A lead's calling journey

After release, a lead receives up to fifteen calls. The calling style chosen for the campaign decides how those calls are spread across the weeks that follow — quick and concentrated, or gentle and distributed. One "stop calling me" request ends the journey immediately, from any point.

The structure of the journey

Each lead's journey is its own durable, supervised process. Thousands of these journeys run at the same time. Each journey is inactive between calls. Each journey continues after a crash or a restart, at the exact point where it stopped.

Two calling styles, one entitlement

Every lead gets the same entitlement: up to fifteen calls, never more. A campaign is set up with one of two calling styles, chosen by LeadRefinery staff when the campaign is created:

Concentrated. Two calls on most business days, one call on the last day, finishing in about two working weeks.
Distributed. One call a business day, finishing in about three and a half working weeks.

An operator decides how many of a client's uploaded leads go into each calling style. The platform never splits a list of leads by itself; it works only the leads a person assigns to a campaign.

Two horizontal rows of day cells. The concentrated row has seven days with two calls each, one more day with a single call, and two reserve days, spanning about two working weeks. The distributed row has fifteen days with one call each and two reserve days, spanning about three and a half working weeks. Both rows total fifteen calls. — Both calling styles deliver **the same fifteen calls**. The concentrated style compresses them into about two working weeks; the distributed style spreads them across about three and a half. Each style also keeps **two reserve days**, unscheduled, at the end.

Every call at a different time of day

At the start of the journey, the platform assigns each lead its own order of times of day for every call in the run: morning, midday, afternoon, or evening. That order is set once and then kept, so the schedule cannot silently change part-way through a run. Across the whole run, all four times of day get used roughly equally, which is what lets the platform learn when a person tends to answer without ever calling at the exact same hour twice.

Four rules apply through the full journey:

a maximum of three calls in any 24 hours;
a minimum of three hours after the last call;
never the same part of the day two times in sequence;
no calls on weekends or public holidays.

Reserve days, and catching a lead up

Both calling styles hold two business days in reserve at the end of the run, unscheduled. A busy signal, a holiday, a temporary pause, or any other delay eats into these reserve days first, not into the calls themselves. This is the mechanism that lets the platform promise a finished run rather than a partial one.

A distributed lead that falls behind schedule can also receive one extra call on a day it is already being called, once its reserve days are used up — never a third call that day, and never past the three-calls-a-day limit. A concentrated lead cannot catch up this way: it already calls twice on most days, which is the legal ceiling in some states, so there is no room left in a single day to add a call.

The journey waits for each result

After every call, the journey stops until that call's outcome is written into the record. The platform cannot redial a lead while its last result is unknown. A one-hour limit prevents a permanent stop if a message is lost.

Every dial must first pass the fifteen-check safety gate. A refused dial does not cost the lead a call, but three limits bound the wait so an endless run of refusals cannot stall a lead forever:

Eight refusals in a row move the lead to the next business day, rather than ending the run.
When the phone network itself refuses dials because it is at its own capacity limit, the platform holds the lead for a growing wait and counts the day done after 24 such refusals, parking the lead until its next calling day.
A call that reaches a busy signal, or a similar carrier failure, does not count against the lead's fifteen calls, but it is redialled at the next open slot rather than immediately; six such redials in a row stop the run, because a number that is busy every single time is a dead number, not a slow one.

A merchant can have more than one phone number on file. Today the platform dials whichever number it currently ranks best, every time, so every call in a lead's journey rings the same line. If that number turns out to be invalid, or the merchant asks it to stop calling on it, the platform automatically promotes the next-best number and the journey continues on the new line.

A lead the platform cannot finish, it does not start

Every campaign has a fixed window of working days in which its released leads must finish their calls. Before releasing any lead, the platform checks whether that lead's chosen calling style can complete inside what remains of the window. A lead that cannot finish stays waiting rather than starting a run it cannot complete. A lead worked partway through is treated as worse than a lead never worked at all, because a partial run distorts the results LeadRefinery reports back to the client. A lead that is held back this way keeps its full entitlement of fifteen calls and can be carried into a later campaign.

The four exits

The journey has four exits. Three of the exits are available at every moment:

Converted. A booking, a sent application link, or a live transfer to a person, ends the journey as a success, at any point. The platform does not call a lead again just to offer a link it already sent.
"Stop calling me." The platform suppresses the number in the same minute, from any point. The journey ends immediately, and every other client holding the same number is protected the same minute too — a "stop calling me" request is never kept private to one client. The same request also closes any callback or pipeline promise still open on the lead.
The campaign is killed. A LeadRefinery staff member can end a whole campaign at once. Every lead of that campaign still in a journey ends immediately, the same minute, from any point. This exit is separate from a campaign's own soft stop, which leaves an already-released lead's journey to finish on its own.
Journey complete. The last call ends with no conversion. The platform closes the lead and keeps its full history.

Two overrides exist, both started by a merchant's own words on a call:

A callback. The merchant gives a day and time within about the next four weeks, or wants a person to call but gives no usable time at all. Either way, this promise belongs to the client's own team, not to the platform. The journey ends there: the platform does not dial it, and the promise appears on the results screen for a person to keep.
A pipeline wake-up. The merchant wants to be contacted further out than four weeks, or vaguely, with no firm date (treated as about three months out), capped at a year ahead. The journey leaves its normal pattern and sleeps until that one later date. On the wake date the platform re-checks the number is still safe to call, then starts a brand-new journey from the beginning, exactly as if the lead had just arrived.

The technical bit, translated

Temporal runs each lead's journey as one durable workflow. This long job keeps a record of its own progress. It stays inactive for days and uses no resources in that time. After a crash, it continues at the exact point where it stopped. This makes a multi-week schedule safe in software.
The lifecycle state machine: at every instant, a lead is in exactly one named stage. The stages are: queued, active on its schedule, scheduled for a promised callback, stopped while it waits for an outcome, suppressed, closed, or ended by a campaign kill. Only permitted moves between stages exist. The dial gate's first check reads this stage. Thus no path can dial a lead that is not in a calling stage, and a killed campaign's leads cannot be dialled again by mistake.
A fixed dialling plan, set once. When a lead is released, the platform writes down its whole call plan — which business day each of the fifteen calls falls on, and which time of day — before the first call goes out. The lead's live progress is then compared against that written plan, so an operator can see, days before the run ends, whether a lead is on schedule, behind, or no longer able to finish on time.