Operators Reference

This page documents all comparison operators available in the audience filter builder. Each operator is listed with its supported data types, behavior, and examples.

Operator Summary

String Operators

Equals

Exact match comparison. Case-sensitive by default.

SQL: country = 'US'

Not Equals

Inverse of equals. Matches entities where the field does not have the specified value.

SQL: status != 'churned'

Contains

Substring match. Returns entities where the field value contains the specified text anywhere within it.

SQL: email LIKE '%@company.com%'

Not Contains

Inverse of contains. Matches entities where the field value does not contain the specified text.

SQL: email NOT LIKE '%test%'

Starts With

Prefix match. Returns entities where the field value begins with the specified text.

SQL: phone LIKE '+1%'

Not Starts With

Inverse of starts with. Matches entities where the field value does not begin with the specified text.

Ends With

Suffix match. Returns entities where the field value ends with the specified text.

SQL: email LIKE '%@gmail.com'

Not Ends With

Inverse of ends with. Matches entities where the field value does not end with the specified text.

Numeric Operators

Equals / Not Equals

Exact numeric match. Same semantics as string equals/not equals.

Greater Than

Strict greater than comparison.

SQL: lifetime_value > 1000

Greater Than or Equal

Greater than or equal comparison.

SQL: order_count >= 5

Less Than

Strict less than comparison.

SQL: days_since_last_purchase < 30

Less Than or Equal

Less than or equal comparison.

SQL: avg_session_minutes <= 2

Between

Inclusive range comparison. Matches entities where the field value is between the lower and upper bounds (inclusive of both bounds).

SQL: lifetime_value BETWEEN 100 AND 500

Not Between

Inverse of between. Matches entities where the field value is outside the specified range.

SQL: age NOT BETWEEN 18 AND 65

Boolean Operators

Is True

Matches entities where the boolean field is true.

SQL: is_enterprise = TRUE

Is False

Matches entities where the boolean field is false.

SQL: has_opted_out = FALSE

Equals / Not Equals

Boolean fields also support equals and not equals with explicit true or false values. Functionally equivalent to Is True / Is False.

Date and Timestamp Operators

Date and timestamp fields have their own operators, because time questions come in shapes a plain comparison cannot express. There are four families.

Time windows — Within and Not Within

A time window is described by four controls: a direction, a count, a range, and — for calendar ranges — whether the period you are currently in counts.

Direction is which side of now the window sits on. Last looks backwards, next looks forwards. This is what makes “renewal date in the next 30 days” possible.

Range comes in two flavours, and the difference matters more than it first looks:

• Rolling ranges (hours, days, weeks, months, years) measure elapsed time from the moment the audience is evaluated. “Last 4 days” means the last 96 hours, ending at this instant.
• Calendar ranges (calendar days, weeks, months, quarters, years) snap to period boundaries. “Last 4 calendar days” means four whole days on the calendar, whatever the time of day.

Including the current period has three settings, shown only for calendar ranges:

The count always means exactly that many periods. So “last 4 calendar days including today” reaches back three days — today plus three earlier ones. The row shows the resolved meaning in words underneath, so you can check you got the window you meant.

At a count of 1 the calendar combinations become the everyday phrases, with no special operator needed:

Not Within means outside the window in either direction — not “before” it. A value two years in the future satisfies “not within the last 4 days”. Rows where the field is empty match neither operator, and the row says so.

Points in time — Before, After, On

These operators want a single point rather than a span, and that point can be a picked date or a relative one: an offset, a range, and ago or from now. An offset of zero means the period you are currently in.

This is what a window genuinely cannot say. A window always runs from now to somewhere; on 4 calendar days ago is that one day by itself, and between 30 and 7 calendar days ago needs two independent points.

The operator name decides which edge of the period the cut falls on, so there is no extra control to get wrong:

That also makes the pairs exact opposites: before selects the same rows as not on or after.

Anniversaries

An anniversary operator discards the year of the stored date and asks whether the resulting month-and-day falls in your window. That is what powers birthday and renewal-anniversary campaigns.

Three things are worth knowing:

• It works across the New Year. A window running 30 December to 2 January matches a 1 January birthday.
• 29 February falls back to 28 February in years that do not have one, so a birthday campaign stays in the same month.
• The window must be shorter than a year. A longer one would match everyone who has a date at all, so it is refused rather than quietly returning your whole audience.

Repeating parts and times of day

Some questions are not windows at all, and no start-and-end can express them.

Date-part operators read one component of the date and ask whether it is in a set you pick. They are periodic — true in every cycle, every year — which is why no start-and-end can express them.

There is one operator per component, so picking the operator is the whole choice:

You can select several values at once, so “month is January, June, December” is one condition rather than three ORed together.

Time of Day reads the clock position inside the day, which the hours range cannot express — “within the last 4 hours” is a duration, never a reading.

A band whose start is later than its end crosses midnight and matches both sides of it. The row tells you when that is happening.

Absolute dates

Every point-in-time operator still accepts a picked date; switch the control from relative to a date.

A range can mix the two: a fixed launch date on one side and current calendar day on the other.

Which timezone all of this is evaluated in

Every claim above — “today”, “this month”, “22:00”, and a picked date’s midnight — is computed in the audience’s evaluation timezone. It is set in three places, each overriding the one above it:

1. Workspace setting — the default for every audience. UTC unless you change it.
2. Audience setting — overrides the workspace for one audience.
3. A single condition — overrides the audience for one row.

The chain exists because one audience sometimes genuinely needs two clocks: store visits on the store’s local calendar day, and email opens on the campaign’s UTC day.

Two related points:

• The week start is a workspace setting, Monday by default. It is written into every query explicitly rather than left to the warehouse, because the warehouses disagree with each other — so the same audience would otherwise give different answers depending on which source it read.
• Fields carrying no timezone of their own (a DATETIME, or a TIMESTAMP_NTZ) are read as UTC. If such a field actually holds local wall-clock time, its readings will be off by that offset near midnight. The condition row says so whenever a non-UTC evaluation zone is in force.

Windows saved before this redesign

An audience saved earlier keeps working and keeps matching exactly what it matched before — nothing changes on its own. Its row is marked (as saved), and hovering it shows the equivalent in the new controls.

The moment you touch any control, the window is rewritten in the explicit form. You will usually see the count change by one — 30 days becoming 31 calendar days including today, for instance. That is not a new window: it is what the saved one always meant, now stated plainly.

List Operators

In

Matches entities where the field value is any one of the specified values. Enter values as a comma-separated list.

SQL: plan_type IN ('free', 'trial')

Not In

Inverse of In. Matches entities where the field value is not any of the specified values.

SQL: status NOT IN ('deleted', 'banned')

Null Operators

Is Null

Matches entities where the field value is null (not set).

SQL: phone IS NULL

Is Not Null

Matches entities where the field value is not null (has a value).

SQL: email IS NOT NULL

Operator Behavior by Data Type

This matrix summarizes which operators are available for each data type:

Next Steps

• Filter Builder — How to use operators in the visual builder
• Creating an Audience — End-to-end audience creation guide