Your credit card statement, your gas/water/hydro usage data, and your daily step count - while these are very different sources of data, they all have something in common. They are all time series data sources. In fact, most of the personal data that you’ll likely look at with personal analytics would be classified as time series data. So what is time series data?

In its most basic, time series data set is made up of records which have (at least) a time field, and a numeric measurement field (with each record capturing the value of the measurement, at that respective time). The time field can be in different formats and different granularities - but don’t worry, Core Insightz Personal Analytics will normalize your data’s date/time fields for you. As well as the time and measurements fields, your data may also contain fields that allow you to look at subcategories or portions of the data separately (e.g. if your data had a gender field, you could look at the “male” and “female” subcategories separately by filtering on the gender field).

Time series data often has (a) recurring cycle(s) pattern in the data. For example, an airline’s bookings will go up and down over the year, depending on the time (e.g month) of the year. Restaurant sales vary in a consistent pattern, based on the day of the week. Patients visits to a hospital vary by the time of the day. This cyclic pattern in the data is sometimes referred to as the “seasonality” of the data. Being able to see the “seasonality pattern” in the data is extremely valuable (e.g staffing for peak periods in the day for the restaurant). At the same time, overall the data may have trends, independent of the cycles (e.g. if the business is doing well, sales may be trending up overall (with the noise of the cyclic patterns obscuring the trend). Being able to see the patterns in the trend AND in the seasonality cycle are key to understanding what is going on in your data (and what actions you should take in response).

Core Insightz Personal Analytics includes a set of tools specifically designed to help deliver insights for time series data.

Time Explorer The best starting place for understanding your time series data is to look at the big picture. The time explorer tool allows you see the whole time series at an overview level, then easily navigate through interesting points you see, with simple gestures for zooming in/out of areas of interest, and scrolling through the time range. While navigating, you can get quick popups with information on any point, and you can see summaries of all the data currently in the visible window.

Seasonality Analysis There is often important insights that can be found in understanding how your data changes in such a “seasonal cycle”. For example, a store knowing how sales vary by time of day, or an airline knowing how flight bookings change by month of the year, are crucial for how they can optimize their businesses. The seasonality analysis tool allows you to choose the period of the cycle you want to explore (e.g. time of day, day of the week, and month of the year), and it aggregates all the data across all the cycles for that period in the cycle (e.g. all data for “Mondays” is aggregated together, in the day of week analysis). Besides showing you just the average value for each point in the cycle, it also shows you a “50%” interval (I.e. the area where 50% of the data falls into, with 25% of the data above and 25% below this area). It also shows the "high water mark" maximum seen for that point in the cycle, and the minimum.

Relative Time Comparison One of the challenges of trying to look at the raw data for a dataset with seasonality cycles/influences, is that it can be a little like comparing apples to oranges (e.g. comparing Augusts hydro usage(with A/C being a big contributor) to the Septembers (where there is less A/C usage). The relative time comparison tool, “sorts the apples and oranges” . It lets you find the trend in your data, by comparing times in the current cycle, to the same time in previous cycles. For example, compares hydro usages for this August, to usage in other Augusts in previous years.

Forecasting One of the most useful things we can do with historical data, is use it to statistically forecast how the measurement is likely to change in the future. The forecast tool provides this functionality.