Two On-Time Delivery Numbers and a Mix Shift
Situation. Two Q3 on-time delivery (OTD) figures reach the same steering committee. Operations reports 94.1%; Finance reports 88.6%. The definitions differ three ways: Ops measures against the original commit date, Finance against the latest commit after customer-approved change orders; Ops drops cancelled lines from the denominator, Finance keeps them as misses; Ops calls a multi-line order on-time if its first line ships by commit, Finance requires every line.
Separately, regional OTD is NA 97.0% to 98.0% year over year, EMEA 94.0% to 95.0%, APAC 82.0% to 83.0%. Mix by shipped lines: NA 60% to 48%, EMEA 22% to 21%, APAC 18% to 31%. The blended figure fell about a point, and the VP opened the meeting with 'so we got worse everywhere.'
1. Show why the blended rate fell while every region improved. What is the mix effect in points?
Read the worked answer
Last year blended = 0.60(97.0) + 0.22(94.0) + 0.18(82.0) = 58.20 + 20.68 + 14.76 = 93.64%.
This year blended = 0.48(98.0) + 0.21(95.0) + 0.31(83.0) = 47.04 + 19.95 + 25.73 = 92.72%. A fall of 0.92 points.
Constant-mix counterfactual - last year's mix at this year's rates = 0.60(98.0) + 0.22(95.0) + 0.18(83.0) = 58.80 + 20.90 + 14.94 = 94.64%, which is about +1.0 point versus last year.
Decomposition: +1.00 (performance) − 1.92 (mix) = −0.92 (observed).
The mechanism is that APAC is roughly 14 points below the group average and grew from 18% to 31% of shipped lines, so every line that moves into APAC drags the average down even though APAC itself improved by a point. Recognising this matters for what the VP does next: the blended number now measures the geographic footprint as much as it measures execution, and it will keep drifting down for a quarter or two purely from mix.
2. Reconcile 94.1% and 88.6%. Which number goes on the executive slide, and what must accompany it?
Read the worked answer
The 5.5-point gap is fully explained by three definitional choices, and each one has a defensible owner.
Original vs latest commit date: Ops' version gives credit for a promise that was renegotiated after the fact. Finance's version measures the date the customer is actually holding in their inbox. If the goal is the customer experience, latest commit wins.
Cancelled lines: if the customer cancelled, counting it as a miss is wrong; if Dell cancelled, it is a genuine failure and Finance is right to count it. Split cancellations by initiating party before either number is used.
First-line vs all-lines: first-line on-time flatters a multi-line order that arrived in three shipments; from the customer's perspective the order was late.
Recommendation: publish Finance's definition as the customer-promise metric - latest commit, all lines, cancellations split by initiator - because it is the one a customer can experience. Keep Ops' original-commit metric as an internal execution measure on the same slide, clearly labelled. Do not choose the higher number for the executive summary and the lower one for the appendix: that is how two teams end up reporting different truths for a year. Prove the reconciliation with a single table attributing each of the three rules to its points of the gap.
3. Rewrite the VP's opening line as something defensible and useful in three sentences or fewer.
Read the worked answer
"Every region improved by about a point. The blended number fell about a point because APAC - 14 points below our average - grew from 18% to 31% of our volume, so on a constant-mix basis we are actually up about a point. The number to manage this quarter is APAC's absolute rate and the ramp plan, because the mix drag continues for at least two more quarters even if execution holds."
Why this version works: it corrects the false premise without contradicting the VP's instinct that the number is bad, it shows the arithmetic in one line so nobody has to trust the analyst, and it converts a reporting argument into a decision about where to spend effort. The general rule for this situation is to always show a rate with its mix beside it, and to name the counterfactual explicitly - a like-for-like figure that nobody asked for is the single most useful number in the pack.