An Introduction to Inventory Management
The day-to-day pressure of inventory management can feel like an endless struggle. Keeping track of your current levels of stock, along with estimating how much stock to order in the future, is difficult to navigate for new business owners.
By using stock forecasting models, you no longer have to guess when to order more or less of particular items in your inventory. Avoid worrying about seasonal changes, monthly fluctuations, and market shifts in demand when you select the right forecasting model for your business.
However, different forecasting techniques are used best in different business situations. By choosing wisely, you can avoid gaps in your inventory and times of overstock.
The simplest version of forecasting is naive forecasting. This method compares past data of a given time period and uses it to predict future results. For example, if you sold 100 widgets last month, it would recommend you keep 100 widgets in stock for this month.
However, naive forecasting does not take any market variables into account. You would not factor in holidays, competitors, or shifts in customer shopping habits. This makes naive forecasting the least reliable inventory forecasting technique.
It is most useful for checking your work against other methods of forecasting. Used as a general baseline, it can point out computation errors or other outliers that you would need to be aware of when managing your inventory.
To use naive forecasting, determine the time frames that you wish to compare, such as month-to-month or the same month from each year. Then you simply carry over the inventory numbers from that time period and apply them to the current month. You can then get a clear picture of what your typical stock amounts should look like, and note any unusually high or low numbers.
Your inventory does not exist in a vacuum. There are numerous outside influences that can cause fluctuations in your desired inventory levels. For example, new product releases might cause high customer demand for that model, while older versions would languish on the shelves.
Demand forecasting attempts to take these outside variables into account, providing a more nuanced and accurate inventory assessment. It uses many different methods to provide a well-rounded picture of the market, which can be broken down into quantitative and qualitative techniques.
By gathering multiple perspectives of the market, you can make informed decisions about which items to stock up on, and which items to minimize in your inventory.
Quantitative techniques take past data and combine it with mathematical formulas to determine future performance. These calculations often include looking at outside influences, such as product launches, holiday seasons, and other demands on the market. Frequently, these formulas look at all of the available past data, to provide more accurate insights.
Some examples of quantitative techniques include:
- Exponential smoothing
- Box-Jenkins models
- Time series models
- Seasonal indexes
- Causal models
- Moving averages
- Data mining
These techniques rely on having sufficient data to make a reasonable assessment. In cases where this data is not available, such as for new businesses or new products, you may have to rely only on qualitative techniques. You may also choose to use both quantitative and qualitative techniques together, for a more well-rounded perspective.
Qualitative techniques are a far more subjective viewpoint on the market factors at play in your business, yet they often provide sharper insights. This method relies more on the educated guesses based on experience and knowledge of how the market works. By looking at the human element within the market, you will become aware of when the demand trends are ready to shift.
Some examples of qualitative techniques include:
- Delphi method
- Prediction method
- Historical life-cycle analogy
- Game theory
- Executive opinions
- Consumer surveys
- Sales force composite
When used in conjunction with quantitative techniques, you can refine your inventory levels much more precisely than using a single method alone. For example, if consumer surveys indicate that curtains are much more popular than blinds right now, you would know to adjust your inventory levels accordingly, despite what last year’s sales numbers might indicate.
Examples of Forecasting Techniques
Familiarize yourself with some of the more commonly used forecasting techniques, and you will begin to understand which ones might be a good fit for your business. Remember that there are multiple variations available, so feel free to experiment to find the right ones for you.
The moving average is a mathematical formula that uses past sales volume data to predict future trends. By drawing from your data over a fixed time period, for example, monthly sales totals from the year-to-date, you can draw reliable conclusions about next month’s sales volume.
As a version of a time series model, it keeps the data fresh by not including outdated data from before your chosen time period. For example, in a 12-month analysis, the 13th month would be excluded. Next month, the current month would be included, and the 12th month would fall away.
Since it is a quantitative technique, the moving average doesn’t take into account any outside market variables. However, this technique is flexible enough to be used in a variety of applications, for both long and short-term demand forecasting.
Exponential smoothing takes the moving average technique a step further, by weighing certain data points with what is known as a smoothing constant. This is a value between 0 and 1, added to the moving average formula, to indicate how relevant you believe that the prior data is to your forecast.
For example, a company might find that more recent sales data is more relevant, so they believe that it should be weighed more favorably. By adding the smoothing constant, your results will be more tailored to your company’s market positioning.
The Box-Jenkins model predicts data within a time series. It makes the time series stationary by measuring the difference between data points. This makes it easier to determine seasonal differences and trends, improving the quality of your forecasts.
This model forms the basis of many ARIMA forecasting models, or Autoregressive Integrated Moving Average. Briefly, ARIMA means that the model includes:
- Autoregression – it relies on the relationship between an observation and related lagging observations;
- Integrated – making the time series stationary by subtracting an observation from a previous time step observation;
- Moving average – it relies on the dependent relationship between an observation and its related residual errors, and applies to lagged observations.
While there are programs available to help with the calculations involved, the Box-Jenkins model requires making some judgement calls to define the time series.
Time Series Model
A time series model allows you to forecast future sales based on a similar time frame in the past. By evaluating sales at the same time last year, it is easy to determine which items will do well seasonally. However, this model can be used for any time frame, such as weekly or monthly sales volumes.
Time series models allow for considerable flexibility, as you can account for unusual events that don’t occur on a regular basis. For example, the winter Olympics only occurs once every four years, and could influence sales during the Olympic season, but no other times. By adjusting for irregular occurrences, you will obtain a more accurate forecast.
It is also possible to filter out random sales fluctuations, which will allow you to see true variations in buying patterns. The resulting forecasts smooth out fluctuations, so you are not at the whims of last year’s trends.
This model builds on the time series model, but its purpose is to determine how much demand will increase when compared to normal levels of demand during a given season or other time period. With this information, you can determine if sales are increasing because of the season itself or because of an increase in overall demand.
First, you must determine what is the normal, or average, demand. This is given a value of 1.00. Then you can compare the seasonal demand, and create a value known as the seasonal relative. So, if sales are up 25% this season, the seasonal relative would be given the value of 1.25. Then you can see if this season has a similar demand to last season, or if there is a sudden increase or decrease.
The causal model looks at independent factors to determine when to expect an increase in demand. While time is a factor that is often included, it is only one piece of the puzzle. Instead, the causal looks at many possible reasons for fluctuations in demand.
For example, if you sell fertilizer, you might expect an increase in demand during the spring and fall months when buyers do most of their gardening. However, a sudden change in the weather might prompt buyers to start their gardening earlier or later, which will affect your overall demand. Any outside variable can be addressed in a causal model, such as stock market fluctuations, company promotions, or new trends.
Data mining is the process of examining existing data, uncovering existing patterns within the data, and using that data to influence future business decisions. It uses mathematical algorithms to identify trends and predict future scenarios.
One possible scenario you might encounter while data mining is noticing that your buyers often purchase the same products repeatedly, at set intervals, such as with perishable products. You might then decide to remind them that it is approaching time for them to reorder each month, or you might choose to offer a subscription service so they do not have to reorder manually each time.
The data mining cycle includes:
- Defining the problem well, so you can identify when you have solved it.
- Collecting data that correlates with the problem, for ease of analysis.
- Building a model that addresses the problem and will help identify potential solutions.
- Applying the knowledge gathered to see if it solves the problem.
Data mining should be used often to identify solutions to existing problems and point to new areas of business growth.
The delphi method attempts to reach a consensus opinion by sending questionnaires to a select group of experts. The results are then gathered and summarized by a facilitator, and another round of questionnaires are sent out. During each round, the experts are allowed to change their opinion, until an overall decision has been reached.
This process was designed to remove the chance that any one individual’s biases would take over the decision-making process, or that vocal members would dominate the discussion. It works well with long-term forecasting, as industry experts often have insights that cannot be determined solely by past data.
The prediction method, also known as predictive analytics, uses patterns found in past data to determine future trends and opportunities. Frequently, scoring is used to identify which customers are more likely to participate in the future trend or opportunity, which businesses can use to target their sales and marketing efforts more effectively.
For example, a predictive model might indicate that customers who pay with a credit card are more likely to seek out other credit-based opportunities, such as an in-store payment plan. Another possibility might be that customers who buy sporting goods during the summer are more likely to purchase winter sporting goods as well.
By identifying these relationships, you can easily assess what offers and products will appeal to your customers and can encourage sales opportunities that might otherwise have gone unnoticed.
Historical Life-Cycle Analogy
The historical life-cycle analogy takes past purchasing decisions and applies them to related new products. This allows retailers to confidently estimate the sales of a new product line.
This method can be very useful when you are introducing a product that is similar, but not exactly the same as what you have previously offered. For example, if you are selling a family-style tent, you can look at the sales of the larger tents you have already sold to get a sense of the potential demand.
Game theory studies how rational individuals act within a given situation, using mathematical models. While used in a variety of applications, in economics it is particularly useful for evaluating how people will make buying decisions in a competitive setting, such as if they are purchasing products to improve their own business.
Besides the normal buying journey, the customer also evaluates whether or not their competitor is likely to buy the same item or similar, and how that would factor into their own success. The customer could also choose to cooperate with their competitor if they feel that they are trustworthy, which would benefit them both.
Game theory is useful for high stakes or complicated financial situations and can account for statistical outliers in some cases. Software can help determine the eventual outcomes and how the individuals within the situation might act, based on prior experiences.
Rather than analyzing all of the data on your own, you can rely on the executive opinions of your trusted advisors. Typically, the head executives of each department analyze the data on their own and make recommendations based on their own experience and area of expertise. Then they review each others’ opinions and combine it all into a single action plan.
This is frequently used to have a reliable result with less time spent overall. It works well in sales, as individuals in different fields may spot different sales trends, providing a comprehensive overview overall. They also have years of experience to draw from, and can often pinpoint shifts in the market.
By asking consumers directly, you can get a sense of what they are looking for, and the anticipated quantities that they wish to buy in the future. Consumer surveys provide these answers in the customers’ own words, which can fuel future marketing promotions.
There are three methods to conducting consumer surveys:
- Complete Enumeration – By asking all of your customers exactly what they plan to purchase, you get a single answer by totaling their responses. However, it is not always possible to survey each one, and customers often have to guess what they will buy in the future.
- Sample Survey – By only surveying a selection of the total customer base, it is easy to extrapolate the potential future sales. If the sample isn’t large enough or accurately representative of all customers, then the results can be skewed.
- End-Use – With this method, you attempt to determine who are the end users for your products and whether your products are often sold to someone who is not the end user (such as materials used to create a new product). You then look at industry trends and forecasts to determine final consumption.
These customer surveys allow you to understand your customers’ buying habits more clearly, and point out any fluctuations in demand.
Sales Force Composite
The sales force composite method relies on the projections of salespeople within their respective regions, as they anticipate future demand. This shows variations in demand by region, and allows you to make adjustments as necessary.
However, if a salesperson makes an inaccurate prediction, there is a risk of having too much or too little inventory on hand. They may also be unaware of shifts in the market on a national level and may miss crucial information.
Inventory management is a never-ending process, with multiple factors affecting the overall outcome. With a good understanding of the various forecasting techniques available, you will have the tools you need to effectively manage your inventory without long periods of overstocked or under-stocked goods.