How to Run a Profitability Analysis

Profitability analysis answers a question that top-line reports cannot: where is money actually being made, and where is it being subsidised by the profitable parts of the business. In most mature businesses, the decomposition reveals a concentrated profit pool and a long tail of break-even or loss-making activity.

The framework here follows the standard cost-accounting treatment^[1], with the ABC (activity-based costing) refinements that have become standard since Kaplan and Cooper^[2].

1. Layer profitability — gross, contribution, operating

Three layers, each answers a different question:

• Gross profit = Revenue − COGS. Answers: is the product economics working at unit level? Gross margin varies by category: Damodaran's January 2026 averages for listed US companies are 71.7% for system and application software, 33.2% for general retail, 35.3% for specialty retail and 33.4% for business and consumer services^[4].
• Contribution profit = Gross profit − Variable selling and marketing costs. Answers: after acquiring and serving the customer, what is left to cover fixed overhead?
• Operating profit = Contribution profit − Fixed overhead allocation. Answers: at the segment/product level, is this activity covering its share of overhead?

Most business reports stop at gross. That is where the money-losing long tail hides — products with 60% gross margin can be operating-loss if they require high-touch onboarding or heavy support.

2. By product or SKU

Compute gross margin per SKU. Rank them, then plot cumulative profit against the share of SKUs. Once cost-to-serve is included, the curve often rises above 100% of total profit before falling back, the "whale curve" Kaplan and Narayanan describe for customers^[5].

For each SKU, also compute a variable-cost-to-serve: support tickets per customer × support cost per ticket, average onboarding time × loaded hourly cost, returns and refund rate × average transaction. Some products with high gross margin are actually contribution-negative after cost-to-serve. These are the quiet losses in most catalogs.

Action triggers from SKU-level analysis:

• SKUs that are contribution-negative after cost-to-serve: reprice, fix the cost to serve, or discontinue unless they are loss leaders with proven attach rates.
• SKUs with low volume and low margin but high complexity cost (SKU count burdens operations): rationalise at next catalog review.
• Top 5 SKUs: protect. These are where operational improvement compounds fastest.

3. By customer segment

Segment by something operationally meaningful: size band (SMB / Mid-market / Enterprise), industry vertical, geography, acquisition cohort. For each segment, compute:

• Average revenue per customer (recognised revenue on the same ASC 606 basis as your P&L, not bookings or cash collected^[3]).
• Gross margin per customer (carries category COGS).
• Fully loaded cost-to-serve (support tickets, CS time, custom work).
• Net contribution per customer.

A common pattern in B2B: SMB customers carry more support per dollar of revenue, so their margin after support costs falls furthest below the company average, while enterprise customers may carry discounted prices but less support per dollar. Mid-market is often the sweet spot — high ACV with moderate support burden. Reveal this in your own data before you shift go-to-market strategy.

A caveat on cost allocation: using revenue as the allocation base for shared overhead (CS, infrastructure, management time) systematically over-allocates to enterprise customers because they have higher revenue but often similar or lower incremental cost. Use activity drivers — ticket count, provisioned seats, time-use, where possible^[2].

4. By acquisition channel

Channels differ dramatically in customer quality, not just cost. Compute CAC, retention, LTV, and resulting LTV:CAC by channel:

• Organic (SEO, direct, content): Typically lowest CAC, highest retention, strongest LTV. Usually the best channel — except that it's capacity-constrained.
• Paid search: Medium CAC, intent-qualified leads, often strong retention. Scales but margin compresses at volume.
• Paid social: Variable CAC, lower intent, often weaker retention unless targeting is tight.
• Partnerships/referrals: Highest conversion rates, best retention, but dependent on partner capacity.
• Outbound sales: Highest loaded CAC, enterprise ACVs, long sales cycle. Economics work only above an ACV high enough to pay for a rep's time; compute that threshold from your own cost per closed deal.

The profitability picture changes by channel. A channel at 2.5:1 LTV:CAC that pays back in 6 months can be healthier than one at 4:1 that takes 30 months, because the second ties up cash for longer and its LTV rests on retention assumptions further into the future.

5. Turn analysis into action

Profitability analysis only matters if it changes decisions. Three typical outputs:

• Pricing changes. Segments with below-target margin often need a price increase, a feature pull-back, or a restructured offer.
• Product rationalisation. Loss-making SKUs and segments get cut. The secondary benefit, reduced operational complexity, can rival the direct margin improvement.
• Investment reallocation. Channels and segments generating the top-decile contribution margins get disproportionate sales, marketing, and R&D investment. Underperforming areas get frozen or cut.

Run the full analysis annually, update the headline numbers quarterly. The concentration of profit in a small share of activity tends to persist; the specific products, segments, and channels that occupy that share shift with market conditions. Knowing which is which, accurately, is what the work produces.

6. The allocation question honestly

Profitability analysis depends on how you allocate shared costs. The two most common approaches — revenue-share allocation and activity-based costing — produce materially different answers.

Revenue-share allocation: if segment A is 60% of revenue, it gets 60% of shared overhead. Simple, but systematically wrong when segments have different cost-to-serve profiles. A segment with high ARPU but low support burden is over-charged; a segment with low ARPU and high support burden is under-charged.

Activity-based costing (ABC): allocate shared costs by their actual driver^[2]. Support time allocates by ticket volume; infrastructure allocates by actual usage; sales team allocates by hours spent per segment. More accurate, more work to maintain.

The practical recommendation for small business: start with revenue-share for the first pass, identify the 2–3 shared-cost pools where ABC would materially change the answer, and apply ABC to just those. Full ABC across every shared cost pool is rarely worth the operational overhead in a small business.

7. Caveats and honest limits

Three ways profitability analysis misleads decision-makers when applied naively:

• Short-term versus long-term profitability. A segment losing money today might be the highest-LTV segment with long ramp. Pure current-quarter profitability kills new-market investments before they compound.
• Covering-overhead fallacy. "This segment is profitable once you exclude shared overhead" is usually a warning sign — shared overhead exists because the business needs it, and every segment needs to contribute to it.
• Strategic value beyond margin. Anchor customers, reference accounts, and segments that unlock distribution into other segments can be worth carrying at thin margin. The analysis should flag these but not decide for them.

The work produces numbers; judgment applies them. The analysis that routinely guides good decisions is the one that surfaces the hidden structure of where profit actually comes from, then gets debated by people who understand the strategic context beyond just the margin column.

8. Numeric worked example — segment decomposition

A $4M-ARR B2B SaaS runs a headline 79% gross margin (above the 71.7% January 2026 average for listed US software companies^[4]) and an 8% operating margin. Decompose the $4M across three segments, with support time costed at $100 an hour:

Segment      ARR    GM%  Gross profit  Support hrs/mo  Support  Var. S&M  Net contrib  Contrib/$ARR
──────────────────────────────────────────────────────────────────────────────────────────────────
Enterprise  $1.6M   82%   $1,312k        120          $144k     $88k      $1,080k      $0.68
Mid-market  $1.6M   78%   $1,248k         80           $96k    $102k      $1,050k      $0.66
SMB         $0.8M   74%     $592k        160          $192k    $120k        $280k      $0.35
Total       $4.0M   79%   $3,152k        360          $432k    $310k      $2,410k      $0.60

Aggregate gross margin looks uniform; the contribution picture is not. SMB uses 44% of support hours on 20% of ARR, more than twice the company's support hours per dollar, and earns about half the contribution per dollar of the other segments. The $2.09M of fixed overhead that takes contribution down to the 8% operating margin decides the rest: allocated by revenue share, SMB carries $418k and shows a $138k operating loss. Allocating the support-related part of that overhead by support hours instead would move still more of it onto SMB and off enterprise and mid-market.

Decision: raise the SMB floor price by 20% and accept that 10% of SMB accounts leave over it. The extra churn costs about $80k of ARR ($59k of gross profit); the increase adds $144k of ARR ($107k of gross profit) on the accounts that stay. Net: about +$64k of ARR and +$47k of gross profit, before the support hours freed by the accounts that leave.

9. Failure modes worth naming

• ABC over-engineering. A ticket-time-based allocation model needs regular maintenance to stay accurate, and a 12-driver ABC model is more upkeep than a small team will sustain. Pick 2–3 drivers that actually move the answer and accept approximation on the rest^[2].
• Conflating profitability with strategic value. A flagship enterprise logo may be contribution-negative on support cost but unlock three similar logos through reference calls. Quantify the strategic dependency (how many deals cite it?) before cutting a loss-making anchor account.
• Stale fixed-overhead allocation. Overhead that was 18% of revenue two years ago may be 25% today after an ops hire wave. Rerun the full analysis when headcount or infrastructure crosses a step function; quarterly deltas are not enough.

Frequently asked questions

How do you run a profitability analysis?

Decompose profit three ways — by product or SKU, by customer segment, and by acquisition channel — using three layers per cut: gross profit (revenue minus COGS, the unit-economics question), contribution profit (gross minus variable selling and marketing, what's left to cover fixed overhead), and operating profit (contribution minus allocated fixed overhead, whether the activity covers its share). Most reports stop at gross, which is exactly where the money-losing long tail hides: a product at 60 percent gross margin can be operating-loss once high-touch onboarding or heavy support is counted.

Why does aggregate profit hide the real picture?

Because profit is concentrated, often more than the 80/20 rule suggests. Kaplan and Narayanan report that the most profitable 20 percent of customers typically generate 150 to 300 percent of total profit, while the least profitable customers give much of it back. SKU-level analysis can find high-gross-margin products that are contribution-negative after cost-to-serve (support tickets, onboarding time, returns). The decomposition surfaces this hidden structure so you can fix or discontinue contribution-negative SKUs, rationalize low-volume high-complexity SKUs, and protect the top 5 where operational improvement compounds fastest.

How should I allocate shared overhead?

Allocate by activity driver, not by revenue. Revenue-share allocation (a segment that is 60 percent of revenue gets 60 percent of overhead) systematically over-charges high-ARPU, low-touch segments and under-charges low-ARPU, high-support ones. Activity-based costing allocates support time by ticket volume, infrastructure by usage, and sales time by hours per segment — more accurate but more work. The practical recommendation for a small business: start with revenue-share, identify the 2 to 3 shared-cost pools where ABC would materially change the answer, and apply ABC to just those.

When can a loss-making segment be worth keeping?

When it has strategic value beyond margin or long-ramp LTV. Three honest caveats apply: a segment losing money today may be the highest-LTV segment with a long ramp, so pure current-quarter profitability can kill compounding investments; the 'profitable once you exclude shared overhead' framing is a warning sign, because shared overhead exists because the business needs it; and anchor customers, reference accounts, and segments that unlock distribution can be worth carrying at thin margin. The analysis should flag these, but judgment — informed by how many deals actually cite the anchor account — decides them, not the margin column alone.