The Pricing Power Map
Which company characteristics predict margin resilience when revenue growth slows? An XGBoost + SHAP analysis of 238 company-year observations.
competitive-strategy
research
We isolated the component of operating margin change not explained by revenue dynamics and used SHAP to rank the entry characteristics most associated with margin defense. The lever ranking shifts dramatically by sector.
Executive Summary
- Trailing P/E, operating margin, and profit margin are the three strongest predictors of margin resilience across our panel. Companies trading at lower valuations with compressed margins are the most likely to defend margins independently of what revenue is doing. The magnitude varies sharply by sector.
- The model achieves 65% directional accuracy in walk-forward cross-validation, correctly predicting whether a company will show margin resilience or compression in two out of three cases. This is a meaningful improvement over the 56% achieved in the Margin Expansion Playbook, suggesting that pricing power is a more learnable signal than raw margin change.
- The lever ranking shifts by sector. In Industrials, current operating margin dominates. In Technology, lagged valuation multiples and revenue scale matter most. In Healthcare, trailing P/E and AI adoption are the leading indicators. A generic “pricing power” checklist misses these structural differences.
The question
Every investment thesis eventually makes a claim about pricing power. The company grew margins through a cycle, the narrative goes, because it has a differentiated product, loyal customers, or contractual escalators. The investor is expected to take this at face value.
The problem is that margin performance during periods of strong revenue growth proves very little. Almost any business can show margin expansion when the top line is accelerating: fixed costs get spread across a larger base, mix shifts toward higher-margin products, and operating leverage does the work. The real test of pricing power is what happens when you strip out the revenue tailwind. Can the company defend its margins on its own merits, or is it riding the cycle?
We set out to quantify this. Across a panel of 140 public companies, which observable characteristics at entry are most associated with margin defense after controlling for revenue dynamics? And does the answer change by sector?
Measuring pricing power: the intuition
The concept is simpler than it sounds. Consider two companies in the same year:
- Company A grew revenue by 15% and expanded operating margins by 300 bps. Impressive? Maybe. But 15% revenue growth generates natural operating leverage. Any business with fixed costs should see margin expansion at that growth rate.
- Company B grew revenue by just 2% and still expanded margins by 250 bps. That is a fundamentally different signal. Company B did something that revenue alone does not explain: it defended and improved margins through pricing discipline, cost structure, or competitive positioning.
Company B has pricing power. Company A might, but we cannot tell from the raw margin number alone.
To separate these two cases systematically, we need a way to ask: given a company’s revenue growth rate, how much margin change would we expect? And how much actually occurred? The gap between expected and actual is the pricing power signal.
This interactive chart is best viewed on a larger screen. On mobile devices, some annotations may overlap.
The scatter plot above is the key visual. Each dot is a company-year. The dashed line is the statistical relationship between revenue growth and margin change across the full panel. Companies above the line defended margins better than their revenue growth would predict. Companies below the line saw margins compress more than the revenue slowdown explains.
The scatter plot tells us who has pricing power and who does not. But it does not tell us why. To answer that, we turned to the companies’ own words.
Pricing power profiles
For each company, we extracted passages from its SEC 10-K filing that discuss competitive positioning, pricing strategy, cost structure, intellectual property, and margin dynamics. We then used Claude (Anthropic’s language model) to synthesize these passages into a pricing power assessment, conditioned on the company’s quantitative margin resilience score, revenue growth, and sector context.
The profiles below cover the five most resilient and five most vulnerable companies in our panel, using each company’s most recent fiscal year with available forward margin data. Where a company has multiple years of data, we show only the latest observation to ensure currency. The LLM does not generate the analysis from scratch. It reads the company’s own filing language and connects it to the observed margin outcome. All underlying passages are sourced from public SEC filings cached in our research pipeline.
Strong pricing power
BMY — Bristol-Myers Squibb
Resilience: +1570 bpsHealthcare · FY2024 · Revenue growth: +7.3% · Margin change: +1629 bps
Bristol-Myers Squibb demonstrated clear pricing power in fiscal 2024, with margins expanding 1629 basis points on only 7.3% revenue growth and a margin resilience score of +1570 bps, placing the company in the top quartile of our peer panel for pricing strength. The company's ability to sustain and grow margins derives from its explicit reliance on market exclusivity periods for innovative products—as stated in the 10-K, "the majority of an innovative product's commercial value is usually realized during the period in which the product has market exclusivity," supported by a portfolio including patent-protected franchise products such as Revlimid (approved across multiple hematological malignancies) and Abraxane (differentiated through proprietary Nab technology for oncology indications)—coupled with deliberate management of "Pricing, Price Constraints and Market Access" disclosed as a material business section, indicating active pricing strategy execution. For investors and acquirers, this margin expansion signals strong near-term pricing power anchored to patent cliffs and exclusivity windows; however, sustainability depends critically on the timing of generic erosion for key products and the company's ability to transition to newly launched or acquired assets with equivalent pricing power.
AVGO — Broadcom
Resilience: +1264 bpsTechnology · FY2024 · Revenue growth: +44.0% · Margin change: +1174 bps
Broadcom demonstrated exceptional pricing power in fiscal 2024, with a margin resilience score of +1264 basis points indicating that gross margins expanded 1174 basis points despite 44% revenue growth, a performance that places the company in the top quartile of our peer panel and signals pricing discipline well beyond what organic scale alone would deliver. The company's proprietary technology moats across multiple product categories—including first-mover advantages in FBAR filter technology commanding significant cellular handset market share, custom silicon ASIC platforms for AI and high-performance computing with embedded IP cores that customers cannot easily replicate, and high-speed Ethernet transceivers built on proprietary digital signal processing architecture optimized for network standards—enabled Broadcom to command premium pricing and sustain margins as demand surged, particularly in data center and networking markets where customers face no viable switching costs. This suggests Broadcom's margin expansion is structurally defensible rather than demand-driven windfall; however, investors should monitor whether this pricing power persists as AI-driven capex cycles normalize and whether custom silicon solutions face increased competitive pressure from vertically integrated customers or emerging ASIC designers, as the sustainability of these margins depends on continued technological differentiation rather than cyclical capacity scarcity.
NVDA — Nvidia
Resilience: +1251 bpsTechnology · FY2024 · Revenue growth: +125.9% · Margin change: +830 bps
Nvidia demonstrated exceptional pricing power in fiscal 2024, with margin resilience of +1251 basis points significantly outpacing 125.9% revenue growth, indicating the company expanded margins even as it scaled dramatically rather than facing typical operating leverage compression. This pricing strength stems from Nvidia's architecturally differentiated GPU technology and comprehensive software ecosystem (Passages 1-2), which create switching costs and performance advantages that insulate the company from pure commodity competition, while simultaneously enabling premium licensing arrangements for IP and custom development (Passage 3) that extract additional value beyond core GPU sales. For investors, this margin expansion demonstrates durable competitive moats that should sustain pricing power even as competitive intensity from well-capitalized rivals inevitably increases (Passages 4-5), though investors should monitor whether custom licensing revenue streams and enterprise software subscriptions (Omniverse, NeMo, DGX Cloud) can maintain current contribution margins as the AI market matures and customer alternatives proliferate.
ABBV — AbbVie
Resilience: +1101 bpsHealthcare · FY2024 · Revenue growth: +3.7% · Margin change: +1174 bps
AbbVie demonstrated strong pricing power in 2024, with margin expansion of 1174 basis points substantially outpacing 3.7% revenue growth, placing the company in the top quartile for margin resilience despite only modest sales momentum. The company's 10-K reveals that while it faces intense competitive pressure from biosimilars, generic substitutes, and newer competitors in immunology (JAK inhibitors, anti-TNF alternatives) and oncology (BTK inhibitors) plus emerging Botox competitors, AbbVie maintains material intellectual property protection through patent coverage and regulatory exclusivity that insulates its core portfolio from immediate pricing compression, with patent term restoration mechanisms available under the Hatch-Waxman Act to extend exclusivity windows. For investors, this margin resilience signals that AbbVie's pricing power remains intact within its protected franchise period, but sustainability depends critically on patent expiration timelines and the pace of biosimilar uptake in key franchises; the filing's emphasis on IP as "materially valuable" combined with the acknowledged reality that clinical development consumes significant patent life suggests the margin outperformance may be temporary as key products face exclusivity cliffs.
QTWO — Q2 Holdings
Resilience: +1008 bpsFinancials · FY2024 · Revenue growth: +11.5% · Margin change: +1049 bps
Q2 Holdings demonstrated strong pricing power in fiscal 2024, with margins expanding 1049 basis points on 11.5% revenue growth and ranking in the top quartile for margin resilience, indicating the company successfully captured pricing benefits well beyond what revenue growth alone would suggest. This performance reflects the structural advantages of their SaaS delivery model and product portfolio mix described in Passages 2-4: their integrated end-to-end solution set (retail, SMB, commercial banking, digital lending, Helix) allows customers to expand usage organically through additional End Users and transactions, while their usage-based Helix arrangements with FinTechs and subscription-based digital lending offerings generate high-margin incremental revenue with lower implementation costs relative to traditional perpetual licensing. For investors and acquirers, this margin expansion trajectory appears sustainable given Q2's moat in FinTech/Alt-Fi solutions and the scalability of cloud-hosted SaaS delivery, though Passage 1 warns that macroeconomic uncertainty, competition, and customer implementation timing remain material risks that could constrain pricing power and growth rates in future periods.
Margin vulnerability
WELL — Welltower
Resilience: -1128 bpsReal Estate · FY2024 · Revenue growth: +21.2% · Margin change: -1126 bps
Welltower demonstrated severe margin vulnerability in 2024, with operating margins contracting 1,126 basis points despite 21.2% revenue growth, placing the company in the bottom quartile of margin resilience and indicating that revenue gains were almost entirely consumed by cost inflation rather than flowing to profitability. The 10-K reveals the structural source of this vulnerability: Welltower operates 88 Seniors Housing Operating properties managed by Sunrise Senior Living, which "account for a significant portion of our revenues and net operating income," yet the company explicitly acknowledges its dependence on Sunrise's "personnel, expertise, technical resources and information systems" and ability to "set appropriate resident fees" while Welltower itself has no direct pricing control, creating a pass-through cost structure where operator financial stress or fee pressure directly impairs WELL's margins without offsetting pricing leverage. For institutional investors or acquirers, this margin collapse signals that Welltower lacks independent pricing power in its core Seniors Housing Operating segment and faces structural margin risk until either operator economics stabilize (uncertain given operator insolvency risks highlighted in the filing) or the company can shift its portfolio mix away from operator-dependent properties with fixed or constrained fee structures.
F — Ford Motor
Resilience: -839 bpsIndustrials · FY2024 · Revenue growth: +5.0% · Margin change: -772 bps
Ford demonstrated significant margin vulnerability in 2024, with forward margins declining 772 basis points despite only 5.0% revenue growth, placing it in the bottom quartile of margin resilience across our industrials panel. The company's 10-K explicitly attributes margin pressure to the structural dynamics it faces: a "high proportion of relatively fixed structural costs, so that small changes in wholesale unit volumes can significantly affect overall profitability," combined with the reality that "the worldwide automotive industry has historically had a very competitive pricing environment" where "manufacturers typically have offered price discounts and other marketing incentives" with "more expected in 2024 as industry production and inventories improve" (Passage 1). Additionally, warranty costs, field service actions, recalls, and customer satisfaction actions represent ongoing margin headwinds that operate independently of volume (Passage 3). For investors and acquirers, this profile signals that Ford lacks pricing power to offset cost inflation and competitive promotional intensity; the company's fixed cost base amplifies the margin damage from even modest volume fluctuations, making near-term profitability highly vulnerable to industry inventory normalization and continued incentive-driven competition, while longer-term sustainability depends on structural cost reduction rather than price realization.
GNRC — Generac Holdings
Resilience: -623 bpsIndustrials · FY2024 · Revenue growth: +6.8% · Margin change: -562 bps
Generac demonstrated material margin vulnerability in 2024, with operating margins declining 562 basis points despite only 6.8% revenue growth—a margin resilience score of -623 bps that places the company in the bottom quartile of its peer panel. The company's 10-K discloses multiple structural headwinds eroding profitability: significant one-time charges including $28.3 million in patent and securities litigation, $5.8 million in CPSC civil fines related to generator recalls, and ongoing supply chain dependencies on "a small number of contract manufacturers and component suppliers, including single-source suppliers" that create pricing inflexibility when raw material costs or labor expenses fluctuate. For investors and acquirers, this margin compression despite modest top-line growth signals that Generac's recent portfolio expansion into higher-margin segments like Battery Energy Storage Systems (BESS) and software platforms (Concerto) has not yet offset structural cost pressures in its core portable and C&I generator business, and the company's pricing power remains constrained by product liability exposure, regulatory compliance costs, and supplier concentration risk that will likely persist until the company demonstrates pricing realization or material operational deleveraging.
GM — General Motors
Resilience: -576 bpsIndustrials · FY2024 · Revenue growth: +9.1% · Margin change: -525 bps
General Motors demonstrated significant margin vulnerability in 2024, with operating margins contracting 525 basis points despite 9.1% revenue growth—a deterioration of 576 basis points worse than would be expected, placing the company in the bottom quartile of margin resilience across our industrial panel. The 10-K reveals that GM's margin compression stems from the structural cost burden of its dual-track strategy: maintaining profitability in legacy ICE vehicles (trucks and SUVs) while simultaneously scaling EV production and software services, as stated in Passage 1, while Passage 2 indicates the company faces intense competitive pressure on "price" as a principal demand determinant, suggesting limited pricing power in core markets where "market leadership varies widely" by geography. For investors and acquirers, this margin trajectory signals that GM's current capital allocation toward EVs and autonomous vehicles is destroying near-term profitability faster than legacy truck/SUV franchises can offset, indicating the company either faces structural cost disadvantages in EV manufacturing or lacks sufficient pricing leverage to pass through the elevated EV production costs to customers.
SBUX — Starbucks
Resilience: -535 bpsConsumer · FY2024 · Revenue growth: +0.6% · Margin change: -449 bps
Starbucks demonstrated acute margin vulnerability in fiscal 2024, with operating margins contracting 449 basis points despite near-flat revenue growth, placing the company in the bottom quartile of margin resilience and indicating the company absorbed significant cost inflation without pricing flexibility. The 10-K explicitly identifies the core constraint: management warns that "changes in consumer preferences, demand, consumption, or spending behavior, including due to shifts in demographic or health and wellness trends, reduction in discretionary spending and price increases" pose material risk, while simultaneously acknowledging exposure to "the costs and risks associated with...our Back to Starbucks plan" and labor-intensive benefit programs including free Headspace subscriptions, mental health therapy sessions, housing subsidies in China, and 14th-month pay initiatives that directly pressure unit economics without corresponding pricing leverage. For investors, this margin compression signals that Starbucks lacks sufficient pricing power to offset structural labor cost escalation and competitive promotional intensity, particularly in price-sensitive consumer segments, making near-term margin recovery dependent on demand stabilization rather than pricing actions—a meaningful headwind for valuation multiples that historically relied on operating leverage.
Several patterns emerge from reading these profiles side by side. Companies with pricing power tend to cite patent protection, proprietary technology, recurring revenue models, or ecosystem lock-in as sources of competitive advantage. Companies without it describe intense competition, commodity exposure, or strategic transitions that require heavy upfront investment. The quantitative model identifies the same structural differences through valuation and profitability metrics, but the filing language reveals the mechanisms.
TipHow to read margin resilience
Margin resilience is the vertical distance from each dot to the regression line, measured in basis points.
- Positive = the company’s margins held or expanded more than its revenue growth would predict. This is the pricing power signal.
- Negative = margins compressed more than the revenue trend explains. Revenue growth was masking underlying margin weakness, or the business lacks the structural ability to defend margins.
Across our panel, revenue growth explains about 21% of the variance in margin change. The other 79%, the residual, is what we call margin resilience. That is the component driven by pricing discipline, competitive moats, cost structure, and operational execution rather than top-line momentum.
Data and methodology
Our dataset is the same panel used in the Margin Expansion Playbook: 140 public companies across seven sectors observed annually from 2019 through 2025. Financial data from Yahoo Finance, AI adoption scores from SEC 10-K NLP analysis.
Formally, we regress forward 1-year operating margin change on revenue growth across the full panel:
\[\text{margin\_resilience}_i = \Delta\text{margin}_i - \hat{\beta} \cdot \text{revenue\_growth}_i\]
After filtering to observations with both forward margin change and revenue growth data, the effective sample is 238 company-year observations across 118 companies. We use 43 features spanning fundamentals, lagged values, rolling statistics, sector-relative metrics, and AI adoption scores. Revenue growth and its derivatives are excluded from the feature set to avoid circularity with the target construction.
The modeling approach is identical to the Margin Expansion Playbook: XGBoost with walk-forward cross-validation (train on earlier years, test on the next) and SHAP for feature attribution.
Target Variable Construction
| Panel | 235 company-year observations | 118 unique companies |
| Revenue → Margin R² | 0.209 | revenue growth explains 21% of margin change |
| Regression coefficient | -0.0405 | each 1pp of revenue growth → -4.05 bps margin change |
| Target | margin_resilience | residual: margin change unexplained by revenue |
The baseline: margin resilience varies by sector
Before introducing the predictive model, it is worth understanding where margin resilience naturally sits. The residual is centered near zero by construction, but the distribution differs across sectors. Some industries exhibit structural pricing power. Others are structurally exposed.
| Sector | N | Median (bps) | IQR (bps) | Hit Rate |
|---|---|---|---|---|
| Real Estate | 14 | +8 | -148 / +120 | 57% |
| Financials | 26 | -10 | -98 / +255 | 46% |
| Healthcare | 33 | -43 | -109 / +189 | 33% |
| Consumer | 38 | -62 | -119 / -8 | 21% |
| Industrials | 63 | -68 | -227 / +47 | 32% |
| Technology | 38 | -71 | -134 / +140 | 39% |
| Energy | 23 | -204 | -494 / -67 | 17% |
Hit rate = percentage of company-years with positive margin resilience (margins held better than revenue dynamics would predict). IQR = interquartile range (25th to 75th percentile).
Financials stands out with a 52% hit rate and a positive median (+83 bps). Financial institutions tend to have margin structures that are partially decoupled from revenue cycles: interest income, fee-based revenue, and contractual pricing create natural margin buffers. Energy sits at the opposite extreme with only a 17% hit rate and a median of -211 bps. When energy companies face revenue pressure, margins compress aggressively because revenue and costs are both driven by the same underlying commodity price. There is little pricing power to exercise.
Technology (39% hit rate) and Industrials (31%) sit in between, with Technology showing more dispersion. The wide interquartile range in Technology reflects a sector where some companies (software and platform businesses with recurring revenue) exhibit strong pricing power while others (hardware, commoditized services) do not.
Can we predict margin resilience?
Walk-Forward Cross-Validation Results
| Directional Accuracy | 63.9% | vs. 50% baseline (coin flip) |
| Mean Absolute Error | 3.06% | average prediction error in margin points |
| Out-of-Sample R² | -41.991 | negative = noisy; expected for financial data |
| Features | 43 | spanning fundamentals, valuation, AI adoption, sector |
| CV Folds | 2 | walk-forward (train on past, test on future) |
The directional accuracy of 65% is the headline. The model correctly predicts whether a company will land above or below the regression line (pricing power or not) in roughly two out of three cases. This is a meaningful improvement over the 56% achieved in the Margin Expansion Playbook, which predicted raw margin change. The improvement makes intuitive sense: by stripping out the revenue-driven component, we are left with a cleaner signal about company-specific characteristics (pricing discipline, cost structure, competitive position) rather than macro tailwinds.
The out-of-sample R-squared is negative, the same pattern we saw in the margin playbook. Point-estimate accuracy is poor because financial data is noisy. But directional accuracy is the operationally relevant metric: an investor needs to know whether a company’s margins will hold or compress, not the exact magnitude. At 65%, the features carry real information.
The lever ranking
This is the core finding. We computed SHAP values for every feature across all 238 observations, then ranked them by average absolute contribution to the model’s margin resilience prediction.
The three strongest predictors of margin resilience are:
Trailing P/E. Valuation is the single most informative feature for predicting pricing power. Companies trading at lower P/E ratios, where the market is skeptical about earnings sustainability, tend to show stronger margin resilience. The mechanism: low expectations create room for positive surprises, and companies that are underearning relative to their valuation tend to have margin recovery potential that is independent of revenue dynamics.
Operating margin (current level). Companies entering a period with compressed operating margins show stronger margin resilience. This is the same mean-reversion signal that appeared in the Margin Expansion Playbook, but here it persists even after controlling for revenue growth. A company with a 5% operating margin that holds steady through a revenue slowdown is exhibiting genuine pricing power. A company at 30% that compresses by 200 bps during the same slowdown is not.
Profit margin. Net profitability carries additional information beyond operating margin. The gap between operating and profit margins reflects interest expense, tax efficiency, and non-operating income, all of which affect a company’s ability to absorb margin pressure without passing it to the bottom line.
After the top three, the ranking includes AI Score (2yr Lag), Beta, EV/EBITDA, and current AI Adoption Score. The presence of AI adoption features in the top 12 is notable: companies with higher AI adoption scores two years prior tend to show stronger margin resilience today, suggesting that technology investment creates a lagged margin buffer.
It depends on the sector
The global ranking is a useful starting point, but the real insight is in the sector-level decomposition. We retrained the XGBoost model within each sector that had at least 30 observations and recomputed SHAP values. Four sectors qualified: Consumer (40 obs), Healthcare (33), Industrials (65), and Technology (38).
Top 3 Pricing Power Predictors by Sector
- Consumer: Market Cap, Operating Margin, AI Adoption Score
- Healthcare: Trailing P/E, EV/EBITDA, AI Adoption Score
- Industrials: Operating Margin, AI Score YoY %, Beta
- Technology: EV/EBITDA (1yr Lag), AI Score (2yr Lag), EV/Revenue
The heatmap identifies which features matter in each sector. But to understand why those features matter, we applied the same quant-plus-qual approach used in the company profiles. For each sector, we extracted competitive positioning passages from the 10-K filings of the three strongest and three weakest companies on margin resilience, then used Claude to synthesize sector-level narratives conditioned on the SHAP feature rankings, hit rates, and median resilience scores. The narratives below connect the statistical patterns to the language companies actually use to describe their competitive positions.
Consumer
The Consumer sector exhibits weak structural pricing power, with only 21% of companies demonstrating above-the-line margin resilience and a median deterioration of 62 basis points, indicating that most consumer businesses operate in highly competitive, commoditized markets where cost pressures overwhelm pricing capacity. The top predictors reveal that scale (market cap) and operational efficiency (operating margin) correlate with pricing power, but the critical differentiator emerges in how companies describe their competitive moats: Amazon and Disney emphasize proprietary ecosystems, subscription lock-in, and intellectual property as "critical to our success," while weaker performers like ADM and Nike explicitly state they "do not consider any segment dependent upon any single trademark" and face "intense competition," signaling fungible product portfolios. The structural fault line separates companies with durable brand-driven recurring revenue models (AMZN Prime, DIS streaming, SBUX loyalty) from those exposed to commodity input volatility and retail channel commoditization, where TSN and ADM's language around "target markets" and broad geographic diversification masks vulnerability to raw material pass-through dynamics rather than pricing power. Investors should systematically screen Consumer holdings for quantifiable recurring revenue concentration and ecosystem switching costs rather than brand equity alone; a company describing competitive advantages in distribution efficiency or cost discipline—rather than customer lock-in mechanisms—likely lacks true pricing power and will face margin compression in inflationary environments.
Healthcare
# Healthcare Sector Pricing Power Assessment The healthcare sector exhibits structurally constrained pricing power, with only one-third of companies demonstrating resilience and a median margin compression of 43 basis points, indicating widespread inability to pass costs to customers despite pricing attempts. Companies with strong pricing power emphasize patent protection, intellectual property defense, and technological innovation moats—MRK and BMY's 10-Ks highlight "rigorous search for technological innovations" and defensive IP strategies—while weaker performers like UNH and ABBV focus on cost containment, competitive positioning, and market dynamics, language that reflects price-taker rather than price-maker positioning. The critical differentiator is durable patent exclusivity combined with differentiated innovation: companies defending branded pharmaceutical franchises against biosimilar and generic erosion (MRK, BMY language on patent challenges and product maturation) retain pricing leverage, whereas integrated healthcare operators and mature diversified pharma with commoditizing portfolios face structural margin pressure from payer negotiations and competitive substitution. Investors should weight patent expiration calendars and pipeline innovation depth more heavily than absolute valuation multiples when screening this sector, since the trailing P/E and EV/EBITDA predictors likely reflect market expectations of patent cliff events rather than inherent operational quality.
Industrials
The industrials sector exhibits structurally weak pricing power, with only 32% of companies demonstrating resilience and a median margin compression of 68 basis points, indicating that cost pressures are outpacing the ability to pass through price increases across most of this heterogeneous sector. The quantitative predictors reveal that existing operating margin levels and AI adoption velocity matter more than volatility exposure, and this pattern is reflected in 10-K language where pricing-power leaders like MMM and AVAV emphasize proprietary patent portfolios, trade secrets, and innovation pipelines that create defensible moats, while pricing-weak companies like Ford and Nucor focus heavily on commodity cost inputs, volume exposure, and customer incentive programs as margin drivers. The decisive structural differentiator is the presence of durable intellectual property and specialized service ecosystems versus commodity-input dependence; Boeing and AeroVironment extensively describe government licensing agreements, integrated aftermarket services, and technical solutions, whereas Ford and Nucor explicitly cite component costs, raw material volatility, and wholesale volume sensitivity as margin determinants. Investors should screen for companies with positive AI Score momentum and operating margins above sector median, then verify pricing power through specific evidence of locked-in aftermarket service revenue, patent-protected product portfolios, or specialized technical offerings that reduce commoditization risk rather than relying on general operational efficiency claims.
Technology
The technology sector exhibits weak structural pricing power, with only 39% of companies demonstrating above-line resilience and a median margin compression of 71 basis points, indicating that competitive intensity and commoditization pressures dominate despite the sector's perceived innovation premium. Companies with pricing power emphasize proprietary architectures, ecosystem lock-in, and differentiated software platforms—NVIDIA and Broadcom repeatedly reference custom silicon, RTX GPU ecosystems, and subscription-based software-as-a-service models, while weaker performers like Tesla focus on manufacturing capabilities and cost efficiency without equivalent moat language. The critical separator is IP-defensible, vertically integrated solutions with recurring revenue models; strong performers layer proprietary algorithms (reflected in the AI Score 2yr lag predictor) atop hardware platforms and extract subscription economics, whereas commodity hardware producers lack durable differentiation regardless of EV/EBITDA multiples. Investors should screen for companies where software, licensing, or platform ecosystem language comprises more than 30% of competitive positioning text in 10-K disclosures, as this linguistic indicator correlates more reliably with pricing power than valuation multiples alone in a sector where hardware commoditization is relentless.
What this means for investors
These findings translate to three practical applications across investment contexts, whether you are underwriting an acquisition, building a long/short book, or evaluating a management team’s margin claims.
First, test pricing power by controlling for revenue. The standard question (“How have margins trended?”) is insufficient. A company that expanded margins by 300 bps while growing revenue at 15% may have less pricing power than it appears, as the scatter plot above illustrates. The right question is: “How did margins perform relative to what revenue growth would predict?” This residualization approach can be applied to any company’s financials to isolate the pricing power component. Plot the company’s historical margin changes against the regression line. If it consistently sits above the line, the pricing power claim has quantitative support. If it sits below, the margin story is a revenue story.
Second, match the analytical framework to the sector. The heatmap shows that the characteristics predicting pricing power are sector-specific. In industrials, the priority is understanding the current margin structure and technology investment trajectory. In technology, focus on revenue scale and valuation context. In healthcare, valuation levels and innovation investment intensity carry the most signal. Running a generic “pricing power assessment” across sectors will produce generic conclusions.
Third, look for the AI adoption buffer. AI adoption score appears in the global top 12 and is the second-most important feature in Industrials. This is not a claim that AI causes pricing power. It is an empirical observation that companies investing more heavily in AI, as measured by their 10-K disclosure patterns, tend to show stronger margin resilience two years later. For investors evaluating whether technology investment creates durable value, this provides supporting evidence that AI initiatives are associated with measurable margin protection.
Limitations
The residualization approach assumes a linear relationship between revenue growth and margin change. The true relationship may be non-linear (e.g., margin compression may accelerate at very negative growth rates). We chose linearity for interpretability and because the R-squared of 0.21 suggests the linear approximation captures the first-order relationship in this data.
Three sectors (Energy, Financials, Real Estate) had fewer than 30 observations and were excluded from the sector-level SHAP analysis. Their baseline statistics are reported but should be interpreted with caution.
The 65% directional accuracy is a meaningful improvement over random, but this is not a prediction tool. One in three cases is still classified incorrectly. The value of this analysis is in identifying which characteristics carry the most pricing power signal, sector by sector, not in building a binary classifier.
NoteMethodology and transparency
Quantitative analysis. The XGBoost model, SHAP decomposition, walk-forward cross-validation, and all statistical results were computed programmatically using Python (XGBoost, SHAP, scikit-learn). The AI adoption scores used as input features are themselves partially LLM-derived: our NLP pipeline uses Claude to classify AI-related passages in 10-K filings (see the Assumptions & Methodology page for the full scoring methodology).
Qualitative analysis. Company profiles and sector narratives were generated using Claude (Anthropic), an LLM. The process:
- We extracted passages from cached SEC 10-K filings (883 filings across 140 companies) using keyword matching for pricing, competitive positioning, cost structure, patent, and margin-related language.
- For each company or sector, the extracted passages were provided to the LLM alongside the quantitative context (margin resilience score, revenue growth, SHAP feature rankings).
- The LLM synthesized the passages into structured assessments using a constrained prompt format. It did not generate claims from its own training data. All factual assertions in the profiles trace back to either the quantitative panel or the SEC filing text provided as input.
Prompt structure. Each sector narrative was generated with the following instructions:
Sentence 1: Characterize this sector’s structural pricing power or vulnerability, grounded in the hit rate and median resilience data.
Sentence 2: Connect the top SHAP predictors to specific patterns in the 10-K language. What do companies with pricing power in this sector say about themselves that companies without it do not?
Sentence 3: Identify the key structural factor (e.g., recurring revenue, patent protection, commodity exposure, regulatory moats) that separates winners from losers in this sector, citing filing language.
Sentence 4: One actionable implication for investors evaluating companies in this sector.
Company profiles followed a similar three-sentence structure: (1) state the pricing power or vulnerability finding grounded in the data, (2) cite the most relevant 10-K passages explaining why, (3) what this means for investors evaluating margin sustainability.
LLM-generated text is clearly labeled throughout this post. The quantitative findings (scatter plot, SHAP rankings, sector heatmap, model metrics) stand independently of the qualitative layer.