Wells Fargo’s EPS Grew Faster Than Earnings. I’d Like the Rest of the Story.

An earnings-per-share headline can make a quarter look as if the entire company went to the gym, ate its vegetables and came back 13% stronger.

Sometimes the company did improve. Sometimes part of the improvement is in the denominator. Sometimes taxes helped. Often several things happened at once, which is inconvenient for a headline and extremely relevant to a finance professional.

Wells Fargo’s latest quarter gives me a useful reason to pull the result apart. I can appreciate higher earnings without allowing one growth percentage to become the company’s entire personality.

What I want from FP&A is an explanation that survives the removal of the prettiest number. If the presentation gets less confident when we ask how earnings turned into EPS, the problem isn’t my question.

Wells Fargo Q3 2026 EPS revenue: the actual comparison

Wells Fargo reported diluted EPS of $1.87 versus $1.66 a year earlier, revenue of $21.870 billion versus $21.436 billion, and net income of $5.951 billion versus $5.589 billion. The quarter’s credit-loss provision was $1.029 billion, up from $681 million. Noninterest expense fell from $13.846 billion to $13.743 billion.

The release also reports $170 million of discrete tax benefits and $2.5 billion of common-share repurchases during the quarter. The October 13 SEC-filed release is the source.

Using those reported amounts, revenue grew about 2.0%, net income about 6.5% and EPS about 12.7%. Those are different measurements of the same quarter.

I am comparing reported results with the prior year. I am not claiming a beat or miss against an analyst estimate I haven’t independently established. The market’s surprise and the business’s economics deserve separate seats at the table.

I don’t want a scoreboard pretending to be an explanation

Imagine a management presentation with a green EPS box, a mildly green revenue box and a sentence saying that performance was strong.

I’d ask what management intends us to do with that information. Approve investment? Increase the forecast? Change compensation? Conclude that the business can return more capital?

Each decision requires different evidence. A scorecard can tell us where we landed. It cannot tell us what to assume next.

I’d put operating performance, risk costs and capital allocation in separate views before connecting them. Otherwise the same positive percentage gets used to justify every proposal, like a corporate universal remote that somehow also approves headcount.

This distinction travels well beyond banking. A manufacturer can improve per-share results through share repurchases while its factories produce similar earnings. A software company can deliver profit growth through a tax movement while customer retention weakens.

I want Finance to explain the mechanism. Color coding is not a mechanism.

The denominator deserves its own slide

Here’s a deliberately simplified hypothetical company with $1 billion of earnings available to common shareholders and 100 million average diluted shares. EPS is $10.

Next year, earnings increase 6% to $1.06 billion and average diluted shares fall 6% to 94 million. EPS becomes approximately $11.28, an increase of 12.8%.

Build the bridge sequentially. First hold shares at 100 million: the earnings improvement raises EPS by $0.60 to $10.60. Then use the lower share count: EPS rises another $0.68 to approximately $11.28.

That attribution depends on the order of the bridge; the interaction between earnings and shares can be allocated differently. The total change does not.

This is an illustrative calculation, not an exact decomposition of Wells Fargo’s EPS. Reported net income is not automatically the EPS numerator: preferred dividends and other relevant adjustments may matter. Average diluted shares are not the same as quarter-end shares.

I would reconcile the actual numerator and weighted-average denominator before claiming a precise company attribution. Dividing two headline figures and calling the result “shares” is how a useful question acquires a false answer.

Buybacks can be sensible and still complicate the story

I don’t consider share repurchases inherently suspicious. A company with excess capital and limited attractive reinvestment opportunities may reasonably return capital to shareholders.

But I won’t call denominator-driven EPS growth an operating improvement. The transaction can be economically sound while belonging in a different part of the discussion.

I’d compare repurchases with available uses of capital, assess the purchase price and examine the capital left behind. A transaction that improves EPS mathematically does not automatically create value economically.

For a hypothetical company, repurchasing shares at an inflated price can reduce the share count while transferring too much value to selling shareholders. Keeping the money also has an opportunity cost. Neither option deserves an automatic moral victory.

The important question is whether management made a sensible capital decision, given available information and alternatives. “Accretive to EPS” answers a narrower question.

I’ve never wanted a finance team to become hostile to a tool because it can improve a metric. I want it to understand which job the tool actually performed.

The operating bridge should be allowed to look untidy

Using the published revenue, expense and provision figures, the year-over-year movements are approximately $434 million favorable from revenue, $103 million favorable from lower expense and $348 million unfavorable from the higher provision. Together, those lines yield $189 million before taxes and remaining reconciliation items.

I would not label that the complete net-income bridge. The actual result requires the full income statement, including taxes and any attribution differences.

A partial bridge is fine if clearly identified. An incomplete bridge disguised as a complete one is not fine because the final bar happens to be green.

Internally, I’d insist that the residual be explained. Material unexplained amounts need owners, not the magical category “other,” where variances go to avoid being interviewed.

I also want the order of the bridge to match the question. Start with revenue and operating cost to discuss execution. Add credit costs to show the risk result. Then reconcile taxes, attribution and shares. That sequence prevents a capital action from wearing an operating medal.

Credit provision is a forecast with an accounting consequence

A provision reflects expected losses and relevant allowance changes; it isn’t simply the cash default bill received that quarter. Treating it as one creates sloppy forecasting.

I’d separate realized charge-offs, recoveries, portfolio growth and changes in expected-loss assumptions. A growing loan portfolio can need more allowance even without a sudden collapse in borrower quality.

That doesn’t mean every provision increase is harmless. It means I need to understand why it increased. Changes in mix, borrower health or forward-looking conditions require different responses.

For FP&A, the practical issue is that revenue and credit costs must be modeled together. More lending can improve interest income while consuming capital and increasing future losses. A growth plan that books the revenue immediately and leaves the risk calculation for another department is incomplete.

Risk isn’t a guest we invite after the budget has been approved. It already owns a seat.

I’d therefore reconcile the planning view with the credit function’s assumptions. Two departments forecasting different borrower worlds won’t become consistent because Finance added their totals.

A loan growth example that earns its own risk

Consider a hypothetical $500 million average incremental loan portfolio. Assume an annual gross yield of 7%, funding costs of 4%, servicing and operating costs equal to 0.5% of balances, and expected credit cost of 0.8%.

The arithmetic gives $35 million of interest income, $20 million of funding expense, $2.5 million of operating cost and $4 million of credit cost. Pretax contribution is $8.5 million.

If expected credit cost rises to 1.6%, the credit line becomes $8 million and contribution falls to $4.5 million. If funding costs also rise to 4.5%, contribution falls another $2.5 million to $2 million.

The lending team can deliver exactly the planned volume while the economics disappoint badly. The volume target wasn’t wrong; it was insufficient.

These are simplified annual assumptions, not Wells Fargo pricing or losses. They omit taxes, capital charges, fee income and portfolio detail. I’d add those before making a real decision.

Even this stripped-down example shows why a finance model needs more than balances multiplied by yesterday’s spread.

The hurdle rate changes the growth conversation

Add an illustrative capital requirement equal to 10% of the $500 million portfolio, or $50 million. Assume a 12% pretax capital hurdle for this example: required contribution is $6 million.

The base case’s $8.5 million contribution exceeds that hurdle by $2.5 million. The higher-credit-cost case’s $4.5 million falls short by $1.5 million. The combined credit-and-funding stress falls short by $4 million.

That isn’t a claim about regulatory capital requirements or Wells Fargo’s hurdle. It is a decision model showing how capital can change the interpretation of apparently profitable growth.

What would I do? Consider better pricing, a different borrower mix, tighter structures, or less expansion. I’d also investigate whether relationship fees improve the full customer economics without double counting them elsewhere.

The answer might still be to grow. A strategic relationship can have broader value. I’d require that value to be evidenced, though. “Strategic” cannot be the word that makes a below-hurdle result stop being discussed.

Cost savings need a dependency map

A lower expense total is encouraging. My next question is what produced it and what must happen for the benefit to continue.

If savings come from removed duplication, they may be durable. If a technology bill moved to next quarter, the “saving” is a calendar event wearing a productivity badge.

I’d distinguish gross efficiencies from reinvestment. Suppose a hypothetical business removes $50 million of annual operating costs but adds $35 million of recurring technology and specialist expenses. Net recurring improvement is $15 million, before transition costs.

If implementation costs $20 million initially, the first-year result depends on timing. Full-year run-rate savings cannot be booked in January when the work finishes in September.

I’d ask whether the removed capacity was genuinely unnecessary or whether remaining staff now perform the same work through overtime and external support. A headcount reduction can create savings. It can also create a new consulting budget with better stationery.

The model needs the operational dependency, not merely the expense reduction.

Tax benefits belong in the result, with the right label

Discrete tax benefits can improve reported earnings. They are part of the actual quarter and shouldn’t be erased because the operating story becomes less dramatic without them.

For forward planning, I would separate recurring tax assumptions from discrete events. A prior-period resolution is not automatically next year’s quarterly run rate.

In a hypothetical company with $100 million pretax earnings and a 25% recurring tax rate, net income would be $75 million. A $10 million discrete tax benefit raises reported net income to $85 million without changing pretax performance.

If someone annualizes the $85 million quarter into $340 million, versus $300 million at the recurring rate, we have manufactured $40 million of forecast earnings by extending a one-time benefit.

Actual tax modeling is more complicated: jurisdiction mix, deductions, timing and other items can matter. The simple example is enough to expose the mistake.

I’d let the quarter keep its benefit and make the forecast earn its own tax assumption. Favorable accounting outcomes do not need to be recurring to be welcome.

My management choices would follow the bridge

Suppose our own business faces modest revenue growth, successful expense work and rising risk costs. I’d compare actions on their net economics.

One choice is to pursue volume aggressively. It may help revenue, but I would test marginal contribution after funding and risk. Another is to preserve pricing discipline and accept slower growth. A third is to shift capacity toward business with stronger risk-adjusted returns.

Across all alternatives, I’d include execution costs and time. Moving resources isn’t free. A business line with a better margin may lack the demand to absorb the capacity.

I would avoid declaring a winner from consolidated revenue growth alone. The best use of resources could produce less revenue and more economic profit.

That is an uncomfortable discussion when bonuses reward volume. It’s still the discussion we need. Otherwise FP&A ends up explaining why everyone met their targets and the company missed its objective. Nobody enjoys that meeting except perhaps the manufacturer of aspirin.

Consensus doesn’t govern my operating budget

Market expectations have a legitimate role in investor communication. Management should understand what external analysts anticipate and explain differences responsibly.

Internally, I want the baseline to match the task. Compared with our plan, which assumptions were wrong? Compared with last year, what changed? Compared with consensus, why was the market surprised?

A miss against consensus could coexist with strong year-over-year progress. A beat could coexist with weak execution against management’s own plan. Neither statement is contradictory.

I would label every comparison rather than letting the word “beat” float around until someone attaches it to whatever decision they prefer.

For a practical reporting design, I’d show actuals against internal forecast and prior year, then include a separate external-expectations note where verified data exist. That prevents investor surprise from overwriting operational accountability.

The stock market can react to a number faster than we can finish a reconciliation. That does not relieve us of finishing it.

The forecast should preserve the choices

I’d build scenarios around specific economic drivers rather than applying one arbitrary percentage to the total profit line.

For the hypothetical loan portfolio, a credit-cost scenario changes expected loss; a funding scenario changes deposit or wholesale funding cost; a demand scenario changes balances and related expense. A combined case shows interaction.

I’d then add management responses separately. Repricing might improve yield but reduce volume. Tighter underwriting might lower expected losses but change customer mix. Cost cuts might improve current expense while increasing service or control risk.

That separation tells us what happens without intervention and what the proposed intervention changes. It also prevents a downside case from magically recovering because the model quietly assumes management will fix everything.

My approach to the drivers behind margins and unit economics starts with the same discipline: identify the actual mechanism before choosing the response.

A scenario that cannot lead to a decision is mostly a weather report with company colors.

The board pack needs fewer adjectives

I’d bring a short, reconciled explanation of performance and a clear recommendation. The performance explanation would distinguish operating contribution, credit effects, tax movements and the share denominator.

The forward view would show which drivers can recur, the capital implications and the actions available if conditions change. Material uncertainty would appear in the assumptions, not in tiny print beneath a triumphant chart.

I would also show how the proposed action affects the measure management is trying to improve. A buyback may improve EPS. An operating investment may improve future earnings while reducing current profit. A risk decision may lower revenue growth and improve return quality.

Those choices deserve deliberate judgment. I don’t want directors choosing between them using a single green box.

Nor do I want an encyclopedia. Finance should do the detailed reconciliation before the meeting and bring the parts that determine the decision. Making a board read forty pages isn’t evidence that we understand the quarter.

My conclusion: make the metric confess

I like higher earnings. I like cost discipline. I am willing to support capital returns and profitable growth when the economics justify them.

What I refuse to support is a result explained entirely through its most flattering percentage.

For Wells Fargo, the reported EPS improvement is worth understanding precisely because revenue, earnings, risk costs and capital actions tell different parts of the story. For our own companies, the same habit should shape forecasting and resource decisions.

My opinion is straightforward: an EPS headline is an invitation to reconcile, not permission to stop. Finance earns credibility by explaining where improvement came from and which part can support the next decision.

If we can’t tell whether the business improved, the tax line helped or the denominator got smaller, we haven’t finished the analysis. The earnings release may be ready for its photograph. The forecast is still getting dressed.

Sources and model notes

Reported company amounts are from Wells Fargo’s SEC-filed third-quarter 2026 earnings release. Growth percentages are calculations using the stated figures. All loan portfolios, tax examples, hurdle rates and management scenarios are hypothetical. They do not estimate Wells Fargo’s undisclosed internal plans or exact EPS attribution.