The decision desk

MIS 78 ▲ +6 WK/WK · OUTCOME: RETAIL VELOCITY

KeepCalled Tue · Jul 13

Hold the early-fringe weight.

The daypart move is the week's biggest contributor. It lifted response without a dollar of added spend, and the effect held for a second week.

+3.2score contribution

$0added spend

2 wkseffect stability

Called by the buying desk · scored by the Impact Engine

ReallocateCalled Tue · Jul 13

Shift two overnight rotations into weekend prime, as a test.

Overnight inventory scored in the bottom decile for the third straight week. Weekend prime is the strongest unowned daypart for this audience. Small test, one week, measured against the baseline.

P9overnight percentile

+18%modeled weekend response

1 wktest window

Called by the buying desk · scored by the Impact Engine

FixCalled Mon · Jul 12

Rebalance retargeting frequency caps.

Retargeting answered same-day but contribution flattened at higher frequencies. Cap adjustment recovers efficiency without cutting reach.

+0.1current contribution

10%of attributed outcomes

Same-dayresponse profile

Called by the digital desk · scored by the Impact Engine

Draft · internalAwaiting Friday call

Consider pausing the estimated-impressions network feed.

Two networks deliver estimates, not measured impressions. The scoring discount on estimates has not been validated. Proposal: pause, measure the difference, decide with data.

2networks affected

0.6xunvalidated discount

Fridecision due

Drafted by the analytics desk · not client-visible until called

Concept C · The Decision Desk · prototype v0.1 · Jul 2026 · all numbers are sample data