The settings that decide how a campaign calls

Every campaign points at a named group of settings, called a calling style. A calling style decides how a campaign calls: how many calls, how fast, at what times of day, and how the start of the campaign spreads across its first days. LeadRefinery staff choose a calling style when a campaign is created; nobody edits it inside the app.

What a calling style is

A calling style is a named group of settings. An operator chooses one from a list when the operator creates a campaign. The calling style then decides how that campaign calls.

A calling style answers four questions about the campaign:

How many calls does each lead get?
How fast do those calls occur?
At what times of the day do they occur?
How does the start of the campaign spread across the first working days?

Three further facts about a calling style:

A calling style belongs to one client. A client never sees the calling styles of another client, for the same reason that a client never sees another client's leads.
Up to two campaigns of the same client can run at once, each with its own calling style — never two running campaigns with the same style at the same time.
Every new client gets two calling styles from the start: concentrated and distributed. Both make 15 calls to each lead. Only the spread is different — see a lead's calling journey.

What a calling style can set

The settings fall into five groups. Each group has its own purpose and its own reason to change.

How many calls, and how fast

This group decides the amount of calling that each lead gets:

The number of calls for each lead. Both built-in styles use 15.
The number of calls on each calling day. The concentrated style uses 2 calls a day, and 1 call on the last day. The distributed style uses 1 call a day.
The number of calling days. The concentrated style uses 8 days. The distributed style uses 15 days.
The number of reserve days. Each style holds 2. The platform uses a reserve day when it puts a call off: a state time window closes, a safety stop runs, or the call must move to a later day. A public holiday does not use a reserve day, because the platform does not count a holiday as a calling day at all.
Whether the platform makes up a call that it missed, and how late a lead must be before the platform does so. A made-up call still obeys the limits in the next group.

The platform holds the daily figure as one line for each calling day, not as one packed value. A new calling style is therefore a new set of lines, and not a change to the platform.

The limits that protect the person called

This group limits how often one person hears from the platform:

The maximum number of calls to one person in 24 hours. Both built-in styles use 3 — the legal ceiling in some states.
The minimum time between two calls to the same person. The value is 3 hours.

This group holds two values, not three. The limit on calls to one person in 7 days is not a calling-style value. It counts calls from every client together, against one shared count for each telephone number, so one client cannot hold a different value for it. The platform holds that limit as a single number in the program: 12.

This group changes on legal advice, and most calling styles of one client use the same values. LeadRefinery staff make the change. The checks before each dial read these values on every call.

When in the day the calls occur

The platform divides the working day into four windows: morning, midday, afternoon, and evening. This group decides the windows:

Today the four windows are 08:00 to 11:00, 11:00 to 14:00, 14:00 to 17:00, and 17:00 to 19:30. Some states open later, and the platform truncates the morning window for them.
Each window has a start time and an end time. Each boundary is a separate setting.
A change to one window leaves the other three windows as they are.
Each time is local to the person called.
A calling style never holds a copy of the legal calling hours. It points at the record of the legal hours for each state, which the platform keeps in one place.

How fresh the screening must be

The platform screens each number against the do-not-call registers before it calls. This group decides how old a screening result can be:

The maximum age of a screening result: 21 working days, and also 31 calendar days. A screening result expires on whichever of the two comes first. The two are not the same length of time: 21 working days is 29 calendar days when no public holiday falls in them, and one day more for each holiday that does.
There is no list of fixed days for a second screening. The platform screens a number again when its screening result expires, and at no other time.
The database refuses any age above 21 working days, and any age above 31 calendar days. Those are the ceilings the contract with the client sets. A calling style can therefore make the rule stricter, never weaker.

This group changes on the contract with the client. It changes on a different clock from the calling pattern, so it is a separate group.

How the start is spread

This group decides how the campaign begins and how long it runs:

The number of working days that the start covers: 2 working days.
How the starting leads divide across those days.
The length of the term: 21 working days.

Three things a setting can never do

Each of these three rules exists for a stated reason. Together they are the reason a calling style is safe to hand to an operator.

1. A setting cannot make a legal limit weaker

The database holds the legal floors itself, and refuses a value that breaks one.
It refuses more than 3 calls to one person in 24 hours. It refuses a gap shorter than 3 hours. It refuses a screening age above 21 working days, or above 31 calendar days.
A calling style can make a limit stricter. A calling style can never make a limit weaker.
The refusal comes from the database itself, not from one program. No program can therefore create an unlawful campaign, even by mistake.

2. A setting cannot change a campaign that already runs

The platform never overwrites a setting. An edit makes a new version beside the old one.
A campaign points at the exact version it started with, and keeps it to the end of the term.
An edit therefore applies to the next campaign that selects the calling style, and to no campaign that already runs.
The reason: a change in the middle of a term must never leave a started lead without its remaining calls.
The second reason: for any call, the record answers one question exactly — which rules governed this dial?

3. A setting cannot be changed in the app

The running platform reads the settings. It cannot write them.
A new value arrives as a reviewed change to the database, not as a form in the app.
The reason: each value decides what the platform dials. A wrong value calls the wrong people, or calls them too often.
The New-campaign screen lets an operator select a calling style. It does not let an operator edit one.
A later decision can give edit rights to a role. That grant is a visible change, and never a quiet default.

The screens

Two screens carry this design. The first is where an operator picks a calling style. The second is what appears once the campaign has started.

The diagram shows one named calling style as a bundle of five groups of settings: the calling pattern with one line for each calling day, the frequency limits, the four time windows, the screening rules, and the start spread. A campaign points at one exact version of the calling style, and each lead in the campaign takes its schedule from that version. — Notice the single arrow from the campaign to one **version** of the calling style. An edit makes a new version beside that one, so a running campaign keeps the settings it started with.

The New-campaign screen. It holds an organization, a calling style with a read-only summary of what it does and the limits it holds, one or more voice agents, a name, and the list of leads with how many to give this campaign. The Create button stays off until every field is set. There is no plan preview on this screen. — Every field is required before Create turns on. Choosing more than one voice agent splits the leads evenly across them, for a fair side-by-side comparison. Notice what is **not** here: no preview of how many leads will be admitted. That answer is solved once the campaign starts.

The day-one report card. It shows four figures: uploaded, admitted, waiting, and released so far, a sentence naming which capacity limit sized the plan, and, since leads are waiting, a note that they can be selected into a follow-on campaign from the campaigns page. — LeadRefinery staff see this the moment a campaign starts. The **waiting** figure and the reason for it are visible on day one, not discovered midway through the run.

You decide which leads go where

The platform never divides a list of leads by itself. A person allocates the leads, when the campaign is created:

An operator imports the leads. The example here uses 10,000 leads.
The operator creates one campaign, selects the concentrated calling style, and gives it 5,000 leads from the list.
The operator creates a second campaign, selects the distributed calling style, and gives it the other 5,000 leads.
Both campaigns can run at the same time. The two results show which calling style works better.

Two campaigns with two calling styles are the way to compare two approaches. Some facts follow from that:

The allocation is visible to the operator, because it is a list of leads on a named campaign. The client does not see the split. The client sees one programme, with one set of totals.
A comparison is only reliable when both campaigns release on the same days. A later start gives a different calendar position and a different phone-number reputation.
The platform does not force the two campaigns to match. The operator keeps them matched, and the platform gives a loud warning when two running campaigns' release days diverge.

An operator does all of this work in the control room.

Leads that wait are not lost

A campaign admits the number of leads that its capacity supports for the whole term. When an operator gives it more leads than that, the rest wait.

This is what happens to a lead that waits:

It stays in the queue with zero calls made and all 15 calls still owed to it.
The platform counts it and reports it the day the campaign starts.
Once the plan has finished releasing, an operator selects the waiting group on the Campaigns page and starts a new campaign from it.
The new campaign calls them when capacity frees up. Capacity frees up as the first leads finish, because the daily calling load falls after working day 7.

A client's leads still get their calls. Waiting changes when a lead starts, not whether the platform calls it.

The technical bit, translated

PostgreSQL is the single system of record. Each group of settings sits in it as small tables of versioned lines. Two things follow. The database refuses a value that breaks a legal limit, so an unlawful calling style cannot exist. And nothing overwrites a line, so a campaign that runs keeps the exact settings it started with.
Temporal keeps the state of long tasks. If a server stops, the task continues. The platform solves one release plan for a whole term, so a plan that a restart could delete would not be a guarantee. No splitting logic lives in these tasks either: a lead takes its schedule from its campaign's own calling style, never from a rule the workflow itself decides.