Purpose: Share a non‑comprehensive set of formula examples to cover the majority of customer use cases

Note: Without the Account Intelligence add‑on product, you can only create 2 formula fields for free. It is still possible to use 2 formula fields to build out a health score, but it

0. Data to power formulas

Here's a list of data you can use in Pylon to power formulas

• Issue data

• Meetings (via Google Cal Integration or Meeting Recorder Integration)

• Custom Activities (note: custom activity types require descriptions to work with formula field generation)

• Custom Account fields

  • Manually created Pylon Account fields

  • CRM synced Account fields

  • Product usage data, or any other data synced in from a data warehouse

1. Standard formulas

These are standard formulas for what you see in Excel or Google Sheets. Type directly into the formula editor and you will see the list of formulas and options. Some common ones we use:

• Logic: If, And, Or, Not

• Math: Sum, Average, Max

• Date: Today, Day, Month, Year

• Month example: To convert the numeric month returned by Month() into a name, use nested If statements (e.g., If(Month=1,"Jan", … )).

2. Query formulas

These look like simple formulas, but are actually querying your customer data. The one below will return the number of issues that each account submitted, that is tied to a Linear ticket, in the last 90 days

Using custom activities in query formulas: To use custom activity types in account‑level formulas:

1. Go to Settings › Account data › Create Field › choose Formula

2. In the formula editor, start typing to add a query on Activities / Custom activities

3. Use the editor’s prompts to filter by your custom activity type (by its name) and aggregate (e.g., COUNT, SUM)

Note: Custom activity types work with specific query functions like querycount, which are primarily for time‑based metrics.

Here are common QUERYCOUNT formula patterns:

Tip: When creating a Formula account field, Pylon offers standard formula templates for common account‑level statistics like issue counts to help you get started.

Count issues by time period

QUERYCOUNT({{ pylon.issue.issues }}, null, 30, "d")

This returns the number of issues created in the last 30 days for each account.

Count issues by status

QUERYCOUNT({{ pylon.issue.issues }}, {{ pylon.issue.state }} != {{ closed }})

This returns the total number of open issues per account. You can then filter an account view based on this field.

Count issues with specific criteria

QUERYCOUNT({{ pylon.issue.issues }}, {{ pylon.issue.linear_product_issue }}, 90, "d")

This returns the number of issues tied to a Linear ticket in the last 90 days.

3. AI formulas

AI formulas require the following sequence to setup:

Creating AI-based categorical fields

To create AI‑based categorical fields (like sentiment):

1. Add the categorical values as options in a new select‑style field

2. Create a prompt for each category option so the AI knows how to categorize it

3. The AI will then automatically categorize based on these prompts

This approach can be applied at both the account level and the issue level to manage issues according to sentiment.

• Create a custom Account notebook with the relevant data you'd like to reference

• Create an AI formula field that uses the Account notebook as its context. In the below explain, Mentioned competitor is an Account notebook AI block. To create such formula, create a Text, Select, Multi-Select, or True/False Account field and set Fill field with AI to Enabled

• Create a custom Account notebook with the relevant data you'd like to reference

• Create an AI formula field that uses the Account notebook as its context. In the below explain, Mentioned competitor is an Account notebook AI block. To create such formula, create a Text, Select, Multi-Select, or True/False Account field and set Fill field with AI to Enabled

Creating AI formulas from the Accounts page:

To create AI formulas directly from the Accounts page:

1. Navigate to the Accounts page

2. Add a column for the formula

3. Select Formula

4. If the AI formula feature is available, you will see a “Generate” button in the pop‑up

Best practices for AI formulas

Craft specific, account‑scoped prompts

Write prompts that are clear and specific to the account level. The AI is already scoped to the account, so avoid redundant phrasing. For example: “Mark as true if there is an executive sponsor identified in {account block}. Otherwise, mark as false.”

Include relevant keywords and properties in your prompts to help the AI identify the information you need.

Test prompts using preview functionality

When iterating on AI field prompts, create a new field from the Accounts view (rather than from Account data in settings) to access the preview feature. Click “Preview changes” to see the field values for up to 30 rows before saving.

Configure sentiment fields effectively

For sentiment analysis, use Select or Multi‑Select field types for best results. Ensure your field option descriptions align with the wording in your prompt to improve accuracy. Save the field as a column on the Accounts View to make results visible once data populates.

Common formula patterns

A. Signal, explanation, next step

Set up an Account notebook specific for a given signal (upsell or risk), and create a series of formulas to extract the relevant insights. In this below case, we created a notebook for identifying customers that are interested in Account Intelligence product

In the below example, Indicated Interest and Next steps are two different AI fields within our custom Account notebook

Overall signal that aggregates multiple signals

The signal is an AI indicator of a possible upsell or risk

The explanation is an AI explanation of why the signal was flagged

Next step is an AI recommended of how to address the signal

B. Health scores

Best practice: Use formula‑based health scores

Use formula‑based health scores instead of AI‑only health scores for reliability. LLMs are probabilistic, not deterministic, so scores can vary for similar accounts—for example, one might score 75 while a similar one scores 65.

Instead, calculate health scores based on formulas using data from product usage, Pylon data, and CRM data.

For AI‑based sentiment, consider using categorical text values (like Low, Mid, High or Advocate, Positive, Neutral, Frustrated, High Risk/Detractor) rather than numeric scores.

Example formula structure

Health scores can be calculated using weighted formulas that combine multiple data sources. Example structure: Weight for (Usage) + Weight for Set‑up Score + Weight for Engagement + Sentiment.

You can also create formulas to detect significant changes in health score over time periods (e.g., last 30 days).

Aggregate health score weighted across multiple sub‑scores

Sub‑health score based on your org's definitions

Implementation steps:

1. Create custom Account fields for each signal you care about (engagement metrics, meeting frequency/recency, support volume, sentiment analysis, issue counts, last meeting date).

2. Populate these fields through manual entry, formula‑driven calculations, or AI‑filled values.

3. Incorporate product usage data by syncing it into Pylon and combining it with conversation sentiment, ticket volume, and engagement patterns in your formulas.

4. Add a Formula field that rolls those signals into a weighted overall score (e.g., (engagement * 0.5) + (sentiment * 0.3) + (meetings * 0.2)).

C. Last conversation tracking

Track the most recent interaction with a contact by combining message and meeting data:

1. Create a trigger for Last Message Received: When an issue has a new message from a customer, set an action to update a custom account field (date type) with the current date.

2. Use the existing Last Meeting Date field, which is automatically populated based on the Google Calendar connection.

3. Create a formula field that takes the larger of the two dates:

IF (Last Message Received > Last Meeting Date, Last Message Received, Last Meeting Date)

This formula field will show the most recent interaction, whether it was a message or a meeting.