Calculated metrics content library - what are you looking for? 📚

🚩 April 2022 update: calculated metrics inside calculated metrics!

This technical introduction will start by taking two steps back and give you a brief introduction to the overall Planhat data structure. Understanding the data structure will impact your ability to understand calculated metrics in-depth.

We're going to approach this like a funnel - start broad, and then narrow down to the end deliverable which is "how to build a calculated metric".


  1. Planhat Models

  2. Data types - fields and metrics

  3. Calculated metrics

  4. Tips & tricks when when experimenting with calculated metrics

1. Planhat Models

First, it's worthwhile to start by explaining models - probably the most fundamental element of Planhat's (and other tools') data structure.

Models (sometimes also called "objects" in Planhat) are the core elements to which all raw data is associated to. For example, when you sell a deal then a license is created with an ARR value, a start date, end date, etc. Those three data points are part of the License model.

Without going into too much depth, let's make one comment on the data hierarchy of Planhat: the company model is at the top/center. For example, all licenses, end-users, conversations, tasks belong to a company. This makes intuitive sense - everything that happens with your customers belong to a customer. What this means is that the company model will often allow for standardised aggregations (e.g., we can see the end-user activities for Company X because all the end-users who performed the activities belong to a company by default).

Where can you find Planhat models in the app? Everywhere! They're literally what your Planhat is built upon.

At the top of the Data Module you find some (but not all) - companies, end-users, etc

In Settings > Fields you find models and their associated fields (think of fields like columns in an Excel spreadsheet)

In Customer Intelligence > Editing a Widget you have to select which models you want to represent data from, for some of the chart types. Do you want to see Company data? License data? End-user data? That's the "model" selection, even though it's called Object there.

2. Data types - fields and metrics

Data is associated with models/objects, and exists in two main types: fields and metrics.

2A. Fields and Metrics

  • Field: a single-point-in-time data point (e.g., MRR, number of active licenses) of any type (list, text, number) that you typically find in the Data Module as columns

    • System fields: standard fields in Planhat

    • Custom fields: your own custom fields

  • Metric: numeric time series data (e.g., active users but could also be MRR, active licenses) that lives in the Customer Intelligence Module

Both of these data types are reference-able across Planhat, but go to "Fields" under Settings or "Trends & Analytics" in the Customer Intelligence Module to find their "homes".

2B. Metrics - the different types

Since we're talking about "building calculated metrics", let's explain the different types of Metrics a bit deeper. We're now in the Customer Intelligence Module:

  • System metrics: standard metrics in Planhat

  • System-generated metrics: perhaps a bit counterintuitively, there are also standard metrics under Conversations, Tasks, Revenue which are a function of data you input elsewhere in Planhat

    • For example, your Conversation Types (...create standard Conversation metrics) and Task activities (...create standard Task metrics)

  • Custom metrics: this is raw data you are bringing into Planhat yourself for your own custom metrics (e.g., weeklyActiveUsers).

  • User activities: this is raw data on usage you can bring into Planhat which can live on multiple models (company, end-user, project, asset)

  • Calculated metrics: custom-built metrics in Planhat - let's continue below on how to build one

👑 Best-practices on bringing in custom metrics

  • You can not directly reference custom metrics everywhere in Planhat, so convert them to calculated metrics first (see below how-to)

  • Push as raw / disaggregated data into Planhat as possible, and build calculated metrics on-top (vs. pushing in pre-calculated metrics). This makes your Planhat instance more adaptable and modular, allowing you to create that new metric in 2 years time which you didn't know you wanted yet

3. Calculated Metrics

If you've seen some examples of calculated metrics, you know they can look like this:

{"type": "metricOverTime", "days": 7, "op": "SUM", "prop": "activities.login"}

{"type": "propertyValue", "prop": "company.custom.expected number of logins"}

{"type": "rawNumber", "value": 100},

{"type": "metricOverTime", "days": 1, "op": "LAST", "prop": "metrics.active_seats"},{"type": "metricOverTime", "days": 1, "op": "LAST", "prop": "metrics.max_seats"}]]

There are several variations between these examples, and we will unpack the structural ones step-by-step below.

3A. Defining the calculated metric "type"

We'll start with the differences in "type" you find at the beginning of each metric or sub-metric. This you can also find on the left-hand side guide when creating a new calculated metric in-app.

Let's follow the structure of the in-app description, perhaps with slight repetition but also expanding a bit. Again, remember the different data types field and metric, what the difference is, and where they live.

{"type": "rawNumber"}: this is just if you want to write a number. Note that while all other arguments are written with " ", numbers are written without citation marks.

{"type": "rawNumber", "value": 100}

{"type": "propertyValue"}: used when referencing a field (which does not exist as a time series). Note that the calculated metric will start tracking the data from today and going forward, and not retroactively. So when building a calculated metric of type propertyValue, then expect that the time series could be empty (value = 0) or constant for historical dates.

{"type": "propertyValue", "prop": "company.custom.expected number of logins"}

{"type": "metricOverTime"}: used when referencing another system/custom metric or calculated metric which already exists as time series data (i.e., "metric over time").

{"type": "metricOverTime", "days": 7, "op": "SUM", "prop": "activities.login"}

{"type": "daysSince"}: less common, measuring days since last occurence, only available for Activities and Conversations.

{"type": "daysSince", "prop": "activities.logins"}

📌 Important to note

As you can see in the examples above, the "type" you choose impacts the other arguments you need to input. For example, with a rawNumber you only need a "value" while for propertyValue and daysSince you only need a "prop".

3B. Calculated Metric Operators

Real power in calculated metrics don't come purely from referencing another data point. but rather in combining multiple ones into insight. For example, this week's number of active users vs. last week's number of active users. Making that comparison requires some type of operation, most likely a division or a subtraction.

On the left side window of creating a new calculated metric, you can find the seven mathematical operators available for group components and how to use them.

One of the tricky ones is the SIGN function - read more on that here.

3C. How to find the name of a metric/field you want to address

In the latter part of the metric, you need to address the data you want to represent. How do I find the name, and what are some naming conventions?

Metric names are always found in light grey below the given metric name when going into details. See three examples below.

activeUser is a custom metric with the name metrics.activeUser

Health Score is a system metric with the name

Recurring Base is a standard revenue metric with the name revenue.rrvalue

Field names are equally found in light gray below the given name, or in the case of standard fields the "given name" = property name.

Company fields, where you can see name, owner, country, mrr

License fields, where you can see renewalStatus, _currency, fromDate

📌 Important to note

You need to spell the name exactly as you find it. That includes both upper and lower case letters and spacing. So if you create a custom field called "No of companies" then the way to reference it in a calculated metric is No of companies.

👑 Almost by the finish line! Here's some space to regain energy before diving into...

3D. Addressable models in a calculated metric

We started this article by briefly explaining models, and now the reason for doing that might become clearer. As you can see, there are metrics and fields living under all models, and we need to clarify what you can and can't address when building a calculated metric.

There are four kinds of calculated metrics you can build:

  • Company calculated metric (again, remember that company model = top/center-of-the-hierarchy)

  • End-user calculated metric

  • Asset calculated metric

  • Project calculated metric

The rules for what models you can address are the following:

  • For company calculated metrics you can reference:

    • Company custom and calculated metrics and company numeric fields

    • End-user, Asset, and Project custom metrics (not fields)

  • For End-user calculated metrics = end-user metrics and end-user numeric fields

  • For Asset calculated metrics = Asset metrics and asset numeric fields

  • For Project calculated metrics = Project metrics and project numeric fields

3F. Referencing different properties in Planhat

Finally, let's bring it all together to explain how to reference properties of various data types.

Referencing calculated metrics

Read more here, but put shortly you use the calculated metric ID as follows:

{"type": "metricOverTime", "days": 5, "op": "AVERAGE", "prop": "calculated.123456789"}  

Referencing system or system-generated metrics

Write "system.[metricname]" or "revenue.[metricname]" or "task.[metricname]"

  • Check this out by pressing "Generating a simple example" when editing a Calculated Metric and you can see how a system field is being referenced

  • Use the {"type": "metricOverTime"} since we're addressing a metric

Average health in the past week compared to last month (the 100x multiplication is a matter of preference, turning percentages from e.g.. 0.08 to 8)


["DIVISION",{"type": "metricOverTime", "days": 7, "op": "AVERAGE", "prop": ""},{"type": "metricOverTime", "days": 30, "op": "AVERAGE", "prop": ""}],{"type": "rawNumber", "value": 100}]

Referencing custom metrics

Write "metrics.[metricname]"

  • Use the {"type": "metricOverTime"} since we're addressing a metric

  • If it's a nested model metric, write [nested_model].metrics.[metricname]

Monthly active users as a share of total users


{"type": "metricOverTime", "days": 30, "op": "LAST", "prop": "metrics.monthly_active_users"},
{"type": "metricOverTime", "days": 30, "op": "LAST", "prop": "metrics.total_users"}],

{"type": "rawNumber", "value": 100}]

The calculation above might seem slightly complicated. What's "LAST"? And why is the metric "monthly" AND we take 30 days? Let's unpack.

  • This is a case where you are pushing in "monthly active users" less than once per day (perhaps once a month)

  • However, the calculated metric needs to calculate a value for each day, and it doesn't make practical sense for the value to be 0 on days where data isn't pushed in

  • Therefore, we use "LAST" to look back in time and pick up the last available value

    • For example, maybe today is Nov 12 and monthly_active_users is pushed in on the 1st of each month, so we use "LAST" to bring in the data from Nov 1

  • However, we only want to look back maximum 30 days to give this month's average active users, which is why we define "days" to be 30.

    • For example, using our previous case, if we would have set "days" to 2 then the calculation would be empty for Nov 12

Referencing fields

Finally, let's address fields.

  • If system field: [model].[fieldname]

  • If custom field: [model].custom.[fieldname]

  • Use the {"type": "propertyValue"} since we're addressing a field

Number of Seats active - field addressed in the denominator

{"type": "metricOverTime", "days": 1, "op": "LAST", "prop": "metrics.active_seats"},
{"type": "propertyValue","op": "LAST", "prop": "company.custom.seats"}

A couple of notes when addressing fields in calculated metrics:

  • This data will now be empty or constant historically because, as mentioned, we're starting to store a single-point-in-time data point as a time series

  • You can not directly reference fields from a different model (so e.g., building a company calculated metric with an end-user field). You can indirectly solve this using something called "formula fields" which can make cross-model references, but that is currently not available to customers in the UI

4. Tips & tricks when when experimenting with calculated metrics

The best learning always comes from trial & error, so we encourage you to bring the theory from above into practice. Try building a few metrics. Examine the results. Go back to this article. Iterate.

So finally, we'd like to pass on some helpful notes to aide in that process.

(Some) Quick notes on error messages

  • If it says "Metric re-build in process", try refreshing the page and see if it remains

  • Note that you will not be able to save/update a calculated metric with an incorrect syntax, so that's a "failsafe" to indicate whether your reference is correct or not

Common root causes of incorrect results

When results from a calculated metric don't match the expected reality, this is often due to one of two factors: either the metric logic is incorrectly built, or the data input is erroneous.

  • Incorrect logic: deeply think about the underlying data and how you manipulate it in the calculation. For example, are you sending in data which is already aggregated, and then summing it up over several days again?

  • Erroneous data: naturally, if the data that is coming in is off, then all downstream data points will be off too. Here, use the "Show raw events" view on the custom metric or activities to examine what data is actually coming in. We suggest looking at this over time and filtering in companies you can sample check

Did this answer your question?