Analytics, Visualisation and Dashboards in Finance and Costing

This chapter turns the vocabulary built in the previous chapter into two things a CMA actually produces: a statistical model that estimates cost or predicts an outcome, and a visual — a chart or a dashboard — that communicates it honestly to someone who was not in the room when it was built.

1. Correlation — how strong is a relationship?

The correlation coefficient (r) measures the strength and direction of a linear relationship between two variables, and it always falls between −1 and +1. A value near +1 indicates a strong positive relationship (both variables rise together), a value near −1 indicates a strong negative relationship (one rises as the other falls), and a value near 0 indicates little or no linear relationship.

$r=\frac{n\sum xy-\sum x\sum y}{\sqrt{\left[n\sum x^2-(\sum x{)}^2\right]\left[n\sum y^2-(\sum y{)}^2\right]}}$

Correlation does not prove causation — two variables can move together closely because one genuinely drives the other, because both are driven by a third factor, or purely by coincidence — and a good analyst always checks whether a plausible causal story exists before acting on a correlation, rather than treating a high r value alone as sufficient justification for a decision.

2. Regression — from a relationship to a usable estimate

Simple linear regression fits a straight line through a set of paired observations, expressing one variable (the dependent variable, y) as a function of another (the independent variable, x), and it is the direct statistical generalisation of the high-low method already used in cost accounting to split a semi-variable cost into its fixed and variable components.

$y=a+bx,b=\frac{n\sum xy-\sum x\sum y}{n\sum x^2-(\sum x{)}^2},a=\bar{y}-b\bar{x}$

In a cost-estimation context, $b$ is the variable cost per unit of activity and $a$ is the estimated fixed cost — but unlike the high-low method, which uses only the two extreme observations, regression (the "least squares" method) uses every observation in the dataset, which makes it a more reliable cost-estimation technique whenever more than two data points are available, since it is not distorted by whichever two points happen to be the highest and lowest.

The coefficient of determination ($R^2$), the square of the correlation coefficient, states the proportion of the variation in y that is explained by x — an $R^2$ of 0.81 means 81% of the variation in the dependent variable is explained by the independent variable in the model, leaving 19% explained by other factors not captured in this simple two-variable model.

3. Choosing the right visual

A chart's job is to make a comparison, a trend or a relationship visible at a glance, and choosing the wrong chart type for the underlying data story is the single most common analytics-communication error.

A pie chart with more than about six slices, or a 3-D chart of any kind, are the two most common visualisation mistakes: too many pie slices make relative sizes impossible to compare accurately, and 3-D effects distort the true proportions of bars or slices through perspective, actively misleading the reader rather than merely failing to help them.

4. Dashboards — turning charts into a management tool

A dashboard consolidates several key performance indicators (KPIs) onto a single screen, and its design is judged on whether a manager can absorb the state of the business at a glance without hunting through separate reports.

Good dashboard design follows a small number of repeated principles: show the most important KPIs first (typically top-left, where eyes land first); use consistent colour coding (such as red/amber/green against a target) rather than decorative colour; provide drill-down capability so a summary figure can be expanded into its underlying detail on demand; and avoid chart clutter — a dashboard crowded with a dozen small, hard-to-read charts communicates less than three or four well-chosen ones.

In finance and costing specifically, dashboards commonly track: budget-versus-actual variance by cost centre (usually a bar chart with a variance highlighted in red/green), a working-capital cycle trend (a line chart of days sales outstanding, days inventory outstanding and days payables outstanding over time), and a cost-driver analysis (a scatter chart of activity level against cost, with the regression line overlaid to show the fitted cost-estimation model).

Worked Examples

Example 1. Machine hours and total overhead cost were recorded for five months: (100, ₹1,200), (150, ₹1,500), (200, ₹1,800), (250, ₹2,000), (300, ₹2,300). Estimate the fixed cost and variable cost per hour using the least-squares regression method. (Σx = 1,000, Σy = 8,800, Σxy = 1,865,000, Σx² = 225,000, n = 5)

$b=\frac{5(1,865,000)-(1,000)(8,800)}{5(225,000)-(1,000{)}^2}=\frac{9,325,000-8,800,000}{1,125,000-1,000,000}=\frac{525,000}{125,000}=4.2$

$a=\bar{y}-b\bar{x}=1,760-(4.2\times 200)=1,760-840=920$

Estimated fixed cost = ₹920 per month; variable cost = ₹4.20 per machine hour.

Example 2. A correlation coefficient between advertising spend and sales revenue is calculated as r = 0.85. Interpret this figure, and state one caution about acting on it.

r = 0.85 indicates a strong positive linear relationship between advertising spend and sales revenue — as one rises, the other tends to rise too. The caution: correlation alone does not prove advertising spend causes the sales rise; a third factor (such as a seasonal demand peak that also drives up ad spend and sales together) could be responsible for both.

Example 3. If R² = 0.81 for the regression in Example 1, what proportion of the variation in overhead cost is explained by machine hours, and what proportion is not?

81% of the variation in overhead cost is explained by machine hours; 19% is explained by other factors not captured in this model.

Example 4. A manager wants to show the market-share split of five competing brands. Which chart type is most appropriate, and why?

A pie chart, since there are only five categories (within the readable limit) and the data story is genuinely about share of a whole rather than a trend or a detailed comparison.

Example 5. A finance dashboard uses a 3-D pie chart with eight slices to show departmental cost allocation. Identify two visualisation errors in this design and explain the harm of each.

(1) Eight slices exceed the roughly six-slice limit for a pie chart to remain readable, making relative sizes hard to compare accurately. (2) The 3-D effect distorts the true proportions of the slices through perspective, actively misleading the viewer about which slices are actually larger.

Example 6. Explain, using the regression method, why estimating fixed and variable cost from all five months of data (as in Example 1) is generally more reliable than the high-low method using only the highest and lowest observations.

The high-low method uses only two data points (the highest and lowest activity levels), so its estimate is entirely determined by those two observations and can be distorted if either is an unusual outlier. Least-squares regression uses every observation in the dataset, giving each data point influence proportional to how well it fits the overall pattern, which produces a more statistically robust estimate whenever more than two observations are available.

Example 7. A CFO wants a single-screen view to track whether each of six regional teams is meeting its quarterly cost-control target. Describe two dashboard-design principles this screen should follow.

(1) Use consistent red/amber/green colour coding against each region's target, rather than decorative colours, so status is visible at a glance. (2) Provide drill-down capability so the CFO can expand any region showing an amber or red status into its underlying cost-centre-level detail without leaving the dashboard.

Summary

The correlation coefficient (between −1 and +1) measures the strength and direction of a linear relationship, while regression (the least-squares method) fits a usable estimation line, y = a + bx, generalising the high-low cost-splitting method to use every observation rather than just the two extremes; R², the square of r, states the proportion of variation explained by the model.

Choosing the right chart type — bar for category comparison, line for trend, pie only for a few shares of a whole, scatter for relationships, histogram for distribution — is the foundation of honest visual communication, and both excessive pie slices and 3-D effects are well-known ways a chart actively misleads rather than merely fails to help.

A dashboard consolidates key KPIs on one screen using consistent colour coding, prioritised layout, drill-down capability and minimal clutter, and in finance and costing specifically is commonly built around budget-variance tracking, working-capital cycle trends, and cost-driver scatter analysis with a fitted regression line.