LeadRefinery — How the Platform Works

The control room

A client's team does all of its work in one web console. Everyone signs in the same way, and the address decides what opens next. A client's team lands in its own workspace. LeadRefinery staff can also open a second, internal control pane that reaches every client at once.

One sign-in, two shells

You sign in one time. Invitations and password resets arrive as branded emails. What opens next depends on the page address, not on a switch you set:

  • Your workspace. A client's team, and a LeadRefinery staff member looking at one client, land here. It shows one client's own data.
  • The control pane. Only LeadRefinery staff can open it. A client account that tries the address gets a locked-door page instead of an empty screen.

Some people work at more than one client. A switcher moves a person between the workspaces they belong to, without a new sign-in.

One sign-in leads to two shells. A teal region is the client workspace: Overview, Leads and results, a note that the workspace has no campaign screens because LeadRefinery staff run every campaign, Settings, and a dashed note on the account-wide calling switch that anyone on the account can pause and resume. A dashed amber region is the internal control pane, open to LeadRefinery staff only, organised into four modes: operations — now including a real Needs review screen for calls the platform did not settle on its own, alongside Campaigns, Campaign control and Organizations, where staff run every client's campaigns — plus system health, performance, and finance, which remain fixture modes. A grey-dot note marks the control-pane screens that still show fixed example figures rather than live ones. A padlock note states that a hidden screen is not a security control, because the server checks permission again on each request.
The teal shell is a client's own workspace. The dashed amber shell is the internal control pane, for LeadRefinery staff only. Operations now carries a real Needs review screen, not only a fixture. Grey dots mark the control-pane screens that still show fixed example figures, not live ones — a preview of the layout, not today's real number.

Your workspace: Overview

  • Dashboard. The at-a-glance view: outcomes delivered daily, weekly, monthly, or quarterly, and a suppressed count. Two headline estimates sit above them: an estimated pipeline value, which prices every lead your agent moved to a callback, a transfer, or a sent application link, using the amount the merchant named on the call or a standard stand-in amount when none was named; and rep time saved, which prices what a person would have spent making these same calls by hand, weighted by how each call went. Every figure is real or left blank; the screen never draws an invented number.
  • Live calls. Every call in flight for your organization right now, plus each one that finished in the last ten minutes, refreshed every few seconds. Opening a row opens that call's full record in a panel on the right, with separate tabs for the opportunity, the call recording and transcript, the lead's own details, and a Lifecycle tab. The Lifecycle tab lists everything that happened to this lead, in order: the upload it came from, each call, every safety check that let a call through or held it back, and any stop request. Opening a lead from the Leads screen below shows the same tab.
  • Needs your attention. A dashboard card that lists jobs a person must still do by hand, soonest first. It carries three kinds of row:
    • a callback due now, overdue, or due within the next week;
    • a transfer still carrying no deal stage, a day after the call;
    • a deal stage that has not changed for five days.
    A count beside Dashboard in the sidebar shows how many rows wait. Each row can be marked done or snoozed until nine the next morning, and both actions reverse. A row leaves the list once its callback closes or its deal stage changes.

A Demo calls screen sits in its own "Try it" group in the sidebar. A prospect in a demo organization uses it to watch the platform work, in four steps:

  1. Verify a phone number by text message.
  2. Set up a persona to answer as: a name, a business name, and an industry.
  3. Make up to five practice calls. Each one comes with a suggested line to try and the result it leads to.
  4. Finish with a short survey on how it went.

LeadRefinery staff open the same screen to run a live demonstration. A practice call can exercise any of the agent's mid-call tools, including a transfer to a specialist.

A prospective client instead gets a whole sandbox organization, marked as a demo. Each invited person in it can:

  • use the ordinary workspace screens, the same as a signed-up client — upload a sample lead list and read outcomes. The sandbox's campaign is seeded for them; nobody in it creates or starts a campaign;
  • also finish with the "I'm done" step and the survey. A real client gets the same screen as the Agent Sandbox, with no "I'm done" step and no survey. Its demo calls stay apart from the client's campaigns.

Each person's demo calls have a fixed lifetime limit, not an hourly one. Thus one person's testing cannot lock another person out.

A demo organization's sandbox also rehearses on its own, with nobody pressing a button: every thirty minutes, the platform clears the previous rehearsal's rows and starts a fresh one on the seeded campaign. Thus a prospective client who opens their sandbox always finds current results on the dashboard and the Opportunities and Pipeline screens, not an empty one from the day it was created.

A person marked as a tester sees a short welcome tour on first sign-in: five slides, and a floating checklist afterward that tracks a handful of first steps — uploading a lead list, confirming a phone number, choosing a voice agent, and completing one demo call. The tour and the checklist appear only for a tester; an ordinary client never sees either one.

Your workspace: Leads and results

  • Opportunities. Every lead ever delivered as an opportunity — by a promised callback, a connected transfer, or a sent application link — newest first, with one figure tile per kind above the list — click a tile to narrow the list to it — plus a Funded tile showing how many deals funded and the total dollar amount. A delivered lead stays on this list and in its tile count for good: marking a callback done, setting a deal Funded or Lost, or a later opt-out never removes it; only its Next step column clears. Clicking the Callback tile also reorders the list by the promised time instead: a callback with no time set sorts first, then the soonest-promised time. A live connection pushes a notice the instant a new result is ready, so the list and the notification bell update within a second; a slower background check keeps refreshing underneath, so a dropped connection never means a missed result. Each row carries its own commercial stage, which you set by hand and which is separate from what the call itself achieved: Working, Application in, Funded (with an amount), or Lost (with a reason). A funded deal is priced at the amount the merchant named on the call; failing that, at your organization's own typical deal size if staff set one, or at $0 otherwise. You can download every result as a spreadsheet, and the download carries the full detail behind each call — the transcript and everything the agent collected — on its own sheet. A recording plays only inside the app, from a link the platform creates fresh each time and that expires soon after; it is never a permanent public link, and it is not part of the download.
  • Pipeline. A forward-looking count only, separate from the list above: merchants who are not ready to decide now, so the platform pauses them until a later wake date and starts a fresh calling run automatically when it arrives. Charted by month, with how many wake in the next thirty days. It is a count, not a value — the dashboard's estimated pipeline value, above, is the only place a dollar figure for this forward work appears. Only LeadRefinery staff see the button to download the list as a spreadsheet; a client sees the count and the chart only.
  • Leads. All leads, fifty for each page, with three tabs: the lead queue, Upload, and Batches. Six status tiles sit above the queue — Not called yet, Being called, Opportunity, Future pipeline, Do not call, Completed — click one to narrow the list to that status, click again to clear it. A Last activity column states the newest thing that happened on each lead in plain words, for example "Called Oct 14 · No answer" or "Asked for a callback · Oct 14". On the queue you can also search by name or phone number and filter by the upload a lead came from.
  • Upload leads. The same import wizard also gets its own place in the sidebar. The wizard has four steps: upload a spreadsheet, match its columns, review the reject report, and commit the import. Getting your leads into the system explains the full process behind these four steps.
  • Batches. The history of each upload, each with its data-quality profile at arrival.

A Billing screen was designed, to show summary figures and an invoices table, but it was never built; the product shows a not-found page at its address today. Until a real one exists, your account manager sends invoices by hand, outside the console.

Campaign screens: LeadRefinery staff only

A client's team does not create, start, or pause campaigns. LeadRefinery staff set up and run every campaign. The two campaign screens are in the control pane, not in a client's workspace. The server refuses these actions from every client account, admins included. A client's team uploads leads and works the results.

  • Campaigns. Staff create a campaign, which starts as a draft. Staff can pick several uploaded lead lists at once, each with its own count of leads to take from it. Staff can assign leads only while a campaign is a draft or paused; pause a running campaign first. The screen shows each campaign's safety configuration as read-only. Nobody can change the pace limits or the compliance limits here.
  • Campaign control. Staff start, pause, resume, or stop a campaign. A setup summary shows the calling style, the pace, the voice agents, and when the campaign was created. Four live gauges show the leads in the queue, the leads released today, the remaining release capacity, and the projected days until the lead list is empty, next to a chart of the release rate against the safe limit. Below the gauges, a day-by-day table lists the last thirty days: leads released, dialled, and denied, next to bookings, transfers, callbacks, and stop-calling requests. The table names each quiet day's reason — a weekend, a public holiday — so a day with no dialling reads as expected, not as a fault. It is the same table a rehearsal shows before a campaign goes live. Running a calling campaign explains what pause and stop each do, and how they differ from the staff-only kill switch.

Your workspace: Settings and the calling switch

Each person manages their own profile, password, and two-factor security from one entry point, reached through the account menu at the top right rather than from the sidebar. Admins get two more screens there: Team, to invite colleagues by email and see each person's role, and Do-not-call, to record a stop-calling request that reached the client outside a call — a letter, an email, a complaint by phone — for any number, with or without a lead attached yet. A client admin's entry stops calls to that number for that client only. Only LeadRefinery staff can stop a number for every client on the platform.

Two more settings screens exist inside the workspace — Phone numbers and Voice agents — but a client admin never sees either one. Only a LeadRefinery staff member looking at that client's workspace can open them, because buying, registering, and rotating numbers and assigning a voice agent's build are LeadRefinery decisions, not a client's task. The Voice agents screen shows the current list of agent builds and lets staff assign one or more to this client; the list follows the voice provider automatically, once an hour, so a newly built agent appears with nobody pressing a button. Staff can also add a number to this client's own pool directly from this screen, without switching into the client's workspace first.

A pill at the top of every page reads "Calling on" or "Calling paused". Anyone on the client's account can press Pause, with one confirmation. New calls stop within seconds, across every running campaign of that client, and across a lead already working through its own calling plan, not only a campaign's new releases; a call already connected finishes normally. A paused state shows an amber banner that names who paused it and when, with its own Resume button. It is the reversible brake: pausing and resuming restarts exactly where it stopped. It is not the same control as the staff-only kill switch described below.

The control pane: one mode open, three held back

LeadRefinery staff open a second shell that reaches every client from one screen. The design groups its screens into four modes, but only one is open in the menu today:

  • Operations — the only mode a staff member can reach right now. Campaigns and Campaign control, where staff set up, start, pause, and stop the campaigns of every client; and Organizations, where staff create and manage each client's workspace. To see a client's own data, a staff member switches the view to that client. The switch never grants a client role. A staff member can also archive an organization from this screen: archiving stops every one of that organization's running or paused campaigns, refuses every further dial for it, and locks every one of its members out, while the organization's own data stays on the record and visible on staff screens. A member locked out this way can still switch to another organization they belong to. Restoring an organization lifts the lock; it does not restart anything on its own. This screen must also hold a client's own valid postal address before its application-link emails can send. Business email law requires a true postal address in the message, so the platform refuses to send without one on file. Here staff can also set a client's own typical deal size, used to price a delivered opportunity when the merchant named no amount on the call; left blank, such an opportunity is valued at $0.
  • Needs review — a staff-only screen that lists every call whose result the platform did not settle on its own, newest first. Opening one shows the excerpt that caused the doubt, the platform's own best guess, and a link to the full call record. A reviewer can confirm or correct the result. For a do-not-call doubt, the reviewer can only confirm the block: lifting a wrongly-applied block is not yet built behind this screen, so the screen says the number stays blocked and asks the reviewer to send the case to engineering by hand. A case still waiting after 48 hours raises its own alert. See how we know the agent is good for why this queue exists.

Three further modes — System health (a live status board, alerts, voice agent performance, phone number health, lead lookup), Performance (client results compared side by side), and Finance (every client's invoices) — exist in the design but were taken out of the menu in late August 2026. Each one showed fixed example figures with no live feed wired behind it. They return, one at a time, as each is connected to real data; nothing was deleted, only hidden.

The Operations mode carries two controls that do not exist in a client's own workspace: kill and the master switch.

  • Kill is staff-only and does not reverse. It ends every still-open lead of one campaign immediately, not only new releases. Stop, by contrast, is a client-facing control that only halts new releases and can be restarted.
  • The master switch stops every real call for every client at once, in one confirmed click. Only a small number of the most senior LeadRefinery staff can reach it — a narrower group than the staff who can pause or kill a single campaign. One click refuses every dial at the last safety check, including a call whose journey is already under way, and also pauses every running campaign and every active rehearsal. Turning it back on re-enables real calling only; each paused campaign still needs its own, deliberate resume. The same switch can also engage itself, with no person pressing it: if the platform ever fails to save a stop-calling request, it throws this switch on its own, because dialling on while a stop might be lost is worse than stopping every client by mistake. The topbar names which of the two happened.