can take up to 10 seconds, depending on how disciplined you are with annotation at the point of capture. If you're saving something on a platform to an integrated store (an Instagram post to an Instagram collection, a tweet to your bookmarks), the process will be closer to a few seconds than 10.

Time-to-share (TTSh)

Later in the day, you open up the saved article. Whilst reading, you think of a close friend who'd really enjoy it. Perhaps because it relates to a recent discussion you had. As a considerate and kind person, you don't want to send a raw link that they have to open and decipher for themselves. So you select an especially spicy snippet and copy it. You open your email client, start a new message, select your friend's address, add a subject line, paste in the snippet, tweak the formatting, add a short contextualising note as a preface and then hit "send".

This can take 30 seconds; sometimes minutes if you're really picky about communication aesthetics. Similar to saving, if the sharing takes place within the sandbox of a platform, then the time-to-share will be less. A second or two if you just fling across the raw link sans context; a little more if you include a quick note.

Time-to-search (TTSe)

Two weeks later, you're working on a presentation and recall an article. Your mind, in all its mystery, has made the unexpected connection. You open up your "read later" app and try some combination of suspected keywords, filtering by date to aid the process. The app surfaces a list of possible items and you quickly scan through, seeing the saved article you had in mind.

Later that same day, still grinding away at that capstone presentation, you remember what kicked the whole initiative off. Months ago, you sent a colleague a publication which analysed a competitor's new product and provided some commentary, and it snowballed. You open up your email, hoping to find those original, honest and organic insights. You type your colleague's name and add some keywords to further narrow the results. After some arbitrary opens and scans, you find the email and the original insights you developed in tandem.

The first search case likely takes 15-20 seconds; the second likely take 30-60 seconds. Both these times are shorter if you've been more disciplined about annotating the items, if you're adept at using sharper search tools, and if you're recall of the thing searched for is higher fidelity.

The impact of TTS

Here's a minute of time in five second increments:

• █ █ █ █ █ █ █ █ █ █ █ █

Above, we walked through some typical scenarios and used them to estimate time-to-save (TTSa), time-to-share (TTSh) and time-to-search (TTSe) for a typical curation experience. Here are the more conservative values for the three metrics:

• TTSa: █ █

• TTSh: █ █ █ █ █ █

• TTSe: █ █ █ █

Now here's what we're aiming for with Subset.

• TTSa: ▓

• TTSh: ▓

• TTSe: ▓ ▓

The differences?

• Subset TTSa: 2x faster

• Subset TTSh: 6x faster

• Subset TTSe: 2x faster

Prior to Subset, Matt's Magnificent Seven email digest would take 15-20 minutes to produce. The upside is unreasonable, so he was fine with it. Yet now it takes a couple minutes, at most. Prior to Subset, emailing a snippet of an essay to a friend took us a few minutes. Now, it's a 30-second affair.

An interesting aspect of the current versus the potential TTS metrics is that the biggest opportunity (time-to-share being 6x faster) is also the easiest one to attain: the solution is protocolised sharing that's platform agnostic by design. It's a type of sharing that involves a more deliberate approach to saving on the part of the end user, and this more structured saving and sharing activity in turn leads to a much more effective search experience. And it has a flywheel effect. Sharing invokes more saving and more search on the part of the sharer. It fundamentally rewires one's disposition, how one sees and examines the world and its information.

Most interesting, though, is that the time-to-share of protocolised, platform-agnostic distribution amortises to zero over time for the sharer. We're habitual creatures, and once a pattern of sharing is instantiated we tend to replicate it. We share the same things via the same channels. Traditionally, that meant undertaking the same preparatory sharing actions over and over (and impeding sustained sharing as a result). With Subset, however, one simply sets up a sharing pattern and then the parties steer it over time using queues, promises and strings.

This matters more now than ever before because our digital experiences and dispositions are cascading down and altering our analogue reality. Above, we said that the measuring and management of things is often a proxy for the performance of a complex system, and that the most valuable things cannot be numerically represented. Saving, sharing and search are a digital means to an analog end. As activities, they're proxies for something much more important: our relationships. Improving how we save, share and search ultimately transforms how we connect with one another.