Turn one order export into a best-customer profile, ad targeting, and a referral offer

You are a direct-to-consumer analyst for beverage alcohol brands. You specialize in turning raw order exports into customer profiles, paid-ad targeting, and referral offers that actually get accepted.

Read every file in this folder. They are order and customer exports from my business.

My situation:
- Business: [BUSINESS TYPE AND NAME, e.g. a 6,000-case winery in Paso Robles, a production brewery with one taproom, a craft gin distillery]
- What I sell and price range: [e.g. estate reds at $38 to $85, a club at $140 a shipment, 4-packs at $14]
- Market: [e.g. we ship to 38 states, 70% of tasting room traffic is local, we self-distribute in two counties]
- Monthly ad budget: [e.g. $1,500]
- Referral reward I can afford: [e.g. a free bottle, $25 credit, a free tasting for four, a brewery tour]
- Column names in my file: [e.g. customer_id, zip, order_date, order_total, sku, source]

Before analyzing, list the columns you found and tell me which ones are missing or unusable. Do not guess at missing data.

Then produce a Word document with these sections:

1. The Top 20%. Rank customers by lifetime spend. Describe the top 20% with real numbers: average lifetime value, average order value, number of orders, months between orders, and how much of total revenue they represent.

2. Where They Are. List the top 15 zip codes and top 5 metro areas by revenue from this group, with customer counts. Flag any zip code that over-indexes against my total customer base.

3. What They Buy. Name the specific product that most often appears in a first order among high-value customers, and the product that most often appears in the second order. Tell me the first-to-second purchase path by SKU. If a product shows up often in first orders but rarely leads to a second order, say so plainly.

4. How They Found Us. Break the top 20% down by acquisition source if the data exists (tasting room, website, event, club signup, wholesale pull-through, referral). Rank sources by lifetime value per customer, not by count. If source data is missing, say so and skip this section.

5. Three Segments. Define three distinct named segments inside my best customers, with the count in each, what makes them different, and which one I should spend money on first. Example output style: "Local Repeat Taproom (412 people, 3 zips, buys 4-packs monthly, never ships)" versus "Out-of-State Gift Buyer (190 people, peaks in November, highest AOV, lowest repeat rate)."

6. Ad Targeting Plan. For my budget, give me: which segment to target, the exact zip code or radius list to paste into Meta and Google, which customer list to upload for a lookalike audience, the product and price point to lead the creative with, and three ad angles written for that segment. Beverage-specific only. No generic ecommerce advice.

7. The Referral Offer. Build one referral offer aimed at the single segment most likely to refer, based on their order frequency and recency in the data. Give me the reward for the referrer, the reward for the friend, the exact email or SMS copy (under 90 words), who to send it to by name of segment and count, and the one metric to watch after 30 days.

8. What I Should Stop. Name one thing the data suggests I am wasting money or attention on.

Rules: every claim must trace to a number in my files, and show that number. If the data cannot support a section, write "Insufficient data" and tell me which column would fix it. Do not invent benchmarks or industry averages. Be direct. If my best customers are a smaller or less flattering group than I think, say it.