I do not own this photo
Identify trends and predict future floral sales data.
The goal of this report is to help the owner of the floral shop to understand the trends and patterns in sales, and use a predictive machine learning model to make predictions for the future sales of the shop. This uses real data from a fine floral shop owner. For the sake of privacy I will not use the actual name of the shop.
The data was provided to me by the shop owner, who uses floranext and Bloom Nation as their sales operating systems. I was provided sales reports downloaded from these sites in csv format as the data i will be using. The Bloom Nation data set was limited in what it could provide as far as useful data relevant to our project goal because it does not require the customer to fill out every option availble to them when making a purchase order.
I downloaded the data sets and combined them together using SQL Lite. I then cleaned the data, addressed the null values, and prepared the data for visualization and then pipeline construction to prepare it for modeling. I used Tableau for visualizing the data after cleaning it because Tableau has so many chocies for making vivd, powerful visualizations. Then, I preprocessed the data and formated it to make sure it was ready for modeling. I tried several machine learning models, the best of which was a Random Forest reggression model to make predcitions based on the sales data I was provided with.
These visualizations helped to identify the trends and patterns for sales throughout the different months of the year for five and a half years. The first of which helps us to see the trends in sales by Quarterly segments for each year organized by the order date. This can help the shop owners to identify what months out of the year they may need temporay help with arrangements and deliveries, or for slower months cut back on ordering extra stock.
- As expected, the months that have major holidays (Christmas, Valentine's day, and mother's day) are the busiest year over year.
- March, April, June and July seem to consistantly be the slower months year over year.
This second visulization shows the order method distribution for each month. This can help the owner's identify what months are more popular for walk-ins and deliveries.
- There are noticable spikes in the walk-in traffic for November and December yearly, obvisously due to the holidays.
- There are overall way more phone orders than walk-in traffic regaurdless of the time of year.
- Florist-to-florist category was a recent addition as an order option which is why we see almost no representation on the visualization.
This shows a comparison of the different payment methods customers used to pay for the different order types.
- As exspected, most customers use a creditcard or the website for a delivery and pickup.
- Credit cards are the most popular across every category, and cash seems most popular for taken and pick-up orders.
This shows an average grand total per order monthly so that the show owner can see trends in the average amount a customer spends depending on The Order Type
- The average for wire-out is so high due to the fact that it represents the shop owner buying flowers in-bulk for another floral shop. They are normal large orders because there are frequently discounted rates on a higher quantity of flowers.
- The lowest average total per order overall is taken which represents a customer walking in and taking the purchased prodcuts out of the store themselves.
- This compairson provides helpful insight that the owner can use to construct pricing strategies, optomize inventory, gain operational efficiency, and serve as a guide for scheduling promotional efforts.
- The 4th quarter consistantly has the most sales year over year, due to a consistant spike in sales for the holiday season.
- December is consistantly the highest selling month (due mostly to christmas and other winter holidays) followed by May (Mother's Day).
- Months that have major holidays (Valentine's day, Mother's Day, and Christmas) have the biggest spike in sales year over year.
- The most popular order method is by phone, followed by Walk-in. This is because most customers want to talk to a designer about customizing their arangement.
- The most popular payment method overall is credit cards for all orders including orders made through the website.
- The most popular Order type is delivery, followed by taken (Products taken out of the store by walk-in customers), and then pick-up (customer picks up in-store, usually at a later date).
- Slower months are consistantly January, April, June, and July year over year.
Throughout my discovery, I diligently constructed, executed, and fine-tuned a range of diverse models. This model demonstrated remarkable performance by effectively explaining 91.2% of the variance present in the test data. This high R-squared score reflects the model's proficiency in capturing and interpreting the underlying patterns within the test dataset. On average, its predictions deviated by a mere $12.7 from the actual values, and even the largest prediction error was limited to just $51.8. This demonstrates the model's remarkable overall accuracy, with room for minor refinements to further minimize prediction discrepancies. Notably, it's important to emphasize that the model exclusively utilizes the sales data provided by the shop owner.
Diving deeper into the model's performance, it's crucial to highlight the role of the coefficient of determination (R-squared) in assessing how effectively the model predicts sales. R-squared serves as a vital metric, revealing the model's capability to elucidate the variability in sales data. In this context, the 91.2% R-squared value signifies that the model comprehensively captures 91.2% of the variation in sales, illustrating its potency in explaining and forecasting sales trends. This underscores the model's proficiency in providing valuable insights and making precise predictions, substantiating its practical applicability for real-world business scenarios.
One recomendation I have for the shop owner would be to make an modification to the website that requires certain information like ocassion, product description, and recipient to be entered to place an order. This way, there would be alot more specific data about product popularity to analyze that gives insight on the most popular ocassion, and who the most common senders and recipients are. This would have to be a conversation between the shop owner and the company that the website is run through. If accomplished, this could make a major impact on the amount of usable data available and provide much more insight about specific trends and patterns for what products are most popular for each time of the year and what is not selling. In addition, because alot of the volume is associated with floral arrangements, finding a method to keep track of which flowers were used for which arrangements would provide insight as to which flowers were used more often at certain times of the year.
This report was limited to the data I was provided, which was limited to only sales statistics because of the inconsistent website data. There are many customers that purchase things from the store in-person just by browsing throughout the store, there are also many customers who have never been to the store but repeatedly order from the website. It is located in a rather rich part of town where alot of elderly people live. It is also next to a popular local coffee shop which brings in a decent amount of foot traffic consistently during the week on fair weather days. The shop is generally closed on sundays (There are rare exceptions) and usually open from 9:00 am - 4:30 pm; but these hours are subject to change via the owner's descresion. The patterns and trends I've help to bring to light, may help the owner to make descisions about opening and closing times and wether or not to open on sunday depending on the time of year or schedule of holidays in a given year. The next steps are to show the shop owner my findings of the trends and patterns in sales so that he may optomize things like payroll during slow seasons, product inventory before and after busy months, what the most popular order methods and types are, scheduling hours of operations for specific times of the year and the most profitable months for specific quarters of the year.