Export last year's sales by week to find your eight busiest peaks, then build each bundle by occasion and headcount (for example 8, 12 and 20 guests), rounding quantities up to whole packs. Ask ChatGPT to name the bundles and draft a countdown of posts for each peak, while the spreadsheet, not the AI, does the arithmetic.
The eight peaks below are a starting guess; your own till history decides. What doesn't vary is the shape of the customer's question. People think in occasions and headcounts ("a birthday for twelve"), not in aisles of plates and cups, so a bundle that answers that question in one line beats a discount on the same items. Bundles also have to be built from stock you can really count, which is where plans tend to go wrong.
Find your eight peaks in last year's till data
Export last year's sales by week for the whole shop, and by category if your till allows it. In a spreadsheet, put the 52 weekly totals in a column and add a second column that indexes each week against the average: =weekTotal/AVERAGE(allWeeks)*100. An average week scores 100; a week that sells 60% more than average scores 160. Sort by the index and read the top of the list.
You can hand the file to ChatGPT and ask for the top ten weeks and the categories driving them, which is quicker than building a pivot table. Check its list against your own sheet before you use it: the index is simple, and a wrong week costs you a month of planning. The ChatGPT and Excel guide covers uploading and checking a sales file. Here is how one illustrative shop's top weeks line up with the occasions behind them:
| Peak | Week index (average week = 100) | Categories driving the week |
|---|---|---|
| Halloween | 215 | Decorations, tableware, treat bowls |
| Christmas and holiday parties | 188 | Tableware, banners, crackers and favours |
| Graduation and end of school | 165 | Banners, balloons, tableware |
| New Year's Eve | 152 | Balloons, tableware, novelty items |
| Easter and spring | 142 | Table covers, egg-hunt items |
| Valentine's Day | 128 | Balloons, banners |
| Summer garden parties | 118 | Tableware, table covers |
| Back to school and team season | 112 | Tableware, banners |
Yours will differ, and one of your biggest weeks may be an occasion you hadn't considered: a school sports day, a local festival, a championship final. Steady-demand occasions such as birthdays, baby showers and retirements may not show up as peaks at all. They need always-on bundles that sit on the shelf all year, so the calendar below is for the peaks only.
The calendar: eight peaks, with build-up and bundles
Treat this table as a hypothesis built on occasions that many party shops see, and replace it with your own top eight once the index is in. "Start posting" counts weeks before the day. Set yours from your suppliers' real lead times, because the date you must order by comes well before the date you start posting.
| Peak | Start posting | Bundles to build | Watch-out |
|---|---|---|---|
| Valentine's Day | 4 weeks before | Date-night table for 2; friends' night in for 6 | A short peak. Mark down the day after. |
| Easter and spring | 5 weeks before | Egg hunt for 8 and 15 children; spring brunch for 10 | Small-parts fillers: copy the age warnings exactly. |
| Graduation and end of school | 6 weeks before | Graduate party for 20 and 40 | Year-dated banners can't roll over, so order few. |
| Summer garden parties | 5 weeks before | Garden party for 12 and 24 | Weather moves demand week to week. Keep posts flexible. |
| Back to school and team season | 4 weeks before | Team party for 15; class party for 25 | Schools may restrict food and decorations. Ask first. |
| Halloween | 8 weeks before | Classroom party for 30; hallway decoration kit; treat-bowl bundle | The biggest dated-stock risk. Set the markdown plan before you order. |
| Christmas and holiday parties | 8 weeks before | Office party for 20 and 50; children's holiday party for 12 | The order-by date arrives long before the peak. |
| New Year's Eve | 5 weeks before | Countdown party for 10 and 20 | A fixed date with no second chance. Keep returns wording clear. |
If your customers celebrate other festivals, such as Diwali, Eid or Lunar New Year, add them as peaks and run the same method. The calendar should follow your customers, not mine.
Bundles are built from packs, not units
A rule such as "two plates a guest" isn't enough, because plates come in packs of eight. A bundle for 12 guests needs 24 plates, which is exactly three packs, but 18 cups is not a whole number of packs of eight, so you sell three packs and the customer has six spare. Rounding up to whole packs is the difference between a bundle your staff can assemble quickly and one that sends them to open packs and count loose items.
Set the rules once in a sheet, with one row per item. These figures are illustrative; your own sales and your customers' feedback set the real ones.
| Item | Rule | Pack size | Cost per pack | List price per pack |
|---|---|---|---|---|
| Paper plates | 2 per guest | 8 | $1.60 | $3.50 |
| Paper cups | 1.5 per guest | 8 | $1.50 | $3.25 |
| Napkins | 2 per guest | 20 | $1.10 | $2.50 |
| Table cover | 1 per 8 guests, minimum 1 | 1 | $1.40 | $3.50 |
| Latex balloons | 1.5 per guest | 10 | $1.80 | $4.00 |
| Banner | 1 per party | 1 | $2.20 | $5.50 |
| Candles | 1 pack per party | 1 | $0.90 | $2.50 |
| Favour bags | 1 per guest | 6 | $3.00 | $6.50 |
Two formulas do the work. Packs needed for a per-guest rule: =ROUNDUP(guests*perGuest/packSize,0). For the table cover: =MAX(1,ROUNDUP(guests/8,0)). Cost and list price are packs times the per-pack figures. Don't ask ChatGPT to do this arithmetic. Ask it to write the formulas if you like, test them on one bundle by hand, and keep the maths in the sheet.
The 12-guest birthday bundle, worked through
For 12 guests (illustrative):
| Item | Needed | Packs sold | Pieces supplied | Cost | List price |
|---|---|---|---|---|---|
| Paper plates | 24 | 3 | 24 | $4.80 | $10.50 |
| Paper cups | 18 | 3 | 24 | $4.50 | $9.75 |
| Napkins | 24 | 2 | 40 | $2.20 | $5.00 |
| Table covers | 2 | 2 | 2 | $2.80 | $7.00 |
| Latex balloons | 18 | 2 | 20 | $3.60 | $8.00 |
| Banner | 1 | 1 | 1 | $2.20 | $5.50 |
| Candles | 1 pack | 1 | 1 pack | $0.90 | $2.50 |
| Favour bags | 12 | 2 | 12 | $6.00 | $13.00 |
| Total | $27.00 | $61.25 |
At list prices the margin is $34.25 on $61.25, or 55.9%. Price the bundle at $54.95, which is a 10.3% saving for the customer, and the profit is $27.95, a margin of 50.9%. The customer gets one decision and a clear price; you give up five points of margin and sell eight lines in one transaction. The spare 6 cups, 16 napkins and 2 balloons are part of the offer ("serves 12, with spares"), but they are also a real cost, which is why the rounding rule belongs in the sheet.
The same rules scale to the other sizes (illustrative):
| Guests | Cost | List total | Bundle price | Saving | Margin at bundle price | Price per guest |
|---|---|---|---|---|---|---|
| 8 | $21.40 | $48.50 | $43.95 | 9.4% | 51.3% | $5.49 |
| 12 | $27.00 | $61.25 | $54.95 | 10.3% | 50.9% | $4.58 |
| 20 | $40.90 | $92.00 | $82.95 | 9.8% | 50.7% | $4.15 |
The falling price per guest is a natural ladder: bigger parties see a lower figure without you inventing a separate discount. For the wider theory of setting a bundle discount that still clears your costs, the tutorial on product bundle pricing and margin goes further, and the gift shop bundle and follow-up plan shows how another retailer found the pairs customers already buy.
Naming and describing bundles from the sheet
Once the quantities are in a sheet, ChatGPT earns its place: names, one-line subtitles and descriptions that list the contents exactly. The key is to give it the sheet rows and forbid it from describing anything else.
You write bundle names and descriptions for a party supply shop.
Use ONLY the sheet rows below (contents, pack counts, headcount, price).
For each bundle give: a name (max 5 words); a one-line subtitle stating the
headcount; a description of 40-60 words that lists the contents exactly as
in the sheet and ends with the price and the price per guest.
Do not claim it is everything you need. No safety, age or "non-toxic" claims.
No licensed character names. Describe nothing that is not in the sheet.
SHEET: [paste the bundle rows]A typical result for the 12-guest bundle (illustrative):
The Dozen Birthday Box. Tableware, balloons and favours for 12 guests. Includes 24 paper plates, 24 paper cups, 40 napkins, 2 table covers, 20 latex balloons, 1 banner, a pack of candles and 12 favour bags. $54.95, which is $4.58 a guest.
That version is after editing. Watch for first-draft phrases such as "a mess-free party in one box", "perfect for all ages" and "12 filled favour bags" (the favour bags are empty). Each is a plausible sentence the sheet doesn't support, and "perfect for all ages" is the dangerous one, because balloons and small items carry age warnings of their own. Put phrases like these on the Project's banned list so the model is told to leave them out.
Kits or virtual bundles: keeping the stock count honest
You have two ways to sell a bundle, and the choice decides whether your stock count stays true.
- Pre-packed kit. A box with its own stock code, ready on the shelf. It sells quickly and looks good, but you pay in assembly time, and unsold kits are stranded stock after the peak. Components must be taken out of stock when you pack the kit.
- Virtual bundle. A price rule on components. The customer takes the listed items from the shelves, a shelf card or a basket lists them, and the till charges the bundle price. Stock comes off by component at the sale, so counts stay clean. Staff spend a minute gathering items.
My suggestion is to pre-pack only your best-selling headcount (here, the 12) in the two weeks before a peak, build the others at the till, and keep a rule for breaking a kit: reduce the kit count by one and add the components back in the same adjustment. A realistic slip (illustrative): in the week after Halloween the system showed 40 classroom party kits and the shelf held 11. Staff had been opening kits to serve customers who wanted a few items, and nobody had adjusted the kit count, so 29 "kits" existed only on screen while the loose plates and cups they came from were missing from component stock.
Online, offer the headcounts as variants (8, 12 or 20 guests) and check with your platform whether a bundle can reduce component stock automatically before you assume it does. If it can't, a pre-packed code with a counted stock level is the safer route. The weekly-box approach in the tutorial on weekly boxes and seasonal labels for a farm shop deals with the same counting problem.
A countdown of posts for one peak
For each peak, plan six posts from six weeks out to the day after. Here is the pattern for a peak whose bundles are for 20 and 40 guests, with the line you take from the sheet for each:
| When | Post | Line from the sheet |
|---|---|---|
| 6 weeks out | "Plan early": theme and colour ideas, headcount in the first line | Bundle sizes (20 and 40 guests) |
| 4 weeks out | Bundles on the shelf, with price per guest | Bundle prices and contents |
| 2 weeks out | Special-order deadline for anything not on the shelf | Supplier lead time |
| 1 week out | Last orders and collection times | Opening hours for the peak week |
| 2 days out | "Forgot something?" post listing the items people often forget | Staff list of last-minute items |
| Day after | Thank you, markdowns on dated items, a nod to the next peak | Markdown plan |
The "forgot something" post draws on the best source you have: your counter staff. Ask them for the three items customers most often remember at the till on the way out (candles, tape, a cake knife, batteries, extra bowls) and give that list to the AI. A post that names real forgotten items reads as helpful rather than promotional, and it reminds people of the small items that lift the basket.
Write the 4-weeks-out post for our graduation peak, 50-70 words, for
Facebook and Google Business Profile.
Use only: bundles for 20 guests ([price], [per guest]) and 40 guests
([price], [per guest]); contents as listed below; start date [date];
collection from [time]. Open with the headcount. No promises about stock:
say "while stocks last". Do not use "unforgettable", "bring the fun" or
"party to remember".
CONTENTS: [paste the bundle contents]Where you post matters as much as what you post. A Google Business Profile post appears on your profile when people look the shop up; as of October 2026 an event post needs start and end dates, and undated posts older than six months are archived, which gives you a natural cleanup rhythm. For the profile itself, the tutorial on local SEO for a retail shop, with stock posts and reviews sets out a routine. Add your Facebook page and any community groups that allow shop posts, plus Instagram (capped at five hashtags a post since December 2025) and one email to your list.
One prompt for eight peaks, without eight identical posts
Reusing a brief across peaks saves time and breeds sameness. If every post says "make it a party to remember", readers stop reading by the third one. Make the brief a template with fields that change each time: peak and date; bundle names, headcounts and prices; the three items customers forget at that peak; the last-order date; and a format instruction that rotates (a list, a question, a checklist, a two-line post). Keep a running list of banned phrases in the Project, and add to it whenever a draft repeats itself.
An example of the variation: for Halloween the forgotten items might be "extra bowls for treats" and "tape for hanging decorations"; for the graduation peak, "a banner with the right year" and "table covers for the extra guests"; for New Year's Eve, "poppers or confetti that the venue allows". Each comes from staff, so each is specific to your customers rather than generic.
Safety wording and licensed characters: copy, never paraphrase
Balloons, small favours and candles carry warnings on their packaging: age limits, choking hazards, flammability. Listings, shelf cards and posts must copy that wording exactly. A model asked to summarise a pack will happily write "safe for all ages" or "non-toxic", and neither claim is yours to make unless it's printed on the pack. Tell the AI to use only text from the supplier sheet for anything about safety, and read each safety line against the packaging yourself.
Licensed characters raise a different issue. Use the supplier's approved images for licensed products. Don't ask an image tool to draw look-alikes of cartoon characters for your posters, and don't ask ChatGPT to describe a licensed theme beyond what the pack states.
Three ideas that sit on top of the bundles
- A party-planner checklist. A one-page checklist ("Planning a party for 12? Here is what to buy and what to order early") that drops into an email or onto the website, drafted from the bundle sheet. It brings people into the shop with a number in mind.
- A headcount grid on the shelf. A card at the plates showing the number of packs for 8, 12 and 20 guests. It answers the question before the customer asks and makes the bundle the obvious shortcut.
- An after-party reminder. If you collect email addresses with consent, send one message a few weeks after a big purchase suggesting the next occasion. Never collect a child's birthday or details without a parent's consent, and keep whatever you do collect to the minimum.
After the peak: leftovers, markdowns and next year's order
A peak ends the day after, and the money is made or lost in what is left. Set a markdown ladder before the peak starts. An illustrative one: 25% off in the first week after, 50% off at three weeks, and after that either fold the item into an always-on bundle or box it for next year, if it isn't dated. Dated items (a banner that says a specific year) get cleared completely.
Then record the numbers, because they set next year's order:
| Peak and item | Bought | Sold | Left | Sell-through | Cost of leftovers | Decision |
|---|---|---|---|---|---|---|
| Halloween plates (packs) | 400 | 316 | 84 | 79% | $134.40 | Order about 10% fewer; shift two designs to plain colours |
| Graduation banners | 150 | 128 | 22 | 85% | $48.40 | Dated, so clear; order about the same next year |
The decision column is yours to set. A rule of thumb I'd test: under 70% sell-through, order noticeably less or switch to evergreen designs; over 90%, order a little more and earlier, because you probably ran out. Paste the review table into ChatGPT and ask for five bullets and a draft of next year's order notes, but keep the quantities in your hands. The wider logic of forecasting for small shops is in the tutorial on forecasting stock and reordering with AI.
How to tell the bundles are working
- Bundle share. Bundle transactions divided by all party-category transactions in the peak week. If it's low, the shelf card or the first line of the post may not be answering the headcount question.
- The first item to run out. It tells you which pack size is limiting your bundles. If balloons always go first, buy more of them or change the rule.
- Average basket with and without a bundle. A bundle that lifts the basket pays for the effort even at a lower margin.
- Leftover units and their cost. The table above, every peak.
- A question at the till. "Is this for a party?" and "Where did you hear about us?" give you two facts a till report can't.
If you do one thing this week, build the 12-guest bundle in a sheet with the pack rounding and look at its margin. Everything else, from the names to the posts, follows from numbers you've checked yourself.
Party bundles and seasonal posts: common questions
How many bundles should I offer for each occasion?
Three sizes and one add-on is a sensible start: for example bundles for 8, 12 and 20 guests, plus an optional extras box. More than that slows the choice and complicates stock. Look at your till data after the first peak and drop the size that nobody picks.
Should a bundle always cost less than buying the items separately?
A small saving usually helps, but it isn't compulsory; the convenience of one decision is part of the value. Whatever the discount, check the margin on the finished bundle, as in the 12-guest example, and never discount below your cost plus the margin you need.
Can ChatGPT forecast how much stock to buy for a peak?
It can summarise past sales and spot patterns in a file you give it, but the forecast is only as good as that history, and it can't see supplier delays or a rainy weekend. Use it to prepare a draft order, then let your own judgement and the sheet's pack maths set the quantities.
Do posts for a peak need hashtags?
They are optional. Instagram has capped hashtags at five a post since December 2025, so choose a few that describe the occasion and the product rather than a long list. On other channels, put the effort into the dates, the bundle names and the price.
Further reads
- Stationery Shop Marketing Ideas: School and Corporate Orders With AI — Another retailer building offers around school and workplace calendars.
- Party Rental Booking Software: Quote and Availability Tools With AI — If you also hire out party equipment, see the booking side.
- Boutique Inventory Management: AI for Size Runs and Reorders — Stock logic for goods that come in sizes and packs.
- How to Write a Customer Newsletter With AI in 30 Minutes — Turn the countdown into an email your customers read.
- AI Marketing for Plumbers: A Worked 12-Month Posts and Offers Calendar — A worked 12-month calendar from a very different trade.
- AI Marketing Automation for Small Businesses: Chocolate Gift Seasons — One chocolate gift-box line taken through Easter, Mother's Day and Christmas, with pre-order reminders, cut-offs and shelf-life labels automated.
- Pop-Up Shop Business Plan: Stock, Rent and Staffing in ChatGPT — One fact block, five prompts and a hand check turn a rent quote, stock costs and a rota into break-even days and a go or no-go rule.
- Art Supply Store Marketing: Workshops and Student Offers With AI — A worked month for an independent art supply shop: two capped workshops, a student-offer email and four brand-of-the-month posts, with the maths done first.
- Baby Store Marketing Ideas: Gift Lists and New-Parent Offers With AI — A gift-list page, a four-email welcome series and milestone offers for an independent baby shop, with prompts and a claims check for safety wording.
- Can AI Help a Convenience Store Drive Sales? Where to Start — A convenience store owner's guide to AI sales tests: find high-margin lines in your till export, run one bundle test first, and read the result honestly.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Google Business Profile Help pages on posts and events (checked October 2026); Instagram's published hashtag limit. Quantities, pack sizes, prices, sales indexes and margins in the examples are illustrative; use your own till data and supplier price lists.