Bridging what the operation needs and what agentic tools can actually do.

Plan on these cards is the plan of record — the number written on day one. The KPI strip above divides by the goal, which is what the funder gate turned that plan into once a funder burned past its planned hours. Homes therefore reads against 6,005 here and 6,025 there, on the same actual. That gap is the point, not a rounding error.

A / Day reading of +N means the metric is already past plan — red on Hours, where overage is cost, green everywhere else. TOTAL / Day adds up the figures in the column above it, counting a program that is already past plan as zero rather than as credit against the ones still owing: finishing early on one campaign does not reduce what another campaign owes tomorrow. TOTAL Rem is the one column that does net, because it reports a single combined balance.

Program attainment — ranked on the gate-corrected goal

Premium Bundle leads the Executive Summary's ranking at 103% and places third here at 98.2%. Nothing about the campaign changed between the two lists. It is the only program whose funding source burned past its planned hours, so it is the only one measured against a goal that grew — 375 homes promised, 395 homes measured, the same 388 delivered. The other four are ranked against goals identical to their plans, which is why their two positions agree.

Tuesdays are the strongest day (avg 360 sales, 0.450 GPH) while Fridays lag (avg 232 sales, 0.387 GPH).

% to Goal is empty until a program is chosen, and that is the product rather than a gap: both pages pass a null SPH goal and the panel returns null while the filter reads All, because a business running five SPH plans has no single rate to divide by. Pick a program and the column fills. Where a program is dialled at two sites on two different plans the rate here is plan-weighted; the source reads whichever agent row it finds first, which reports one of the two plans as if it were both. Hours print to a tenth with no separator on every row of this table and rounded with one on Trends — two tables, two formats, both the source's own. The date drill-down opens the day at the filter you are looking at, and its % to Goal divides GPH by the campaign's plan SPH, the same measure as the column above it; the source divides by a prorated daily goal this schema does not carry. Agent mode never collapses a name into two campaign rows, because nobody here carries two campaigns in a month and the expansion that exists for that cannot fire.

A week that is not finished must not be read against a whole one, so a partial week compares against only the prior days that share its weekdays and says so — (prior matched to 3d). This month opens on a Wednesday and closes on a Monday, so the first and last weeks carry the label and the three in the middle line up exactly and do not. The Total row is the two full months and is deliberately not matched: it is the same comparison the MoM tab reports, 5,840 against 6,482. Click a week to open its days; a prior weekday with no current day to sit beside gets its own row rather than being dropped.

Charted are the two programs both sites run — the campaign filter above reaches the other three. Each point is a week's own SPH over that program's plan SPH, and a week under ten hours is not drawn, which is the source's own rule. The axis is fixed at 80–115% over every filter state rather than recomputed from what is on screen: the source rescales per render, which makes two different readings look like one change. Site A leads Site B on both shared programs for four weeks and loses the last one — the one-day week its ten remaining agents carry without the fifteen who had already stopped dialling. That is the wind-down read as a rate rather than a headcount.

Plan on these cards is Site A's plan of record — every goal row at the number written on day one, dependents derived per row and summed. The KPI strip above divides the same homes actual by the gate-corrected goal, so homes reads against 4,360 here and 4,466 there. That gap is the point, not a rounding error.

Funder gate — Site A

A goal is not a constant. A campaign's goal scales by its own hours — but only once the whole funding source has burned past its planned hours. Below plan, every campaign inside it keeps the goal it opened the month with. Expand a funder to see what its campaigns are measured against.

A funder's Scale is its own hours ratio, which is the gate test — above 1.000 and the gate is open. Each campaign inside it then scales by its own ratio, so a funder's scaled goal is the sum of its campaigns' scaled goals rather than its own factor applied to its own plan: 2,363 × 1.035 would give 2,445, and the column adds to 2,449. TOTAL is 4,360 of plan turned into 4,466 of goal against 4,430 sold — 101.6% one way and 99.2% the other.

Supervisor ranking — Site Asite SPH goal 0.436

Agent trays show today's snapshot — the only agent-level figures this dataset models. Ranking rows above are month to date, so a tray will not sum to its row. Each supervisor here owns exactly one campaign, so the tray's Program column repeats down the tray; in the product a supervisor can span several.

Plan in the Homes group is the gate-corrected goal from the funder table — which is why Mobile Propensity's plan reads 1,248 and not the 1,399 its own hours would have bought. Internet and Mobile goals are derived from each row's own homes goal, so they scale with the gate too. The tier buttons scale all three sales groups; hours are never scaled, at any tier.

Quartile distribution — Site A

Tuesdays are the strongest day (avg 272 sales, 0.456 GPH) while Fridays lag (avg 175 sales, 0.388 GPH).

% to Goal is empty until a program is chosen, and that is the product rather than a gap: both pages pass a null SPH goal and the panel returns null while the filter reads All, because a business running five SPH plans has no single rate to divide by. Pick a program and the column fills. Where a program is dialled at two sites on two different plans the rate here is plan-weighted; the source reads whichever agent row it finds first, which reports one of the two plans as if it were both. Hours print to a tenth with no separator on every row of this table and rounded with one on Trends — two tables, two formats, both the source's own. The date drill-down opens the day at the filter you are looking at, and its % to Goal divides GPH by the campaign's plan SPH, the same measure as the column above it; the source divides by a prorated daily goal this schema does not carry. Agent mode never collapses a name into two campaign rows, because nobody here carries two campaigns in a month and the expansion that exists for that cannot fire.

A week that is not finished must not be read against a whole one, so a partial week compares against only the prior days that share its weekdays and says so — (prior matched to 3d). This month opens on a Wednesday and closes on a Monday, so the first and last weeks carry the label and the three in the middle line up exactly and do not. The Total row is the two full months and is deliberately not matched: it is the same comparison the MoM tab reports, 5,840 against 6,482. Click a week to open its days; a prior weekday with no current day to sit beside gets its own row rather than being dropped.

Charted are the two programs both sites run — the campaign filter above reaches the other three. Each point is a week's own SPH over that program's plan SPH, and a week under ten hours is not drawn, which is the source's own rule. The axis is fixed at 80–115% over every filter state rather than recomputed from what is on screen: the source rescales per render, which makes two different readings look like one change. This page is scoped to Site A, so only its line is drawn. The comparison these two programs exist to carry is on the Overview tab, where the other site's wind-down reads as a rate rather than a headcount.