Introducing the #AISummer2023 series, a continuation of the original #30DayAIChallenge in Spring. As the field of AI evolves, this series delves deeper into the latest advancements in AI tools and capabilities. It represents not just my practical tinkering over the Summer, but a metaphorical ‘summer’ of AI growth and evolution.

Given the power of ChatGPT’s Code Interpreter (now rebranded “Advanced Data Analysis“), I wanted to build on a past couple of tests e.g. the one where we analyzed sales for a Pizza restaurant to maximize profitability. For today’s experiment, I wanted to look at a typical Food & Beverage venue that sells a combination of food and drinks per customer. Hunting around for a suitable dataset, I found this really simple file for a Bakery with a day’s worth of sales transactions.

Bakery Case Study: What does the business look like today?

Here’s some basic analysis run on the data to understand key trends:

• Count: 1089 sales transactions are recorded (300 unique transactions).
• Mean: The average sale amount is approximately $4.08.
• Std: The standard deviation is approximately $4.34, indicating the variability in the sales amount.
• Min: The minimum sale amount is $0.79.
• Max: The maximum sale amount is $15.95.
• Quartiles:
  • 25% of sales are below $1.25.
  • 50% of sales (median) are below $2.75.
  • 75% of sales are below $3.50.

What’s Popular? How Do We Make More Money?

Certain items are frequently bought together. For example:

• “Truffle Cake” and “Gongolais Cookie” are bought together 22 times.
• “Lemon Cake” and “Lemon Tart” are bought together 17 times.
• “Strawberry Cake” and “Napoleon Cake” are also bought together 17 times.

Promotional Ideas

Based on the insights gathered, some promotional strategies to consider:

1. Bundling: Offer bundle deals on frequently bought-together items, like “Truffle Cake” with “Gongolais Cookie” or “Lemon Cake” with “Lemon Tart.”
2. Upselling: Use point-of-sale promotions to upsell drinks with popular food items, considering that food items are more commonly bought.
3. Discount Tiers: Implement a tiered discount system based on transaction value or the number of items in a transaction. For example, offer a 10% discount for transactions over $20.
4. Promote High-Revenue Items: Use in-store signage or online advertising to promote high-revenue items like “Truffle Cake,” “Casino Cake,” and “Opera Cake.”
5. Happy Hour for Drinks: Since drinks are less popular, consider a “happy hour” promotion to boost drink sales.

Creating a Profit and Loss Statement

We’ll make some assumptions to have ChatGPT generate a P&L for this Bakery:

• The data provided is only for a day and the Bakery is open 6 days a week.
• The Fixed Costs for this small Bakery including labour (one man operation) and all utilities, rent, etc. is about $200,000 per year.
• The average gross margin on food is about 20% and on drinks is 50%.

Here’s the generated P&L:

Show Me The Money! Which Promos Will Net Us More Profit?

Let’s reconsider the promotional ideas we previously discussed with an eye on how this may impact profit:

1. Bundling: This could encourage customers to buy more items per transaction, increasing the average transaction value. However, it could also marginally reduce the gross profit percentage due to the discount on the bundled items.
2. Upselling: Promoting drinks alongside popular food items could boost drink sales, which have a higher gross margin (50% compared to 20% for food). This could improve the overall gross profit percentage.
3. Discount Tiers: Offering discounts for higher transaction values could encourage customers to buy more, but it might also reduce the overall gross profit percentage.
4. Promote High-Revenue Items: Promoting high-revenue items could increase the average transaction value and potentially the net profit, assuming these items have a decent margin.
5. Happy Hour for Drinks: This would likely boost drink sales, which have a higher gross margin, thus improving the overall gross profit percentage.

Best Promotion to Explore?

The Upselling of drinks alongside popular food items seems to be the most promising. Drinks have a higher gross margin (50%), and the bakery’s current sales from drinks are relatively low. By increasing the sales of higher-margin items, the overall gross profit percentage—and thus the net profit—could see a significant improvement.

Upselling Scenarios

Let’s consider the following scenarios:

1. Low Success: 10% of food transactions result in the sale of an additional drink.
2. Moderate Success: 20% of food transactions result in the sale of an additional drink.
3. High Success: 30% of food transactions result in the sale of an additional drink.

Tangible Targets for Staff

Based on the scenarios:

• For a 10% success rate, the staff needs to upsell approximately 29 more drinks each day.
• For a 20% success rate, the staff needs to upsell approximately 58 more drinks each day.
• For a 30% success rate, the staff needs to upsell approximately 88 more drinks each day.

Drinks to Focus On

The staff should focus on upselling the following drinks, listed in descending order of profitability:

1. Vanilla Frappuccino
2. Lemon Lemonade
3. Raspberry Lemonade
4. Orange Juice
5. Chocolate Coffee

Best Food-Drink Combinations

Here are some of the best food-drink combinations based on existing sales data:

1. Chocolate Cake with Chocolate Coffee
2. Apple Danish with Cherry Soda
3. Casino Cake with Chocolate Coffee
4. Lemon Cookie with Lemon Lemonade
5. Apple Tart with Cherry Soda

These combinations are good starting points for the staff to focus on when trying to upsell drinks with food items.

Editorial Note

It took about 15-20 minutes of analysis to figure out a way to potentially make $15-45k more for this little Bakery. All the above work was done by ChatGPT. Clearly, with item sales data and a P&L from an actual operating restaurant, we could do this even better and more accurately, plus test different promos over time. The point of this experiment is to showcase the power and versatility of this tool.

Considering a ChatGPT subscription costs $20 per month, assuming a day’s usage the cost of generating these insights was a mere $0.66 (66 cents).

BONUS – An AI Generated Promotional Flyer!

To take this a step further, here’s a little promo, created using Canva (free) and Midjourney (which costs $8-10 per month…so roughly another 26 cents in cost. Or you can use free tools like Stable Diffusion, Dall E, Ideogram, Leonardo, etc. to generate images or ideally use actual product photography if available). ChatGPT, of course, does an excellent job of coming with suggested marketing copy. All this took another 10-15 minutes.

Wrap Up

There you have it! About half an hour of work, less than $0.92 in cost and a way to potentially drive an additional $15k+ in annual venue profit, with an actionable promo plan and daily targets to boot!

It goes without saying that when you use mainstream (especially free) AI tools, please be mindful and avoid using overly sensitive or proprietary data. With ChatGPT introducing their Enterprise option though, it feels like some of these tools may become a lot more suitable and widely available, even in larger organizations that choose to deploy them. I’m sure Microsoft CoPilot isn’t far behind…and Google has just released Duet AI more widely for Workspace.

What did you think of this experiment? Anything we could do better? Have you tried something similar? Leave a comment or DM to let me know.