DuPont ROE Decomposition Inside a KPI Tree

ROE Fell 3.05 Points While the Equity Multiplier Rose

DuPont ROE decomposition in a KPI tree splits return on equity into the ratios that produce it, so a movement in ROE traces back to margin, asset efficiency, or the mix of debt and equity funding the balance sheet.

The framework is old and still standard. DuPont built it in the 1920s as an internal performance measure, and it sits in the CFA Level 1 curriculum as the canonical decomposition of return on equity, read on August 14, 2026.¹

The awkward part is structural. ROE is a ratio, its three factors multiply, and a KPI tree adds. A tree whose parent nodes are sums of their children cannot hold a multiplicative chain without either storing ratios as data or accepting a residual.

Both shortcuts are avoidable. The rest of this article uses one company whose equity multiplier rose from 2.50 to 2.94 while ROE fell 3.05 points, and shows the arrangement that reconciles.

What Is DuPont ROE Decomposition in a KPI Tree?

DuPont ROE decomposition splits return on equity into net profit margin, asset turnover and the equity multiplier, which multiply back to ROE. Inside a KPI tree such as kpitree.io, those three ratios are not stored anywhere. Each is derived by dividing two summable columns drawn from an additive layer of line items underneath.

The identity works because the middle terms cancel. Net income over revenue, times revenue over assets, times assets over equity, reduces to net income over equity.

That cancellation is the whole trick, and it is also the warning. Any chain of ratios that cancels can be written; it does not follow that each link describes a separate cause. Margin and turnover move together in most businesses.

A driver tree decomposes one metric into the levers that produce it, and DuPont is the oldest published example of the pattern.³ The difference between DuPont and a general driver tree is that DuPont stops at three financial ratios. It says nothing about why margin moved.

That is why DuPont belongs at the top of a tree, not at the bottom. It frames the question. The additive layer underneath answers it.

The Three-Step and Five-Step Identities

The three-step identity is the one most people mean by DuPont. Return on equity equals net profit margin, times asset turnover, times the equity multiplier.¹

The five-step version splits the margin term into three. Return on equity equals tax burden, times interest burden, times operating margin, times asset turnover, times the equity multiplier, where tax burden is net income over pre-tax income and interest burden is pre-tax income over EBIT, read on August 14, 2026.²

The five-step version earns its extra complexity in one specific case: when a company funds itself with more debt, the equity multiplier rises and the interest burden falls at the same time. The three-step version hides that offset inside a single margin number. The five-step version shows both halves.

Use three factors for a board summary. Use five when the financing question is live, which is exactly when the headline ROE is least trustworthy.

Build the Additive Layer First, Then Derive Every Ratio

The correct construction order is bottom up, and most teams do it the other way.

Layer one holds summable stocks and flows only. Revenue, cost of goods sold, operating expenses, EBIT, interest expense, tax expense, net income. Total assets as a sum of asset classes. Shareholders' equity as a sum of paid-in capital, retained earnings and reserves. Every node in this layer is a parent that equals the sum or difference of its children.

Layer two holds nothing but division. Net margin is net income divided by revenue. Asset turnover is revenue divided by total assets. The equity multiplier is total assets divided by equity. ROE is net income divided by equity, computed directly rather than reconstructed from the product of the three.

This order matters because averages and ratios do not add. A stored margin column aggregated across segments returns the average of averages, which is not the margin. Storing the numerator and denominator instead is the same discipline that governs the three decomposition patterns in the KPI tree template.

A Worked Example That Reconciles

One company, two fiscal years, no rounding until the last step.

FY2025. Revenue 400.0 million dollars. EBIT 40.0. Interest expense 6.0. Pre-tax income 34.0. Tax 8.5. Net income 25.5. Total assets 320.0. Equity 128.0.

Derived: margin 6.375 percent, turnover 1.25, equity multiplier 2.50. Their product is 19.92 percent, which equals 25.5 divided by 128.0.

FY2026. Revenue 440.0. EBIT 39.6. Interest expense 9.0. Pre-tax income 30.6. Tax 7.65. Net income 22.95. Total assets 400.0. Equity 136.0.

Derived: margin 5.216 percent, turnover 1.10, equity multiplier 2.94. Their product is 16.88 percent, which equals 22.95 divided by 136.0.

ROE fell 3.05 points. Revenue grew 10 percent. The equity multiplier rose. Read the ratio layer alone and the story is contradictory.

Read the additive layer and it is not. Net income fell 2.55, and that splits exactly: EBIT down 0.4, interest expense up 3.0, tax down 0.85. Minus 0.4, minus 3.0, plus 0.85 equals minus 2.55. The company borrowed to buy 80.0 million dollars of assets and paid 3.0 million more in interest for them.

The additive split reconciles to the dollar with no convention and no residual. The ratio split does not. When the two disagree about what caused a move, the additive layer is the one that is arithmetically true.

Why Do the Three DuPont Effects Not Sum to the Change in ROE?

Because the three factors multiply. Changing one at a time while holding the others at base values ignores the cross terms created when two or more move together. In the worked example the three single-factor effects sum to minus 2.50 points against an actual decline of 3.05 points, leaving 0.55 points, about 18 percent of the move, unallocated.

Compute the three effects the naive way and watch the gap open.

Margin effect: the new margin of 5.216 percent, times base turnover 1.25, times base multiplier 2.50, less base ROE, is minus 3.62 points. Turnover effect: base margin, times the turnover fall of 0.15, times base multiplier, is minus 2.39 points. Equity multiplier effect: base margin, times base turnover, times the multiplier rise of 0.44, is plus 3.52 points.

Sum: minus 2.50 points. Actual: minus 3.05 points. The 0.55 point gap is not rounding and it is not a data error. It is the interaction between three factors that all moved at once.

There is an exact statement available, and it needs no convention. For any ratio, the new value divided by the old equals numerator growth divided by denominator growth. Net income fell 10.00 percent, equity rose 6.25 percent, and 0.90 divided by 1.0625 is 0.8471, a decline of 15.29 percent. Apply that to 19.92 percent and you get 16.88 percent exactly.

That is the arrangement an additive tree supports natively, because both the numerator and the denominator are sums.

Three Ways to Arrange ROE in a Tree

There is no single correct arrangement, and the choice is driven by what the reader needs to do next.

A board wants a level explained. An analyst wants a change diagnosed. A bridge chart wants terms that add to the total even at the cost of order dependence. These are different jobs and they justify different structures.

What is not negotiable is documenting which one you chose and holding it for the year. A DuPont bridge computed in a different order each quarter produces a series nobody can compare. The table states what each arrangement produces and when to reach for it.

How Do You Test a DuPont Tree for Correctness?

Run four checks. The three factors must multiply back to ROE computed directly from net income and equity. No ratio may be stored as a column. Equity must use one convention, ending or average, in both periods. And any effect split must state whether it reconciles exactly or carries a residual.

Test 1. Product reconciliation. Multiply margin, turnover and the equity multiplier and compare to net income divided by equity. A gap means one of the three uses a different denominator than you think, most often assets net of something.

Test 2. No stored ratios. Margin, turnover, multiplier and ROE are all derived by division at whatever node you are reading. A stored margin column that is summed or averaged across segments produces a number that is not the margin of the whole.

Test 3. One equity convention. Ending equity is simpler and matches the balance sheet. Average equity matches the income statement period better. Both are defensible. Mixing them across two periods is not, and a mid-year change silently breaks the whole series.

Test 4. Declared residual. If a change decomposition does not sum to the actual change in ROE, print the residual as its own line. A residual quietly folded into one of the three terms is the most common way a DuPont bridge misleads.

Failing any of the four returns a number that survives a slide and not a question.

Where DuPont Is the Wrong Tool

DuPont explains the shape of a return. It is a poor instrument in four situations.

Banks, insurers and asset managers. Umbrex notes that the standard decomposition does not work cleanly for financial firms, because assets and funding structure mean something different there, read on August 14, 2026.³ Analysts use sector-specific variants instead.

Negative or near-zero equity. A company with negative equity produces a negative ROE that improves as losses deepen. The ratio is arithmetically defined and economically meaningless. Return on invested capital is the usable substitute.

Cross-industry comparison. A utility and a discount retailer have structurally different turnover and margin. Comparing their DuPont factors side by side measures the industry, not the management.

Causal attribution. The tree says margin cost 3.62 points. It does not say whether that came from input costs, discounting, or product blend. That question is answered by a price, volume and mix bridge on the revenue line, or by repeatable root cause analysis on the cost line.

Naming the limits is part of shipping the method.

Building the ROE Tree in kpitree.io

kpitree.io is a self-service KPI tree builder for finance, business and product analysts. This is the only section of the article about the product.

The mechanics match the four tests. You upload a CSV of period rows carrying the income statement and balance sheet line items: revenue, cost of goods sold, operating expenses, interest expense, tax expense, asset classes, equity components. Inside the tree, derived KPIs are computed by dividing two summable columns, so margin, turnover, the equity multiplier and ROE are all division results at every node rather than stored values. That is what Test 2 requires.

Node identities are addition and subtraction. Net income as a chain of subtractions from revenue, and total assets as a sum of asset classes, are therefore native structures, and the additive reconciliation in the worked example becomes a property of the tree rather than a spreadsheet check.

Three honest limits. The evidenced ingest path today is CSV upload, so a DuPont tree is refreshed by uploading a new file rather than through a live connection. Capabilities the site marks as coming soon are not available. And a tree output is an arithmetic decomposition, not financial, investment or accounting advice.

The same additive discipline runs through the worked KPI tree examples.

Frequently Asked Questions

Is DuPont analysis a KPI tree?

It is the earliest published one, and a narrow one. DuPont decomposes a single ratio into three financial ratios. A general driver tree keeps going into operational levers such as price, units and headcount.³

Should I use ending equity or average equity?

Either, consistently. Ending equity matches the balance sheet date. Average equity, the mean of opening and closing, matches the income statement period better and is the more common choice in equity research. Fix one and record it in the metric definition.

Three-step or five-step?

Three for communication, five for diagnosis. The five-step version separates tax burden, interest burden and operating margin, which is the only way to see a financing change and its interest cost at the same time.²

Can a semantic layer compute ROE for me?

It can define it correctly. The dbt Semantic Layer specifies a ratio metric as an explicit numerator and denominator rather than a stored ratio, read on August 14, 2026.⁵ That is the same rule as Test 2. What a semantic layer does not give you is the tree that shows which line item moved.

How common is driver-based practice in finance teams?

Still uneven. The 2026 AFP FP&A Benchmarking Survey on integrated planning draws on 332 finance professionals across 54 countries and examines how far driver-based methods have actually spread, read on August 14, 2026.⁴

Closing: Rebuild Last Year's ROE Bridge Before the Next Board Pack

The next time an ROE chart is presented and someone asks what moved it, a three-factor chain will not settle the question. It will produce three numbers that do not add up to the answer.

One CSV is enough. Two fiscal periods, one row each, with revenue, EBIT, interest expense, tax expense, net income, total assets and equity. Nothing else.

Build the additive layer first and check that net income equals revenue less every cost line. Then derive margin, turnover and the multiplier by division and confirm their product equals net income over equity. If it does, you have a defensible answer in an hour. If it does not, you have found a definition problem worth more than the answer would have been.

Start with one entity and two years. Upload that single CSV and decompose the return your board already argues about.

kpitree.io is a self-service KPI tree builder for finance, business and product analysts.

Sources

• DuPont Analysis and ROE Decomposition, CFA Level 1
• DuPont Analysis: Formula and Ratio Calculator
• DuPont ROE Tree Explained
• 2026 AFP FP&A Benchmarking Survey Report: Integrated Planning
• Creating metrics