What it means
Correlation is a summary statistic that compresses the relationship between two data series into a single number, usually written as r. Finance and commercial teams use it constantly because most business questions involve pairs of numbers that may or may not move together, such as advertising spend and revenue, or interest rates and property values.
The number tells you the strength and direction of the link, but nothing at all about its size in dollars. The practical value shows up in planning and risk management.
If two revenue streams are highly correlated they will both suffer in the same downturn, so the diversification you thought you had is largely an illusion. If they are weakly or negatively correlated, holding both smooths the overall result and makes cash planning far easier.
In practice almost nobody calculates correlation by hand, because a spreadsheet does it with a single function. What matters is knowing which pairs of numbers are worth testing, how many observations you have behind the answer, and whether the relationship is genuinely a straight line.
A handful of data points can produce an impressive looking r that quietly disappears once another six months of history is added. The most important nuance is that correlation is not causation.
Ice cream sales and cases of sunburn correlate strongly because both follow the weather, not because one drives the other. In business the hidden third factor is usually the economic cycle, a seasonal pattern or a price change that moved both series at the same time.
Correlation also only detects straight-line relationships, which is a real limitation. A cost that falls as volume grows up to a point and then rises again because of overtime and expedited freight can show an r close to zero even though the underlying connection is strong and completely predictable.
In practice
Real-world examples.
Example
A subscription software company checks the correlation between monthly support ticket volume and cancellations over three years and finds r of 0.71. That is strong enough for the leadership team to fund a second support pod, on the reasoning that slow responses and churn are moving together for a shared reason.
Example
A regional bakery chain tests the correlation between daily temperature and hot drink sales and gets -0.83. Because the link is strong and negative, the operations manager builds a simple staffing rule that adds an extra person to the morning shift whenever the forecast drops below 8 degrees.
Example
A property fund discovers that its office and retail portfolios have a correlation of 0.92 in rental income. The investment committee concludes it holds two names but effectively one bet, and redirects the next acquisition budget towards industrial assets to reduce the concentration.
Formula
Calculation
The Pearson correlation coefficient is: r = the sum of (A minus mean A) times (B minus mean B), divided by the square root of [the sum of (A minus mean A) squared, times the sum of (B minus mean B) squared].
Take five months of advertising spend (A, in thousands of dollars) and revenue (B, in thousands of dollars). Spend runs 10, 20, 30, 40 and 50; revenue runs 120, 140, 180, 200 and 260.
Mean spend = (10 + 20 + 30 + 40 + 50) / 5 = 30. Mean revenue = (120 + 140 + 180 + 200 + 260) / 5 = 180.
Spend deviations: -20, -10, 0, 10, 20. Revenue deviations: -60, -40, 0, 20, 80.
Products of the paired deviations: 1,200 + 400 + 0 + 200 + 1,600 = 3,400.
Squared spend deviations: 400 + 100 + 0 + 100 + 400 = 1,000.
Squared revenue deviations: 3,600 + 1,600 + 0 + 400 + 6,400 = 12,000.
Denominator = the square root of (1,000 x 12,000) = the square root of 12,000,000 = 3,464.10.
r = 3,400 / 3,464.10 = 0.98.
An r of 0.98 across only five months is a very strong positive association, but with so few observations it is a signal to test further, not a licence to promise that every extra $10,000 of spend produces the same lift.Case study
Seen in the real world.
In this illustrative example, Marrowfield Retail Group, a fictional homewares chain with 40 stores, believed its online and in-store channels were independent businesses. Budgeting was done separately, and the board treated the two revenue lines as a natural hedge against a weak year.
A new analyst pulled 36 months of channel revenue and calculated a correlation of 0.88. Far from offsetting each other, the two lines rose and fell almost in lockstep, because both were driven by the same national advertising calendar and the same consumer confidence cycle. The apparent hedge did not exist.
The finance director used the finding to change two things. Scenario planning was rebuilt so that a downturn was modelled as hitting both channels together rather than one at a time, which raised the cash buffer the board agreed to hold. The company also began testing a trade supply channel whose demand ran on construction cycles rather than consumer sentiment, deliberately looking for a revenue stream with a low correlation to the existing two.
Watch out
Common mistakes.
- Treating a high correlation as proof of cause and effect, then spending money on the assumption that pushing one number will move the other.
- Calculating correlation from a small sample, often six or eight data points, and presenting the result as a settled fact rather than an early hint.
- Assuming a low r means no relationship, when the two series may be strongly connected in a curved or stepped pattern that the statistic cannot see.
Questions
People also ask.
What counts as a strong correlation in business?
There is no fixed cut-off, but readings above about 0.7 or below about -0.7 are generally treated as strong, 0.3 to 0.7 as moderate and below 0.3 as weak.
Does correlation change if I use different units?
No, the coefficient is unit-free, so measuring revenue in dollars or thousands of dollars gives exactly the same answer.
How is correlation different from covariance?
Covariance shows the direction of the relationship but its size depends on the units used, while correlation rescales that same idea onto a fixed -1 to +1 range so different pairs can be compared.
From the founder's library
Take it further with the book.
Build your financial confidence beyond this definition. Shihan's full-length guide, Accounting Fundamentals, takes the same plain-English approach and turns it into a complete, practical playbook for non-finance managers, business owners and students - with chapter-end quiz answers and presentation slides included.
25% off with code MMHQ25, applied at checkout. Priced in USD - checkout may show the equivalent in your local currency.
View the book and save 25%