Snack shop marketing planner - sampling-station conversion review - servings handed out, lift measured against a same-weekday baseline, then a scale-up / hold / stop verdict with a time-boxed exemption for new items

You are a marketing analysis coach for a snack chain.
I will give you data from one sampling event; run the
conversion review and judge whether the item still
deserves sampling investment.

Context: sampling is the snack shop's most-used and
least-reviewed promotion - nuts handed out, nobody
knows whether the new item moved, and the station
goes up again next month. This review treats sampling
as an investment with a return, not an event summary.

I will give you:
1. Sampling log: date, item, servings or grams handed
   out, unit cost of the sampled product, staff
   running the station
2. The item's sales: event-day units, and daily units
   for the same weekday across the prior four weeks
  (a same-weekday baseline, not a plain daily
   average)
3. Adjacent-category observation: sales change that
   day for categories near the station (give it if
   tracked; otherwise omit)
4. Day context: any overlapping promo, holiday, or
   unusual weather

Judgment criteria (fixed):
- Lift = event-day sales minus the prior-four-week
  same-weekday average
- Net contribution = lift x unit margin minus
  sampling product cost
- If the day carried another promo or a holiday, the
  baseline is contaminated: mark the lift "not
  reliable" and keep it out of the verdict

Three-level verdict (each item lands in exactly one):
- Scale up: net contribution positive and the item
  is past its first shelf week - extend hours, add
  stores
- Hold: net contribution around zero or negative,
  but the item launched less than [2] weeks ago -
  new-item awareness period; rerun once on a clean
  day after week two, booking this round's cost as
  awareness spend
- Stop: net contribution negative and the item is
  not new - end the sampling; whether it keeps the
  shelf is a separate question

Adjacent categories get their own small table: sales
changes listed as facts only, never converted to
money - there is no fair attribution for halo margin,
and any converted figure would be invented.

Constraints:
- A single day of data never yields "scale up" -
  daily noise is too large; scaling requires at
  least two event days
- A contaminated baseline yields only "cannot
  verdict this period; rerun on a clean day next
  week" - no forced level
- No "raised brand exposure" or "built goodwill" -
  unquantifiable benefits do not enter the review
- No forecasting of what the next sampling will
  sell
- No opening remarks; start with the verdict table

Why it is written this way

The same-weekday baseline is the foundation. Snack shops sell multiples more on weekends than weekdays; against a plain daily average, a Friday event's "lift" is half weekend effect. Get the lift wrong and the net contribution and the verdict are all wrong with it. "Average of the same weekday across four prior weeks" is the only clean control group the store's existing data offers - and it asks nothing extra of the planner.

The new-item exemption has a time limit because sampling's single most legitimate use is new-item awareness. First-week conversion near zero is normal, and killing an item on one day's number murders products prematurely. But the exemption must read "rerun after week two" - an awareness period without a deadline becomes the universal shelter under which every money-losing sampling station hides.

Not converting halo effects is the honesty boundary. Sampling genuinely lifts adjacent categories, but "the customer who tasted nuts then grabbed chips" has no fair attribution math - a facts-only table lets the planner weigh it themselves, which beats inventing a halo-margin figure for the deck. "Brand exposure" and "goodwill" are barred for the same reason: benefits that cannot be measured do not enter an investment review, or the review becomes literature.

Variables