Get with the times!

We’ve all felt it. We’ve seen it on the faces of our colleagues, our families, and our business associates. The dreaded ‘spring forward’ day – where daylight savings time begins and our sanity ends (even if only temporarily). This is written on April 1, and you’re quite possibly still feeling the effects of your lost hour from several weeks ago. Defenders of the status quo will mention the increased daylight during the typical workday, the supposed reduction in energy consumption, and the historical precedence. Defenders of using horses as the primary means of transportation use the same trite arguments.

For the visionless, the answer is to get rid of daylight savings time (or, more preferably, standard time). For those who think strategically, however, there are much better solutions. What are those solutions? Well, it depends. When we help clients build a predictive model we don’t always use LightGBM, even though it is the best. Your organization shouldn’t always use Coordinated Universal Time, even though it is the best. Below, we’ll discuss some options for second-tier time coordination and considerations for selecting which one is useful for you.

Sometimes when building predictive models it is difficult to beat a naïve model, which in regression is typically just predicting the mean.

The equivalent in time coordination is the static time approach: rather than exhaust resources keeping and coordinating time (clocks, calendars, time zones, etc.), just always assume the average time (12:00 when translated to UTC).

Organizations often want to use complicated approaches that require all users to spend time keeping time, but is it worth it? Call it 12 and be done with it.

Some organizations require fluctuations in time. Imagine a database that tracks the time that customers have to wait in line, or that it takes to fill a container, or that it takes an employee to create a task. Static Time makes those calculations incredibly efficient but provides no variability or useful measurement. Monotime allows for the calculation of time differences but centralizes the time keeping process to avoid renegade time keeping. Imagine a single location, likely your organization headquarters, that keeps solar time. The rest of the organization can simply use that time.

When your global company schedules a meeting at 4pm, everyone knows when to log in.

“It’s 4pm here, what time is it there?” now becomes “It’s 4pm”.

It is still possible to measure the passing of time, but there is no wasted effort into translating times between two places. This is perhaps the greatest balance in capability and flexibility.

Monotime keeps track of a solar time for a given location. If you have a powerful organizational leader that travels often (Bob), then the monotime could be set to his or her solar time, rather than to a static location. Or, have a major client? Make them your reference solar time! What a show of goodwill, and a useful tool to cater to the client’s schedule.

With monotime people in your organization might have meetings in the middle of the night (though the definition of night should be tailored to meet your organizational needs). Perhaps your teams rely on the position of the sun in the sky? In that case, a team that sits in New York shouldn’t report the same time as the California team. You might think that UTC (the typical time zones) account for this, but data scientist have often reported on the problems of chunking time into ‘zones’, such that all clocks within the zone are set to the same time.

With continuous time, your time wouldn’t be kept arbitrarily the same as other people in the same ‘zone’. The solar time in New York City is very different from that in Buffalo, so why lose that distinction? With continuous time each individual machine would keep it’s own time so that the time for a computer in the Bronx would be different from a computer in Queens.

Even different computers in the same office would have unique times, a supremely accurate reflection of the true solar time. Measurement error would likely start to become a consideration as you measure the time for different sides of the same microchip.

Solar time vs standard time

In 1999 the United States spent roughly $100 billion to prepare for the “Y2K glitch”. This was because the standard for encoding a date was to use only two digits, but when the first two digits in the year changed, the last two digits were no longer sufficient. This was a classic case of organizations not being prepared for the future and being stubbornly unwilling to address how they track time.

Is your organization falling into the same trap now and keeping only terrestrial time? Sure, the position of the sun in the earth’s sky is important. The last two digits of the year are also important. But strategic organizations will realize that a time system that also accounts for the galactic position of the solar system and the universal position of the galaxy are better positioned to stand the test of time.

This may be coming sooner than you think, and many organizations already rely on extraterrestrial timekeeping: The biggest hurdle to this method is that the name “Coordinated Universal Time” is already taken.

In 1905 (per UTC and conventional time and date keeping) Albert Einstein proposed special relativity, which postulates that it is impossible to say whether two events occurred at the same time. This is because the frame of reference is relevant, and there are an infinite number of reference frames. It doesn’t seem plausible to collect data on every event from an infinite number of reference points. (While Moore’s law still holds, it will be an infinite number of years before our computers are able to store an infinite amount of information.)
So, due to special relativity, there is no absolute time, only conditional time. When we see a timestamp on a record, then, there is some uncertainty about where it was recorded relative to where we are when viewing the information. This section has words like ‘conditional’ and ‘uncertainty’, which surely means that a Bayesian solution would be appropriate. Since this is impractical, it is only relevant for academia.

These are only a few of the possible solutions for keeping time at your organization. Coordinated Universal Time is but another option, and just because it is what most people have used, that doesn’t mean it is the best approach. Why must this global, digital, connected world use the same time-keeping that our ancestors used during WWII? We don’t!

Though it may be April 1 on all of our calendars now, here’s to a (relative) future where nobody knows the time or date.

Oh….and happy April Fools Day!