How the numbers on this site are estimated

The starting point: a published yearly figure

The United Nations Population Division publishes population estimates and projections, including information about deaths, through World Population Prospects. An annual worldwide figure means the estimated number of deaths during a calendar year. That figure is not a list of names or times. It is assembled from demographic evidence and statistical methods, which may be revised when better data become available.

To judge a displayed estimate, look for the publisher, the dataset edition, the year and geographic grouping, and whether the number is an estimate or a projection. A projection describes a calculated future outcome under demographic assumptions; it is not an observed total. The date beside a citation does not by itself prove that the precise numerical input was checked.

An undocumented moving figure is more misleading than a clear unavailable message. If the numerical display cannot be traced to a published input, do not treat its movement as evidence of actual events. You can still use the explanatory sections and guides without relying on the counter.

Why the number advances smoothly

A simple display spreads the annual estimated total evenly over the seconds of that year. There are 31,536,000 seconds in an ordinary 365-day year and 31,622,400 in a 366-day leap year. Dividing the yearly total by the appropriate number of seconds gives a rate for the display. Multiplying the rate by the seconds that have elapsed gives the amount shown for that period.

Try the sample below. It begins with an invented total of 36,500 deaths in an ordinary year. That would be an average of 100 a day, so the model would show approximately 50 by noon UTC. Change to a leap year without changing the annual total and the modeled daily rate becomes slightly smaller, because the same yearly amount is spread over another day. This exercise demonstrates arithmetic; it does not estimate the real world's deaths.

Real deaths do not arrive at evenly spaced moments. Seasons, events, regions and ordinary variation affect the pattern, and reporting can be delayed. The moving display makes the scale of an annual figure easier to consider. Its motion is a visual calculation, never evidence that a particular death occurred at the instant a digit changed.

Why Today begins again at midnight

Today refers to the portion of the current calendar day that has elapsed in Coordinated Universal Time, abbreviated UTC. UTC gives the site one shared day boundary even when visitors live in different time zones. At 00:00 UTC, Today's elapsed time is zero; the modeled number starts again. A person may still be on the previous local date when that happens. The display needs to say UTC clearly rather than silently using the viewer's device time zone.

At a UTC year boundary, the demonstration uses the seconds appropriate to the new calendar year; an ordinary year and leap year have different denominators. A displayed figure also needs a source value for that specific year. If one is unavailable, the old year’s figure cannot responsibly stand in as the new year’s estimate. Historical does not restart when Today does.

The digital figures require additional assumptions

No demographic dataset observes the number of photographs, email accounts, devices or subscriptions held by people when they die. To illustrate possible scale, a model can combine estimated deaths with an assumed share of people who used a category of technology and an assumed number of relevant records per person. A separate assumption may describe how long some records persist. These factors must be named and kept apart from data that were actually published.

The assumptions cannot be universal shortcuts. Internet use and device ownership differ across countries, ages, income groups and time. Information about all living users cannot automatically be applied to those who died in a particular year. The International Telecommunication Union publishes useful evidence about global connectivity and the divides that remain; it does not measure deceased people's private digital belongings. Where reliable data for a particular group are missing, the project must label a proxy as an illustration, not quietly give it the authority of a population survey.

A category also needs a clear unit. One person might have two phones, multiple email addresses and thousands of photographs; a household bill is not the same thing as a phone. A 'digital trace' might mean an account, a device, a file or an activity record. These units are not interchangeable and must not be added into a grand total that looks precise but has no consistent meaning.

Historical: what may have existed, and what might remain

Historical looks across earlier years rather than restarting each morning. The view called Ever Associated adds together records that the model imagines could have been associated with people who died during the period being studied. It is cumulative within the defined model window. This does not mean those records are still online today—or that the project has identified any actual person's accounts.

Potentially Remaining asks a different question: under stated assumptions about deletion, account closure, device failure and preservation, how many modeled records could still exist? Older years may contribute less to that second figure because some records are presumed lost over time. A file can also remain but be inaccessible, encrypted or intentionally private. 'Still exists' does not mean 'available to family members'.

The page's teaching example lets you change a hypothetical yearly survival factor and observe how the remaining estimate changes while Ever Associated stays the same. Those are fictional inputs used to explain a distinction. They are not the website's validated historical estimates, statistical confidence intervals or recommendations about how long anyone should keep information.

Reading the ten kinds of digital traces

The opening experience groups digital traces into ten familiar areas: devices, email, photographs, social accounts, cloud files, subscriptions, websites or domains, AI accounts and histories, digital assets or intellectual property, authentication factors, and other activity records. Their labels are a way to explore different questions, not a claim that every person had one of each. Some areas overlap in everyday life—for example, a photograph may be on a phone and also in a cloud library—so the model needs explicit category rules to avoid treating one record as several identical records.

Other Digital Traces can refer to browsing records, searches, location history or app activity, but these examples must be bounded by the model’s published category definition and kept separate from records already counted elsewhere. Where an adoption or retention assumption has little evidence behind it, the result is illustrative—not a tally of records known to survive.

The safest reading is to treat the cards as questions about types of information that may outlast their original use. If you want guidance about a real phone, photograph or bill, choose the practical guides instead of interpreting a modeled category as a personal inventory.

What low, middle and high illustrations mean

An assumption can be changed to see whether a result depends heavily on it. A lower, middle or higher scenario may use different adoption rates, item quantities or survival periods. These alternatives reveal the consequences of chosen inputs. Unless a statistical method and evidence actually support probability ranges, they must not be called confidence intervals, odds, or predictions of what will happen to an individual family.

The most important sources of uncertainty are not always small rounding differences. A category may lack worldwide historical data; older services may have changed their rules; the population that used a technology may differ from the people who died in that year; and content can be duplicated or deleted in ways a broad model cannot observe. A result with many digits after the decimal point has not resolved those questions.

When you compare versions of an estimate, check the source edition, date, units, category definition and assumed model settings. A revision may change a number without changing the underlying question. Old printed explanations can still show the arithmetic, but they cannot establish the latest provider procedures or source values.

How to use the numbers responsibly

Use the modeled displays to understand why digital records deserve attention, not to infer what a particular deceased person owned. If you find yourself thinking about a relative's email, a treasured photograph or a household payment, the practical guides offer a better next step. They explain possible outcomes, what can be prepared and what an appropriately authorized helper may need to check.

You can begin without creating an inventory, and you need not reveal account names or credentials to this project. A small private note about where an independent copy is kept or which provider handles an essential bill may be more helpful than an impressive number. If a topic concerns another person's records, their wishes and the relevant authority matter alongside technical feasibility.