Using Detailed Customer Transaction Data from Farmers' Markets to Analyze Opportunities for Increased Sales

Final report for ONE18-316

Project Type: Partnership
Funds awarded in 2018: $12,214.00
Projected End Date: 04/15/2020
Grant Recipient: Cornell Cooperative Extension- Tompkins County
Region: Northeast
State: New York
Project Leader:
Matthew LeRoux
Cornell Cooperative Extension- Tompkins County
Expand All

Project Information

Summary:

Electronic point-of-sale (POS) systems provide a quick and easy means for farms to gather customer transaction data at farmers markets. With this project, we wanted to test sales data collection with farms at market and determine if and how the resulting data reveals opportunities to improve market sales. We also piloted a methodology for data collection which is documented for others to repeat.

In the summer of 2018, eight Ithaca-area vegetable farms participated in research using POS. Combined, the 8 project farms participated in 16 markets per week for a total of 204 market days observed. Among all farms and markets, the project observed nearly 20,000 customer transactions.

We used the POS application Square on iPad and Samsung tablets. The POS software allows for detailed data collection in just a few seconds. Data captured includes: date, time, unique transaction identification (an alpha-numeric string), items in the order and their prices, order total, and payment type. In addition, we collected data on market traits including weather and the number of vendors (market size).

The participating farms sold primarily vegetable crops including some fruits. Additionally, a few farms sold high-value items including honey, jam, and sauce. Project assistants were provided to farms at market, to train and assist them. While most farmers eventually learned to input sales on their own, a few used project assistants throughout the project, to “shadow” them, entering data as the farmer made sales. Compared to a total cash tally at the end of each market day, we found that farms recorded 85-90% of transactions accurately. Missed transactions and input errors most likely occurred during the peak sales of each market.

The transaction data proved to be as valuable as we hoped. Even before we began analyzing data and providing farms with feedback, one farmer exclaimed, “How did we ever NOT do this?”  The benchmarks that we created from the aggregate data provided a basis for comparison for individual farms. We provide Daily Market Summary reports for farms, but perhaps more useful were Season Market Summary reports created at the end of the project.

 

This data can be analyzed to reveal growth opportunities for market sales. This article is a highlight of early findings from the project. We found that the average customer spent $6.47 per transaction. This represented 1.9 items per transaction with an average item value of $3.38. The highest customer counts occurred in the first hour of markets, with customer counts and transaction size both falling as the market continues.

A surprising data point revealed by the project was the proportion of customers spending a small amount with the farm. We found as many as 45% of customers spent $3 or less in some markets with the group average at 28%. This finding led participating farms to discuss how to alter either unit price or item price as a means to increase customer spending. For example, instead of selling a head of garlic for $1.50, bundling 2 heads together for $3. Such practices would help increase the average customer transaction size, though not necessarily guarantee increased daily sales.

To share this information, a presentation was offered in Ithaca, open to all farms. In addition, a detailed presentation was given to the participating farms. An article was published by Cornell’s Smart Marketing available here: https://dyson.cornell.edu/outreach/smart-marketing-newsletter/

This pilot project was successful and taught participants the value of data collection and analysis. All of the participating farms report that the project changed the way they evaluate their success at markets. All of the farms plan to introduce new market tactics, informed by their data and the group benchmarks, in an effort to influence daily sales. Six of the eight farms plan to use POS in the 2020 market season, with two farms holding out, citing “too busy” as the reason. We recommend that others repeat and build on this project.

Project Objectives:

Objective 1: To identify opportunities for participating farms to increase sales based on consumer transaction data. Our primary objective was to collect and analyze detailed customer transaction data from a group of small scale vegetable farms in the Ithaca, NY area. We provided each farm with a detailed report on their transactions complete with interpretation and suggested methods to increase customer spending.

Objective 2: To pilot and develop a methodology for collecting data on customer transactions. In order to complete our primary objective, we developed and implemented the methodology for programming the POS system and collecting data consistently across all farms and markets.We produced a guide detailing our methods so they are repeatable.

Objective 3: To develop generalized marketing advice based on conclusions drawn from individual farm and aggregated data. We aggregated data and created benchmarks for farmers market sales. We shared recommended marketing changes with participating farms.

Introduction:

Our Marketing Channel Assessment Tool (MCAT) research was among the first to measure performance metrics of farmers markets for small-scale vegetable growers. In that project, we measured each channel’s performance for five factors: sales volume, labor required, profit, risk (of lost sales), and lifestyle fit. We ranked the relative performance of each marketing channel utilized by a farm and created benchmarks for dollars of gross sales/hour of marketing labor for channels. The data-driven approach of MCAT gave farmers information to make informed marketing decisions, specifically, which channels to expand, improve, or drop to increase overall efficiency of marketing labor (dollars gross sales/hour of marketing labor).

Hardesty and Leff (2010) found that the labor to revenue ratio was highest in the farmers market channel. Similarly, our team, LeRoux et al. (2010) and Schmit and LeRoux (2014) found that farmers markets had the lowest overall financial performance rankings relative to other channels. This work led us to look for ways to improve the farmers market channel because it remains popular with small farms. Increased sales at markets can have impacts on farmer qualify of life. Increased sales increase channel profit since nearly all costs for market participation remain fixed. Also, as one farmer expressed in his support letter, increased sales in two markets may allow him to drop a third.

MCAT determined that farmers markets perform poorly for farms, but they are easy to enter and remain popular among farmers. So, what can vegetable farms do to improve the profitability and gross sales/hour of farmers markets? We determined that studying customer spending, measured in dollars/transaction, is the best opportunity for individual farms to make improvements.

This project is a logical next extension of MCAT. While the determinants of average customer transaction size have been evaluated previously using more aggregated data (Schmit and Gomez, 2011; Varner and Otto 2008) or farmer surveys collected during the off-season (FMFNY), we know of no prior farmers market study utilizing transaction-specific data akin to the scanner data utilized in retail stores. If FMs are to provide improved farm returns, further research is needed to investigate the determinants of their success, and more micro-level data on customer transactions is needed. By utilizing transaction-level data from vendors one can ascertain more precisely the nature of individual purchases by consumers rather than relying on aggregate daily/weekly sales and customer number estimates.

Our reports for participating farms will recommend actions which can be categorized similar to work done in the retail industry and used to increase customer spending. Recommendations in the report follow the work of Lobel of SwiftIQ with suggested improvements in five key areas:

1. Localization: Product price and placements,

2. Inventory optimization: market-specific product assortments,

3. Fact-based merchandising: Patterns in consumer behavior inform signage and promotions,

4. Specific, relevant promotions at the market or by email, and

5. Labor management: number of employees needed at a market at any given hour. 

Cooperators

Click linked name(s) to expand

Research

Materials and methods:

In 2018 the project launched and can be described in three main sections: creating the methods, data collection in the field, and data analysis.

Creating the Methods

We created a set of detailed instructions documenting how to collect data in an accurate and consistent manner across all farms. Our team assisted farms in setting up their products on their account in the initial weeks. We then joined them at market to collect data and train them to ring up customers on their own. We worked one or more markets with each farm to assure that set up was correct and that we were capturing the data as intended.

Meanwhile, in the office, directions were written describing Square set up, collection in the field, downloading data to Excel, and handling Excel spreadsheets to create reports for the farmer. These documents were created to make sure that data was collected and handled in a consistent manner, but also so that others wishing to repeat the work have clear instructions. We created a Daily Market Summary report for each successful market day of data collection.

Data Collection in the Field

The project lead and three summer interns conducted the initial farmer training on-site during markets. Once that was complete, interns were dispatched to farmers markets nearly every day of the week, “shadowing” the busiest farms and entering sales into the POS for them. In addition, project staff checked in with each participating farm and handled downloaded data in the office. With eight actively participating farms, each doing multiple markets per week, the data quickly piled up. We focused on data collection at markets from June through the end of September, to capture the “peak season.”

Data Analysis

Beginning in the 2018 market season, we began processing data and creating Daily Market Summary Reports for each farm. The data downloaded from each farm’s Square account is first “processed” which involves adding additional data points, such as number of other vendor, number of items, and weather and by making calculations. The processed data set is then used to make a summary report, which is given to the farm and shows customer spending, item sales, and other totals and averages from the day’s sales.

Early in 2019, all usable market data had been processed and turned into a daily report. Next, we combined daily market sets into seasonal market sets, meaning, one large file with all of one farm’s data from the season for each market they used. From this data set, the project lead created a Season Market Summary Report. These files were also sent to Prof. Todd Schmit for statistical regression analysis. Prof. Schmit combined all of the project data into one file for this.

Season Market Summary Reports were presented to the farms in October 2019 during a presentation. The presentation reported and discussed the range and average of key market metrics including daily sales, customer counts, item value, and average customer transaction size. Each farm received recommendations tailored to their results and marketing techniques were discussed among the group.

Research results and discussion:

We theorized that studying market metrics such as customers/hour and dollars/transaction, would reveal opportunities to favorably influence average daily gross sales (ADGS). Beginning in June 2018, farmers set up the POS system on tablet devices, including each item they were selling and its price. A programmed POS allows the farm to “ring up” a customer quickly and easily and saves the details of each transaction for export to a spreadsheet. As a customer’s order is rung up, the POS software collects: items in the shopping cart, price of each, order total, date and time, and payment method. In addition to transaction details, we observed the weather, number of items that farm had available, and the number of other vendors at each market.

After a season, the data set was large enough for analysis. We used farm data to calculate valuable market metrics including, customers/market, customers/hour, dollars/customer transaction, customer item count, items sold count, average item value and total daily sales.

For this project, we collected sales data from one season, creating a baseline for each farm and the aggregate of all farms in the study. Therefore we have not tested any techniques to increase ADGS. We have basic statistics that serve as useful benchmarks for farms to compare their own market data against. Such a comparison informs farms where they can most benefit from changes.  Below is a summary of observations and suggested (but untested) techniques which have the potential to increase sales.

Customer Counts: Among the 204 market days observed, the average customer count per market was 98 customers. Regardless of market start time or day of the week, the trend for Ithaca-area markets was for the highest customer spending and highest customer counts to occur in the first hour of the market. In fact, we observed anecdotally and from recorded transactions that shoppers are eager to make purchases before the market’s official start time. Therefore, farms might be able to capture more sales by being completely ready for customers before the market’s official start time (when permitted by market rules). Later in the market, customer spending and customer counts drop steadily. This suggests that farms might benefit from trying to engage late-market shoppers with signage, samples, and specials.

Market Size: There has long been debate among farmers market managers and vendors about whether to limit the number of vendors selling similar products. While our research did not directly address this question it did reveal interesting results. The observed markets ranged in size from 11-93 total vendors and 4-15 vegetable vendors. When controlling for all other factors, we found that larger markets increased customer transaction size. This seems to indicate that shoppers at larger markets are prepared to spend more. However, both customer counts and daily sales take a hit at larger markets. We presume this is because there are more vendors from which customers may choose and customers’ desire to “spread the wealth,” making small purchases with several vendors.

Item Value and package size: At many of the observed markets, a surprisingly high proportion of shoppers spent $3 or less with farms. The average item value across all transactions in the project was $3.38.  The proportion of customers spending $3 or less across the 16 markets ranged from 14-45% with the average at 28%. For farms seeking to increase customer spending due to a high proportion of low-spenders, one potential technique is to pack products into sizes priced at $3 or higher and eliminate products priced below $3. For example, instead of selling garlic for $1.50/head, farms might try to pack garlic heads in twos and charge $3. This is not a unit price increase, just an increase in package size in order to increase the minimum a customer will spend with the farm.

Unit price increases: The overall average customer transaction size for the project was $6.47. This average includes transactions with some relatively high-priced value-added products sold by participants, such as honey, jam, and sauce. Prices for individual products varied among the participating farms. Of course, an opportunity for farms to increase customer transaction size might be to simply raise prices but at the risk of reduced sales. While our project did not calculate price elasticities for individual vegetable crops, we do have some indication of consumer willingness to pay. Are there opportunities for farms to increase prices on best-selling items? Depending on the size of the price increase, it is possible to lose some customers (sell fewer units) and still reach the same sales target. If sales drop, at least it results in reduced harvesting, washing, and packing which is a labor savings. However, if sales levels are maintained or drop only slightly, the farm stands to increase revenue.

Items per customer: The average customer across all farms and markets purchased 1.8 items per transaction. For farms looking to increase customer item count, the simplest approach is to try to influence single-item customers to purchase a second item. Our farmer participants brainstormed a few techniques that have potential to accomplish this. One farm introduced a loyalty program, with every $10 purchase earning a stamp on a reward card. Farms also considered offering a variety of items in units that are priced the same. This allows for promotions and signage such as “Any 2 items for $7,” and “3 for $10.” Farmers also considered bundling items into an “all this for $20” pack. Another idea is to offer more varieties and sizes of the most popular crops to incite impulse purchases. A final idea is to suggest items to customers when they cash out such as “do you want to get some garlic to go with these tomatoes?” Anecdotally, farms report that they are often successful with well-crafted suggestions.

Customer spending: While we set out to document the average customer transaction size with vegetable farmers at markets, another metric that we had not previously considered, and which farmers were surprised to learn, was the proportion of customers that spend $3 or less. Consideration of this measure of sales came about while sorting customer transactions while processing data. After sorting transactions by “net sales” and scrolling through the spreadsheet, the large volume of small transactions became apparent. Therefore, full season reports given to each farm at the end of the project included this statistic for their market(s). Farms saw a range of customers spending $3 or less from 14-45% of all customers, with 28% the average. Thought on this percentage then informed discussion about item value, package size, and pricing and ideas to improve ADGS.

Documents associated with project:

Anonymous Daily Market Summary Report2018

Daily Market Log and Instructions sheet

Directions for creating the Daily Market Summary Report for farms – how to summarize data from Excel entries for farmers

Farmer instructions SARE Report version – how farmers can collect their own data

Office instructions SARE Report version – how to have field assistants collect data in the field

Research conclusions:

We endeavored to collect and study detailed data from Point of Sale (POS) software at farmers markets. We piloted the methodology for precise, consistent data collection across multiple farms and markets over time. Taking electronic tablets with POS into the field, we investigated the practicality of collecting transactional data at market and addressed farmer concerns about using it. We piloted data handling and analysis for useful outputs. Data collection and processing methods are documented to facilitate replication of our methods by others. While our data analysis was assisted by Cornell faculty and statistical software, we conclude that farms wishing to replicate this can get valuable results by exporting data into a spreadsheet and calculating summary statistics such as sum and average.

The data analysis revealed benchmarks for valuable market metrics such as, customer counts, item value, and customer spending. In turn, by comparing individual farm results with the aggregate, we learned to identify the best opportunities to increase daily sales for each participating farm. We also brainstormed marketing techniques with the potential to improve market performance, though testing the techniques was not part of this project.

This project realized its goal. In addition to testing the feasibility and utility of POS data collection, the project succeeded by changing how participants define and measure their success at farmers markets. Most participants report intent to adopt POS data collection for all transactions in the coming season and all participants report that they will test new marketing techniques as a result of participation.

Participation Summary
8 Farmers participating in research

Education & Outreach Activities and Participation Summary

2 Consultations
1 Published press articles, newsletters
3 Webinars / talks / presentations

Participation Summary:

15 Farmers
3 Number of agricultural educator or service providers reached through education and outreach activities
Education/outreach description:

We first held a group meeting early in Spring 2018 to orient the farms on how we would proceed. During the market season our team was in contact with the participating farms daily, through email and at the markets. In April 2019 we held an evening Farmer 2 Farmer discussion group open to all area farms. We discussed the early results of the project and presented anonymous and aggregated data.

In this early results discussion, we focused on a few trends that seemed consistent across farms and markets. One trend was for the highest customer counts and customer spending to occur early in the market’s hours. The second trend was that, for many farms and markets, what seemed like a high percentage of customers spent $3 or less with the farm. We used these trends to guide a discussion on the best opportunities to increase average daily gross sales (ADGS).

During the 2019 market season we did not collect data; however, we continued to work with the 2018 data. As a result, we held another discussion group in October 2019 to go over the project results. This meeting included representatives from all of the eight participating farms as well as the manager of the Ithaca Farmers Market. Non-participating farms were not invited as we felt the information discussed was a privilege reserved for participating farms. Participating farms were given their Season Market Summary reports at this meeting as well as presentations by Matt LeRoux (project lead) and Prof. Schmit.

We wrote a “Smart Marketing” article which was published at https://dyson.cornell.edu/outreach/smart-marketing-newsletter/  for the January 2020 edition. Smart Marketing is circulated to regional media and farm publications.

Presentation at the NOFA-VT Winter Conference, Burlington Vermont, February 15-16. 2020 was with 7 farmers and 3 market managers (one counted as a farmer, too).  Current PPT attached.Square project Final Powerpoint

Learning Outcomes

9 Farmers reported changes in knowledge, attitudes, skills and/or awareness as a result of their participation
Key areas in which farmers reported changes in knowledge, attitude, skills and/or awareness:

At the end of the project, a survey was circulated to the eight participating farms. All the farms report that the project changed how they assess their success at market. The data revealed different opportunities for growth for each farm. When asked to list two specific actions they would take to increase sales, answers were varied including, change product size to raise item price, arrive at market earlier, alter quantities and varieties of crops, experiment with different product arrangements, and increase unit prices. The metrics that farms plan to study in the coming 2020 season include: customer counts, pricing, items/customer and best/worst selling items.
After the presentation on the final project results only two farms report that they will not adopt regular use of POS citing “too busy” as the reason. It was rewarding to observe several of the participating farms, perhaps categorized as “recordkeeping-averse,” adopt POS after the project. For these farms, perceptions of inconveniencing customers by using a tablet to “ring up” orders were overshadowed by the power of the captured data. Two weeks into the data collection period, one previously wary participant exclaimed, “How were we ever NOT doing this?” That farmer went on to recruit our eight participating farm for us.

Project Outcomes

9 Farmers changed or adopted a practice
2 Grants applied for that built upon this project
Project outcomes:

This project reached the goal of piloting customer transaction data collection, creating a methodology to do so, and discovering the value of data for creating specific techniques with the potential to increase daily sales at markets. To this end, we produced detailed instructions for data collection and processing, including directions for creating the Daily Market Summary Report. Combining all farm data also created benchmarks for key market metrics such as customer spending, average item value, and customer counts.
Participating farms received individualized recommendations suggesting the best opportunities to increase average daily gross sales. Additionally, we held two presentations on the project, one open to all farms, the other just for the participants. A Cornell Smart Marketing article was written about the project, to be published online and shared with agricultural media in January 2020. Project manager Matt LeRoux also presented on the project at the NOFA-VT Winter Conference in February 2020.
The best outcome of the project was the adoption, by farmers, of using POS technology for the long term. Six of the eight project farms report that they will use POS to record all transactions in the 2020 market season. Of the six, only one used the system before this project. This outcome is particularly meaningful because it demonstrates that participants recognize the value of recording and analyzing transaction data. Even without statistical regression, farmers will be able to review and summarize their sales data and, perhaps most importantly, test new marketing techniques for an impact on daily sales.

Assessment of Project Approach and Areas of Further Study:

When we set out to conduct this project, we knew that some farms would take some convincing. We initially recruited 10 farms but it quickly became clear that two farms would fail to follow through, despite interest. To that end, we budgeted to buy and loan iPads. We also budgeted for summer interns to “shadow” farmers at market, recording their sales for them. We also attempted to get all eight farms up to speed at the beginning of the project and to coordinate the use of four iPads among seven of the eight farms.
Once in progress, we learned that supplying iPads is important to some participants who may not own the technology. Our project partner, Todd Schmit, provide extras funds for the purchase of an additional four tablets, this time, Samsungs. The interns, who were all Cornell undergraduates, recommended that we buy Samsungs because they are less expensive but later reported that they did not work with the POS app, Square, as well as the iPads. This resulted in lost data.
Providing interns to the farms was critical to our success with the project because 7 of the 8 were unfamiliar with Square. With nearly every farm, the interns were an appreciated asset, but not completely necessary after the farmer and their staff became familiar with the POS. One noteworthy exception was one farm that refused to use the POS, or even direct their hired staff to use it. When repeating this project, it would be great to have interns available, but not completely essential.
If I were to do this again, I would not bring all farms on-board at once. Instead, I would stagger them, getting 1-2 all set, then moving to the next few. Additionally, I would not work with farms that need convincing, but rather, let motivated farms self-select. This would eliminate the need for student interns, though they will always be an asset if available. Tablets should be made available for loan as farmer possession of a device is not an indicator of interest. Based on our experience with tablet performance, I would only use iPads, despite their higher cost.
Given our experience with this project, I recommend that other organizations and individual farms conduct customer transaction data in a similar fashion. Provided farms are dedicated to using POS in a consistent manner for all customer transactions, the resulting data set will provide insights to improve daily sales. For further study, I would expand this project to additional types of farms, a broader range of farmers markets, and even other direct to consumer channels.

Any opinions, findings, conclusions, or recommendations expressed in this publication are those of the author(s) and do not necessarily reflect the view of the U.S. Department of Agriculture or SARE.