Stop treating the on-premise as the first budget you cut

You are a beverage alcohol channel strategist advising a [winery / distillery / importer / distributor sales manager / retail wine and spirits shop] in [state or market]. I am deciding how to reallocate selling effort between on-premise and off-premise for the next 90 days.

Here is my situation:
- Portfolio and price points: [list your top 5 SKUs with FOB and shelf or menu price, for example estate Sauvignon Blanc $24 shelf, small-batch bourbon $27 shelf]
- Current split of selling days: [for example 70% off-premise, 30% on-premise]
- Sales team: [number of reps, in-house or distributor-led, territory size]
- Sample and incentive budget for the quarter: [dollar amount]
- Accounts I already have: [number of chain retail, independent retail, restaurants, bars]
- What is actually depleting: [paste last quarter's depletion or sales data by SKU and channel, or describe it]

Produce the following:

1. A channel reallocation recommendation with a specific new split of selling days, and the math behind it using my own numbers, not industry averages.
2. A target list structure: how many on-premise accounts to pursue, what type (neighborhood restaurant, hotel bar, steakhouse, wine bar, cocktail bar), and which of my SKUs fits each type and why.
3. For each of my top 3 SKUs, the single on-premise placement to chase (by-the-glass, well pour, cocktail list, bottle list) and the price point that makes the account's margin work.
4. A staff training one-pager per SKU: three things a server or bartender can say in ten seconds that would make a guest order it.
5. What I should keep doing in off-premise, specifically, so I do not abandon a channel that still pays the bills.
6. The three ways this reallocation could backfire, and the early warning sign for each.

Rules: use only the numbers I gave you. If a recommendation depends on data I did not provide, say so and tell me exactly what to pull. Do not suggest anything that requires hiring. Push back if my price points do not support on-premise margins, and tell me what price they would need to be. If any step takes more than 90 days to show results, flag it and give me a version I can run this month instead.