For: the WFM diagnostic

Adherence and investment live on the same grain.

Two builds of the same day. A is the single chart with two y-axes. B is the panel version: off-phone investment, the day’s demand against forecast, and adherence, stacked on one hour axis. Both put off-phone investment and adherence on one interval grain, and neither claims that one produces the other. They shade on different rules, stated on the face of each drawing: A shades the day’s three lowest adherence readings, B shades the day’s three largest off-phone investments. The rules pick different intervals, which is why the pair has to be read as two readings rather than one argument. Pick whichever you want on the page.

A: one chart, two axes · shaded by lowest adherence

Off-phone investment and adherence on one chart Stacked bars show agents taken off the phones for coaching, training, team meetings and one-to-ones against the left axis. A line shows adherence for the same interval against the right axis. Schedule, activity codes and adherence are each carried to the same interval grain before anything is laid against anything else. The two series are independent readings of the same intervals: the chart places them side by side and does not assert that either causes the other. Shading marks the day's three lowest adherence readings. Coaching Training Team meetings One-to-ones Adherence Agents off the phones shaded: the day’s three lowest adherence readings Adherence % right axis, starts at 80 0 2 4 6 8 10 80 85 90 95 the plan assumes 90 11% 3% 17% 3% 3% 18% 7% 15% 7% 7% 18% 3% 89.9 86.1 88.2 8 9 10 11 12 13 14 15 16 17 18 19 Schedule, off-phone activity and adherence all ETL'd to the same interval grain. None of them compares to another until they are.
Two scales, both declared. Left axis and bars are agents off the phones and start at zero. Right axis and line are adherence and start at 80, which the chart says out loud. Each axis is drawn in the color of the marks it governs.

Shading rule: the day’s three lowest adherence readings: 13:00, 14:00 and 11:00, at 86.1, 88.2 and 89.9 against a day of 90.2%, all of it simulated.

The two series are not connected here, and that is deliberate. Coaching, training, team meetings and one-to-ones are scheduled capacity: planned, funded, and already accounted for in the bars. Adherence is a separate reading of the same intervals: how closely the day was followed, whatever the day was made of. Nothing in this chart models development as a cause of adherence loss, and nothing in it tests for one.

B: three panels, one shared axis · shaded by largest investment

Off-phone investment, demand against forecast, and adherence, by interval Three panels sharing one hour-of-day axis. The upper panel stacks the agents taken off the phones for coaching, training, team meetings and one-to-ones. The middle strip shows demand against forecast: contacts over forecast as a percentage of forecast, zero in most intervals. The lower panel plots adherence for the same interval. Schedule, activity codes, demand and adherence are each carried to the same interval grain first. Shading marks the day’s three largest off-phone investments, and carries through all three panels so the readings for those same intervals can be read off. Investment and adherence are independent readings of the same intervals. The figure does not assert that either causes the other. Coaching Training Team meetings One-to-ones Adherence Agents taken off the phones shaded: the day’s three largest off-phone investments 0 2 4 6 8 10 8.8 5.9 7.2 Demand against forecast contacts over forecast, % of forecast. Zero means the interval ran to plan 0 +20% +2% +22% +13% Adherence for the same interval an independent reading, not derived from the investment panel 80 85 90 95 the plan assumes 90 90.5 86.1 91.7 Scale starts at 80, not zero. The whole day sits in a narrow band, which is the point. 8 9 10 11 12 13 14 15 16 17 18 19 Day, schedule-weighted: 90.2%
One axis, three readings. The top panel is who was taken off the phones and what for. The middle strip is what the day threw at the plan: contacts over forecast, zero in the intervals that ran to plan. The bottom is adherence for that same interval.

Shading rule for this build: the day’s three largest off-phone investments: 10:00, 15:00 and 13:00, releasing 8.8, 7.2 and 5.9 agents respectively. The shading carries down into the lower panel so the adherence reading for each of those intervals can be read off directly: 90.5, 91.7 and 86.1, against a day of 90.2%. Two of the three sit at or above the day’s level, on this seeded day, a description of the illustration rather than a finding.

Different rule, different intervals. Chart A above shades the three lowest adherence readings: 13:00, 14:00 and 11:00. Only 13:00 appears in both selections. That is why each figure states its rule where it is drawn: shade by one rule and caption by another, and a reader will fairly conclude the figure is claiming a cause it never tested.

The middle strip exists to back a sentence. An earlier build of this caption explained the day’s worst interval by demand running over forecast while drawing no demand anywhere: the same caption-one-thing-shade-another move this page warns against. Now it is on the face of the figure: +22% at 13:00, +13% at 14:00 and +2% at 10:00, and zero in every other interval. On this simulated day the overruns are also where adherence dips. That is drawn so it can be seen, and it is still illustration, not evidence.

Nobody misbehaved. Adherence counts being somewhere other than where the schedule said, and this day loses most of it in one place: 13:00, where demand ran +22% over forecast (the middle strip) and contacts overran the start of whatever was scheduled next. That is a consequence of the operating design, not of conduct, and it is not the classroom either. The metric is doing exactly what it was built to do.

The reason to draw these together and still refuse to join them is that the join is what everyone assumes. Once adherence is a target, the cheapest-looking move is to stop scheduling development in the intervals that read badly. These two figures give that move nothing to point at: neither figure measures the relationship the move assumes, and readings on a simulated day are illustration, not evidence either way.

None of this is readable without the plumbing. The schedule, the activity codes, the contact volumes and the adherence calculation each arrive from a different system on a different clock. Laying them against each other requires every one of them ETL'd to the same interval grain first. Skip that and the comparison is not wrong so much as meaningless, which is the quiet reason so many operations never get to have this conversation at all.