Running a calling campaign

A campaign works one list of leads, at one calling style, at a pace the whole platform can sustain. The pace for the entire run is solved once, before the first call goes out. Emergency stops at every level can still stop all calling in less than one minute.

What a campaign is

A campaign is created in a single step. LeadRefinery staff pick:

• the client;
• the calling style — concentrated or distributed;
• one or more voice agents, each pinned to one specific build;
• a name;
• the list of leads to work, and how many of that list to take.

Nothing else is set here. Dayparts, region exclusions, and frequency limits belong to the calling style itself, which is a reviewed, shared configuration — not a field an operator can mis-click on one campaign. The agent selection is per campaign for a reason: the pilot intentionally runs different sales approaches side by side, and results then identify the best one. LeadRefinery staff decide which builds a campaign dials with, so a change to the agent never reaches a running campaign by surprise.

Picking more than one agent splits the campaign's leads evenly across them, so their results can be compared on equal terms:

• each new lead goes to whichever chosen agent currently holds the fewest leads, which keeps the split even even when leads arrive unevenly;
• a lead keeps the same agent for every call of its own journey — the platform never switches a merchant to a different voice partway through;
• an agent removed from the campaign keeps the leads it already holds; only new leads stop going to it.

A new campaign starts as a draft. An operator then starts, pauses, resumes, or stops it. Starting a campaign, or its rehearsal, on a client with no usable phone number in its pool now warns first and names the fix, instead of every dial silently failing one by one. Up to two campaigns can run for the same client at once — one for each calling style — but never two campaigns running the same calling style together, so a client's list is never worked twice by mistake. The safety configuration is visible in the console, but no one can change it there. The frequency limits and the compliance limits are not self-service settings. You can see the exact values; you cannot change them.

The release plan: solved once, then executed

Before a campaign's first call, the platform solves a release plan for the whole run. The plan answers one question for every day ahead: does the planned number of calls fit inside that day's real capacity? Capacity is checked against:

• the number of leads on the list;
• the capacity that the client's phone-number pool has;
• the capacity that the voice provider and the phone provider can sustain;
• the client's own team's capacity to answer live transfers;
• the client's spend limit.

Leads that fit are admitted and start calling on the days the plan gives them. Leads that do not fit stay queued — full call entitlement kept, zero calls made — and this is reported the same day the campaign starts, not discovered partway through the run. A queued group can later be moved into a new, separate campaign once capacity opens up.

Once a campaign is running, the platform executes the stored plan rather than deciding fresh each day. It still checks in about once a minute so a pause takes effect quickly, and so a day the platform missed — a breaker, a pause, a restart — is made up on the next day it runs, without ever releasing more than the plan admitted.

The emergency stops

Emergency stops exist at six levels:

• the whole platform — reachable in one click from a master switch, open to only a small number of senior LeadRefinery staff;
• one client;
• one campaign;
• one phone number;
• one provider;
• one lead — when a complaint or a mention of a lawyer arrives, the platform suppresses that person everywhere and pauses that client's dialling until a person reviews it.

A set emergency stop means zero calls. The final check of the safety gate refuses to dial while any emergency stop is set. Thus calls that are already scheduled also stop at their gate.

The emergency stops fail safe. If the platform cannot read the status of a stop, the platform operates as if that stop is set. A stop that is set for a compliance reason stays set until a person clears it. These stops never reset themselves.

The stop mechanism itself is fully live, and an operator can set a stop by hand in one action that takes effect in less than one minute. The design also lists conditions that must set a stop automatically. Two now do:

• spend above a set daily limit for one client — the platform totals the day's real spend as it is recorded, and blocks every further dial for that client the moment the total reaches the limit. (built)
• a large drop in a number's reputation — the daily number-health check counts how many of a client's numbers were pulled from service in the last week, and blocks every further dial for that client if too many were. (built)
• provider errors on more than a quarter of calls. (open gap)
• a sudden increase in opt-outs. (open gap)

A third condition, not in the original design, is the most severe of the four: if the platform ever fails to save a “stop calling me” request, it blocks every further dial for every client on the whole platform, not only for the client on that call — the same master switch a senior staff member would press by hand. It does not dial a number it cannot confirm is suppressed, for any client, until a person checks that the request was recorded another way and clears the stop. (built)

A fourth condition, also not in the original design, reacts the instant it happens rather than to a running total: if the calling service itself flags a campaign's voice agent as fraudulent, every further dial on that campaign stops at once, because the calling service has stopped connecting that agent's calls and they would only fail. When no campaign is known for the flagged call, the stop reaches every dial for that client instead. (built)

Every automatic stop above needs a person to clear it before dialling resumes, the same as a stop an operator sets by hand.

Pause, stop, and the staff-only kill switch

A campaign moves through a fixed set of stages: draft, running, paused, completed, or killed. Only the listed moves between stages are allowed; the record-keeping system itself refuses any other move. Three controls change a campaign's stage. Only LeadRefinery staff can use them; a client's team cannot start, pause, stop, or kill a campaign:

• Pause stops new releases in less than one minute. Pause does not cancel the calls that are already scheduled for today. Those calls still occur, and each one must still pass the safety gate before it rings. A paused campaign can resume.
• Stop is the soft end. It stops new releases, the same way pause does, but it does not reach a lead that a previous release already sent into its own journey. Those journeys keep running to their normal end. A stopped campaign can start again, so a stop pressed by mistake does not strand any lead behind a dead campaign.
• Kill is the hard end, and only LeadRefinery staff can press it. Kill stops new releases and also reaches every lead still open in that campaign, ending each one's journey immediately. Nothing leaves the killed stage.

The safety gate checks a campaign's own stage again, immediately before every single dial. A dial made for a killed or archived campaign, or for a lead no longer on it, is refused there too, even if the kill itself somehow missed that lead. This second check exists because an operator once ended a campaign by editing its record directly, bypassing the kill button entirely, and its released leads kept dialling for a day before anyone noticed.

The platform itself moves a campaign to the completed stage on its own, the moment the plan has released everything it admitted and no released lead is still active. Nobody has to press a button for a campaign to finish cleanly.

The console states these differences clearly. Thus, after a pause or a stop, the scheduled calls that continue through the afternoon are not a surprise.

A separate switch sits in every page's header, open to anyone on the client's account. It is stronger than a single campaign's own Pause: it reaches every dial for that client at once, including a lead already working through its own journey, not only a campaign's new releases. It is reversible, the same way a campaign's own Pause is. See the control room for how it reads and who can use it.

The live view

The campaign-control screen shows how many of the list's leads are still arriving, waiting, working, suppressed, or finished, next to today's totals: calls released, calls placed, and how many of those calls counted toward a lead's fifteen. A separate day-one report, visible to LeadRefinery staff, names exactly how many leads were admitted, how many are queued, and which capacity limit set the size of the plan. A client's Live activity screen reads the same live figures as campaign control, so the two views of one campaign never disagree. A change of stage — a pause, a stop, or a kill — also posts a notice to the affected team, so a change made by one person is visible to the rest without a refresh.

If two of a client's campaigns are both running and their release days fall on different dates, the platform raises a loud warning. Matching release timing across a client's two calling styles keeps a later comparison of the two fair; the warning exists so that drifting apart is never silent.