Exponent, Inc. (NASDAQ: EXPO) — The Proof Print Arrived; Price Has Caught Up Faster Than the Thesis

Independent Investment Research
Target: Exponent, Inc. (NASDAQ: EXPO)
Sector: Industrials — engineering and scientific consulting
Date: 2026-08-23 · Coverage: UPDATE — follow-up to the 2026-07-08 report · Price at writing: $70.21 (8/21 close) · Shares outstanding: 47.546M (7/31/26) · Market cap: ~$3.34B · Cash: $66.6M · Funded debt: $0 · Indicative EV: ~$3.27B (operating leases excluded)
Leadership: CEO Dr. Catherine Ford Corrigan · President Dr. John D. Pye · CFO Eric Anderson · Chair Karen Richardson

This update re-verifies the durable analysis and integrates the Q2 FY2026 Form 10-Q, the July 30 earnings release and call, filings through August 18, and market data through August 21. Company operating metrics use revenue before reimbursements (“net revenue”) unless stated otherwise. FACT, INTERPRETATION, and OPEN QUESTION labels distinguish evidence from judgment. Third-party market data are unofficial and reconciled where possible to filings.

⚡ Claude’s Take

This is Claude’s subjective opinion, outside any institutional house view and not investment advice. It is the only section containing a recommendation or directional valuation zone; the institutional analysis below contains neither.

Verdict: HOLD / watchlist at $70.21; do not chase the proof print. Q2 delivered the operating evidence the July memo demanded: net revenue grew 12%, billable hours 9%, technical FTE 6%, utilization improved two points, and EBITDA margin expanded 90 basis points. The literal trigger was not a clean pass because consolidated utilization was 74%, one point below the pre-committed 75% hurdle, although the moat-bearing Engineering segment reached 76%. The more important conclusion is that the recovery is volume-plus-rate, not rate-only. That materially weakens the near-term AI-deflation bear case.

The price now reflects much more of that improvement. EXPO rose 15.2% from the July 8 report and 35.9% from the May five-year low. At $70.21, AZI shows 32.6x trailing EPS, 5.89x trailing sales, and a 24th-percentile composite valuation versus EXPO’s own history. The percentile still looks modest because the 2016–2021 comparison set was exceptionally expensive; the absolute multiple is again demanding for a business whose full-year utilization guide remains 72.5%–73.0%. The business is better resolved, but the prospective return is less asymmetric.

Directional zone: I would become constructive again in the low-to-mid $60s, where a 27–29x earnings multiple would better compensate for project concentration and a possible utilization ceiling; I would lean in below ~$60 if the operating evidence stayed intact. My broad base-value zone is ~$64–78. Above ~$80, the stock would require sustained high-single-digit volume growth, stable high-20s EBITDA margins, and no multiple contraction. The low-$50s bear case still exists if the large AI-device studies fade, utilization reverts to ~72%–73%, and the market values EXPO as a merely good consultancy. Conviction is high on business quality, medium on recovery durability, and low-to-medium on the current entry.

What changed my mind from BUY to HOLD is price, not franchise quality. Q2 broke the “one-quarter rebound” version of the bear case and contradicted the predicted rate-only signature. Yet one large user-research study was 4% of quarterly net revenue, twice the level management had expected, and is forecast to normalize to about 2% in Q3. Even excluding the unexpected two points, Q2 growth was approximately 9.8%, so the result was not only one project; still, Q3 is the clean repeatability test. The balance sheet also became less of a valuation cushion after $146.1M of first-half repurchases reduced cash from $221.9M to $66.6M. Those purchases were economically sensible near the lows and shrank the diluted share base, but the prior “$4.50 per share of net cash” floor is gone.

What flips me. A Q3 print near the 8%–10% guide with hours growing faster than FTE, utilization at 74%–75%, and the large study normalized would raise the quality of the recovery. A pullback toward the low $60s without a thesis break would improve the entry. Conversely, consolidated utilization back near 72%–73%, flat or negative hours, or E&H volume weakness spreading to Engineering would revive the structural-ceiling thesis. Tag: “The receipt checked out; the stock already collected much of the refund.”

Changes since 2026-07-08

📈 Stock Price Action — Five-Year Event Map

Prices are FACT from AZI adjusted daily data through 2026-08-21; causal attributions are INTERPRETATION unless tied directly to a company event.

At $70.21, EXPO is 41.6% below its November 2021 high, 13.5% below the February 2026 high, and 35.9% above the May low. It sits above the 21-day ($66.49), 50-day ($64.24), and 200-day ($66.12) EMAs. That is a materially healthier tape than July, but it is corroboration rather than a substitute for the Q3 volume test.

1. Executive Summary

Exponent is an asset-light engineering and scientific consultancy built around roughly 950 consultants across more than 90 disciplines. It is hired when product safety, litigation, regulation, or complex innovation makes technical error unusually costly. The economic moat is a reputation intangible reinforced by multidisciplinary scale and the selection risk clients face when choosing an expert whose work must survive adversarial scrutiny. Ten-year ROIC averaged approximately 20%, free cash flow has exceeded net income in each of the last three full years, and there is no funded debt.

Q2 materially improved the thesis evidence. Net revenue rose 12% to $148.9M; total revenue rose 21% because a large study carried unusually high reimbursable costs; billable hours rose 9%; technical FTE rose 6%; utilization reached 74% from 72%; EBITDA rose 16% to $42.7M; and EBITDA margin reached 28.7%. Engineering & Other Scientific, 85% of net revenue, posted 13% net-revenue growth, 10% hours growth, 7% FTE growth, and 76% utilization. Environmental & Health improved in Q2 but remains the weak spot on a first-half basis.

The prior report’s near-term bull trigger is a partial pass: growth and hours cleared the hurdle; consolidated utilization missed 75% by one point. The bull falsification test is not hit because Q2 did not resemble 73% utilization with flat hours and rate-only growth. The bear’s one-quarter-blip formulation is broken, while the broader bear is not yet falsified because Q3 and Q4 remain unreported and full-year utilization is still guided near 73%.

AI is currently a net demand benefit, but the form of that benefit creates a new concentration question. One user-research study expanded to approximately 4% of Q2 net revenue and should moderate to about 2% for the rest of the year. Management reports more clients, studies, industries, and device types, but gives no client counts. The durable thesis therefore shifts from “is volume returning?” to “is returned volume broad enough to persist after the exceptional study normalizes?”

Capital allocation was aggressive: $146.1M of first-half repurchases and $31.3M of dividends, followed by a $50M authorization increase. The purchases appear well-timed against the May–June lows and reduced diluted shares 5% year over year, but cash fell to $66.6M. Balance-sheet risk remains low because there is no funded debt; valuation support from excess cash is much smaller.

At $70.21, market capitalization is approximately $3.34B using the July 31 share count and indicative enterprise value is about $3.27B excluding operating leases. AZI reports 32.6x trailing EPS, 5.89x sales, and a 24th-percentile composite valuation. The stock is still below its exceptional own-history median, but current absolute expectations require the recovery to persist. The research conclusion is therefore two-sided: franchise quality and volume evidence strengthened; margin of safety narrowed.

2. Business Overview

2.1 What the company actually does

Exponent is a multidisciplinary engineering and scientific consulting firm. Its consultants — engineers, scientists, physicians, regulatory and environmental specialists — are hired to investigate accidents and failures, support litigation as expert witnesses, evaluate the safety and reliability of products before and after launch, and navigate regulatory requirements. The work splits two ways that matter more than the formal segments:

• Reactive work (~60% of the business — management, Q1 2025 call). Failure analysis, accident reconstruction, product-recall investigations, and litigation/dispute support. This is the high-margin core, it is counter-cyclical and largely non-discretionary (failures, lawsuits, and recalls happen regardless of the economy — sometimes more in downturns), and it is where the brand/credibility moat is most valuable.
• Proactive work (~40%). Product-development engineering support (dominated by consumer-electronics user research and human-factors studies), reliability testing, regulatory and environmental/health consulting. This tracks client R&D and capital budgets — and, as the deep-dive below establishes, it is concentrated enough in a few large technology clients that their product cycles effectively set EXPO’s firmwide growth rate.

Management stated the approximately 60/40 split on the Q1 2025 call and has not restated it since. By Q3 2025, reactive work was up 18% while proactive was approximately flat; proactive then led the Q1 and Q2 2026 recovery. Applied to FY2025 net revenue of $536.8M, 40% implies approximately $215M of proactive revenue, but this is an analytical estimate rather than a reported segment.

2.1a The proactive book, decomposed (new since the prior report)

Management’s per-industry mix commentary (Q1’25 call) lets the proactive book be decomposed with useful precision:

Interpretation (derived, not management-stated): of the ~$215M proactive book, consumer electronics is ~40% (~$75–90M, i.e., ~14–17% of total firm revenue), utilities ~25%, chemicals + life-sciences regulatory ~25–28%, other ~10%.

The anchor-client fact pattern (FY2023 10-K, Note 15): no client exceeded 10% of revenue in 2023–2025, but one client was 13% in 2021 and 15% in 2022. The client is unnamed; identifying it as any specific consumer-electronics company would be speculation. The drawdown and recovery nevertheless correlate with consumer-electronics product cycles. Q2 adds an important update: one user-research study reached approximately 4% of quarterly net revenue, above management’s prior 2% expectation, before moderating late in the quarter. Management says the practice now spans more clients, studies, industries, and device types, but disclosed no counts. Concentration improved from 2021–2022, while project lumpiness remains material.

2.1b Origins — why the history matters

Exponent traces to Failure Analysis Associates, founded in 1967 by Stanford engineering professors to investigate why structures, vehicles, and products fail. The firm’s identity and credibility were built over nearly six decades of high-stakes forensic work — crashes, fires, explosions, structural collapses, product defects — long before “litigation consulting” was an industry. The rebrand to Exponent (1998) and the broadening into litigation support, product-safety/reliability, regulatory, and environmental/health sciences layered adjacencies onto a forensic core whose reputation was already established. The practical consequence: the moat (§4) is the cumulative product of ~60 years of Daubert-surviving testimony — it cannot be bought or built quickly, which is why it is durable and why the firm earns ~28% returns in a nominally “low-barrier” industry.

2.2 Reporting segments

• Engineering & Other Scientific (~85% of net revenue). Mechanical, electrical, materials, civil/structural, biomechanics, polymer science, electronics, vehicle engineering, human factors, and data science. Q2 net revenue rose 13%, hours 10%, FTE 7%, and utilization reached 76%. It is the healthy engine and contains the strongest moat-bearing work.
• Environmental & Health (~15%). Chemical regulation/EHS, toxicology, epidemiology, and environmental and health sciences. Q2 net revenue rose 9%, hours 3%, FTE was flat, and utilization improved to 66%. First-half hours still fell 2%, leaving this segment the soft spot and the largest concentration of standardized work plausibly exposed to automation.

2.3 How it makes money — the consulting-firm KPIs

• Net revenues (revenues before reimbursements) — the primary top line (~92% of gross; reimbursements are zero-margin pass-throughs). FY2025: $536.8M.
• Utilization — billable share of available consultant hours; the single most important profitability driver (§6).
• Technical FTE headcount and billable hours — the capacity.
• Billing rates / revenue per FTE — the pricing. Standard professional rates run $225–$1,375/hour, raised annually at fiscal-year start (FY2025 10-K).

Because the firm sells expert hours, its only “inventory” is the time of credentialed people, and margin is set by how fully and at what rate that time is billed — the asset-light, talent-driven model that produces both the high returns and the capacity constraint.

2.4 Customers, end markets, and recurrence

Clients are corporations, law firms, insurers, and government agencies. FY2025 end-market mix: consumer products ~21%, energy/utilities ~20%, transportation ~16%, chemical ~11%, plus life sciences and construction. Revenue is project-based, not contractually recurring, but a more durable kind of recurrence operates: a continuous flow of failures, disputes, recalls, and regulatory events, plus repeat corporate and law-firm clients who return because the prior work product holds up in court. Client concentration at the top has improved (no >10% client since 2023 versus 13–15% in 2021–22), though the proactive book remains few-client-lumpy. EXPO finances itself entirely from operations.

3. Industry Dynamics

3.1 A good business in a structurally mediocre industry

The broad science & engineering / expert-witness consulting industry is structurally mediocre: highly fragmented, talent-mobile, and characterized by low barriers to entry at the individual-discipline level. No decision-grade TAM or market-share denominator was identified in the update. EXPO’s own 10-K concedes that barriers to entry are low and competition is increasing, while also flagging client independence concerns.

The resolution of the paradox — a mediocre industry housing a 28%-ROE business — is that EXPO competes in a defensible niche within that industry: low barriers to enter any one discipline coexist with high barriers to replicate the integrated platform — a ~60-year Daubert-tested brand, ~1,000 consultants across 90+ disciplines under one conflict-clean roof, and the independence that makes it a credible third party precisely because it is not the client’s in-house team. The honest framing stands: a good business in a mediocre industry.

3.2 The value chain and the competitive set

EXPO’s competitors are heterogeneous, and none replicates the full bundle: technical/forensic consultancies (ESI, S-E-A, Rimkus, Robson Forensic — smaller, regional, or narrower), litigation-economics firms (CRA International — the closest public comp — FTI, Cornerstone — skewed to economic/financial expert work), in-house corporate engineering teams (which lack court-required independence), individual academic experts (credible but un-scaled), and large E&C/EHS firms (ERM, Ramboll, Arcadis, WSP — environmental-side competitors, not litigation-grade failure analysis). EXPO sits as the premium, independent niche leader in court-credible multidisciplinary failure analysis; the barrier is the integrated, conflict-clean, brand-name bundle, not any single discipline.

3.3 Demand drivers and cyclicality

Overall cyclicality is low-to-moderate, set by the 60/40 reactive/proactive mix:

• Reactive demand (~60%) is relatively defensive — driven by litigation volume, recall frequency, product complexity, insurance matters, and regulatory enforcement rather than only GDP. It is not fully non-cyclical because clients can still delay or settle matters.
• Proactive demand (~40%) is pro-cyclical and — the sharper formulation this update adds — product-cycle-driven with client concentration: the swing factor is a handful of CE clients’ device pipelines (plus utility risk budgets and chemical-regulatory activity). The 2023–25 soft patch and the current re-acceleration are both, at bottom, the same client-cycle phenomenon.

The net business is far less cyclical than most industrials but not acyclical: the recent slowdown was precisely the pro-cyclical proactive book, not the reactive moat.

3.4 Secular tailwinds — and the AI question, now with numbers

Tailwinds: rising product complexity (EV batteries and thermal events, AI/data-center hardware, autonomous systems, medical devices), litigation and regulatory intensity, and chemical-regulation pipelines. Regulation is a structural support: when technical evidence is offered through expert testimony, Daubert and FRE 702 require a qualified witness using reliable methods. Software can assist the work but cannot itself be cross-examined, which supports the human-expert role in the reactive core.

On AI, this update can be far more precise than the June report:

• The demand side is currently winning. Management previously sized AI-related work in the mid-teens percentage of the business, directly or one step removed. Q2 provided hard operating corroboration: user research for AI-enabled hardware helped drive a 4%-of-net-revenue study while company hours rose 9%. Management described work across glasses, immersive systems, audio-first products, wearables, medical devices, data-center power and cooling, ADAS, and utility infrastructure. These use cases require physical experiments, human subjects, and cross-disciplinary interpretation rather than only desk research.
• The displacement side is real enough that the company itself now flags it. The FY2025 10-K (filed 2026-02) added a first-ever standalone risk factor: AI technologies “may reduce demand for our services… may offer automated alternatives as a substitute for our services… may also render some of our existing service offerings less valuable or obsolete, particularly those that are more standardized in nature.” That final clause is the map to the soft spot: standardized, document-heavy work — dossier preparation and literature review inside chemicals/life-sciences regulatory, commodity testing — concentrated in the Environmental & Health segment (which is already the weak segment: Q1’26 hours −5%). The industry backdrop is consistent: the UX-research labor market contracted sharply in 2024–25 (~21% of companies laid off UX researchers in 2025), and synthetic-user research is credible for well-documented software behaviors while remaining weakest for novel physical hardware.
• The subtler risk is efficiency deflation, not substitution: EXPO bills time and materials; if AI compresses hours per engagement, revenue per project deflates unless offset by volume or rate. FY2024–2025 showed rate-led growth with falling hours. Q1 2026 hours rose 6%, and Q2 accelerated to 9% alongside approximately 4% rate growth. The predicted deflation signature is therefore absent in the latest two quarters; Q3 after the large study normalizes is now the key test.

3.5 Verdict — industry & niche

Unchanged: the broad industry is mediocre; EXPO’s niche within it is structurally attractive and defensible. The legal moat (Daubert-required human experts), the counter-cyclical reactive core, secular complexity/regulation tailwinds, and an AI dynamic that is currently net-feeding demand make EXPO’s position far better than the fragmented “consulting” label suggests. The live risks are end-market/client-cycle concentration in the proactive book and the AI-automation exposure of its standardized slice — cyclical and contained, respectively, on today’s evidence; neither is moat erosion. The verdict would change if the deflation signature (rate-only growth, falling hours) reappeared and persisted.

4. Competitive Position

4.1 Naming the moat (Greenwald)

• (a) Intangible asset — the primary moat. A ~60-year, court- and regulator-tested reputation for credible, independent technical analysis. Clients who could do the engineering in-house retain Exponent specifically for third-party credibility that an internal team structurally cannot supply. The barrier is time and track record — decades of Daubert-surviving testimony.
• (b) Economies of scale within a niche — the reinforcing moat. 90+ technical disciplines and 700+ doctorate-holders under one roof let EXPO staff complex cross-disciplinary failures (a battery fire is simultaneously a materials, electrical, thermal, and human-factors problem) that no boutique can — and the scale is conflict-clean and independent.
• © Selection / track-record costs. In litigation you hire the firm whose experts win; switching to an unproven expert carries real case risk. A softer, per-engagement switching cost that reinforces repeat retention.

There is no classic contractual demand captivity — but credibility-based selection in high-stakes matters functions similarly.

4.2 Where the moat shows up in the numbers

• Pricing power. In FY2025 both segments grew revenue on higher billing rates even as billable hours fell ~2% — raising price into softening volume is the behavior of inelastic, credibility-based demand. Net revenue per FTE rose ~16% over two years to ~$552k. Standard rates run $225–$1,375/hour, raised annually.
• Margins. ~22–23% operating margin on net revenue — roughly double the ~10–13% at peer consultancies (CRAI, FTI, Huron).
• Returns. Ten-year ROIC averages approximately 20%, with FY2020–FY2025 annual readings around 18%–28%. ROE has generally exceeded 25%, though buybacks mechanically elevate it by reducing equity.

If you removed “the Exponent name,” billing rates, utilization, and returns would all deteriorate — the moat ties directly to financial outcomes, satisfying the §9 test.

4.3 The honest counter-evidence, reconciled

The bear must contend with EXPO’s own disclosures — “barriers to entry are low,” “competition is increasing,” client “independence concerns,” and now the new AI-displacement risk factor. The reconciliation stands: barriers are low to enter any one discipline and high to replicate the integrated, court-credible, conflict-clean platform; the moat sits at the firm/reputation level. The genuine vulnerabilities are (1) talent — the moat is the PhD bench, the model is capacity-constrained, and utilization has historically topped out in the low-to-mid 70s; and (2) the newly-specific automation exposure of standardized work — real, but bounded (the affected slice is at most ~a quarter of the proactive book, i.e., ~10% of firm revenue, and it sits furthest from the brand moat).

4.3a The talent engine — the moat’s power source

The profit-sharing bonus pool (33% of pre-bonus pre-tax profit) does triple duty: aligns the workforce with firm profitability, stabilizes margins (the accrual flexes down when utilization falls — §6.4), and retains elite PhDs by tying compensation to the collective enterprise rather than portable individual books. The client value proposition is the Exponent platform and brand, not a single star expert — so individual departures damage less than at a boutique. The flip side: growth requires continuously recruiting scarce elite talent and billing it at high utilization, which caps the achievable growth rate regardless of demand.

4.3b Direct contrast — EXPO vs. CRA International

CRA (CRAI) remains the closest public comparable and a useful mirror. CRA’s Q2 revenue grew 12.8%, utilization was 77%, and six practices—including Forensic Services—grew double digits. FTI’s Forensic and Litigation Consulting segment grew 4.1% with higher rates but lower utilization. These reads argue against an industry collapse and support differentiated demand at EXPO. CRA skews toward economic/financial expert work, while EXPO’s hard-science and engineering niche is narrower and more multidisciplinary. EXPO’s margins remain materially higher, although definitions differ. A valuation premium is directionally justified; its size is the debate.

4.4 Verdict — competitive position

Unchanged: a durable competitive advantage — intangible asset (court-tested brand) plus multidisciplinary scale, validated by pricing power and ~28% ROE. Durability risks: key-person/talent dependency and the utilization ceiling; reputation/independence (a single compromised high-profile engagement); and — added this update — automation of the standardized tail of the proactive book. None is currently in evidence in the numbers; all warrant monitoring. EXPO remains a higher-quality business than CRAI, FTI, Huron, or ICF on margins, returns, and balance sheet.

5. Growth History and Forward Opportunities

5.1 Historical growth — steady, organic, talent-gated

EXPO’s growth is organic (essentially no M&A; goodwill flat at $8.6M) and historically a high-single-digit net-revenue compounder: FTE growth × utilization × billing-rate increases. Net revenues: $497.2M (FY2023) → $518.5M (FY2024) → $536.8M (FY2025). Headline rates (+4.3%, +3.5%) understate the underlying trend because FY2024 was a 53-week year; on a 52-week-equivalent basis FY2024 grew only ~2.3%, making FY2025’s +3.5% an acceleration like-for-like — material to the “slowdown” narrative and frequently missed.

The choppy 2023–25 optics trace to a utilization cycle of management’s own making: technical FTE +10% in 2023 against +2% billable hours (utilization 74%→69%); FTE −8% in 2024 restoring 73%; FY2025 flat FTE with growth entirely on rate. A normalization cycle, not a structural decline — and, per the deep-dive, its demand-side trigger was specific: the anchor CE client’s product-development drawdown.

5.2 Forward opportunities

• Cyclical recovery (the near-term lever). Q3 2025 net revenue grew 10%; Q1 2026 grew 10% with hours +6%, FTE +5%, and utilization 76%; Q2 grew 12% with hours +9%, FTE +6%, and utilization 74%. FY2026 guidance is now 9%–10% growth. The recovery spans three prints and both volume and rate.
• The AI-device cycle. Novel form factors require instrumented human-factors validation. The Q2 4%-of-revenue study proves material demand, while its expected normalization and undisclosed client breadth make repeatability the essential question.
• Reactive secular tailwind. Record recall volumes, rising litigation/Daubert demand, EV/battery failure analysis, AI-product-liability as a nascent category, PFAS/EHS regulatory pipeline.
• Pricing power. Demonstrated ability to push rates (~+5–6% implied on Engineering in FY2025) provides a growth lever independent of headcount.
• Operating leverage on utilization. If utilization holds the mid-to-high 70s on a roughly fixed cost base, incremental margins are high — the bull-case earnings lever.

5.3 The constraint, and the verdict

The binding limit remains talent capacity. Utilization has historically topped out in the mid-70s, and management still guides full-year 2026 to 72.5%–73.0%. Whether hours continue to outgrow FTE after the exceptional study normalizes separates a return to the historical 8%–10% algorithm from a rate-led mid-single-digit grower. Verdict: high-quality, organic, talent-gated growth; Q2 confirmed the reacceleration, while Q3 must establish its breadth and repeatability.

6. Financial Quality

6.1 Q2 reconciliation — net revenue is the controlling top line

FACT. Reimbursable expense equals reimbursement revenue. The $29.65M increase in gross revenue included $13.66M of economically neutral pass-through, so 20.9% gross growth is not the correct organic-growth number. Net-revenue growth of 12.0% is controlling.

FACT. Consolidated operating income contains a deferred-compensation presentation trap. Q2 corporate expense fell partly because deferred-comp expense was $5.18M lower year over year, while the corresponding investment gain in miscellaneous income also fell. The pairing is approximately pre-tax neutral. EBITDA margin and segment operating income, which rose 14.4%, are cleaner reads on operating leverage.

INTERPRETATION. Q2 was a high-quality operating quarter. Hours grew faster than FTE, utilization improved, realized price remained positive, and margin expanded. EPS growth exceeded net-income growth because Q2 diluted shares declined 4.9%. Stock compensation is the counterweight: Q2 SBC increased 27% to $6.68M, or 4.5% of net revenue. Buybacks more than offset issuance, but the compensation burden itself did not decline.

6.2 Segment operating bridge

Engineering & Other Scientific grew through consumer-electronics user research, utility risk and infrastructure work, and reactive engagements in consumer products, chemicals, and transportation. Environmental & Health grew on projects evaluating chemicals’ effects on health and the environment. The latter improved sequentially, but its first-half hours were still down 2%, utilization was flat at 68%, FTE fell 1%, and segment operating income rose only 1.9%. The evidence therefore supports two simultaneous statements: company-wide hours deflation did not occur, and the most standardized segment remains the weak link.

6.3 Earnings history and TTM bridge

The FY2025 net-income decline was largely tax and timing noise: pre-tax income was flat, while the stock-comp tax-benefit swing, Arizona land-lease straight-line cost, biennial managers’ meeting, and lower interest/rental income obscured higher segment profit. The TTM period now shows real operating acceleration, but not the historical cash-conversion premium.

6.4 Cash flow — the new yellow flag

FACT. Accounts receivable reached $218.5M, up 20% from year-end and 28% year over year, ahead of net-revenue growth. Billed receivables rose roughly 35%, while unbilled receivables rose about 12%. The allowance remained approximately $7.5M; first-half write-offs of $2.07M were covered by a similar provision.

INTERPRETATION. Working capital is a watch item rather than an earnings-quality break. The unusually large study and reimbursement spike plausibly increased billed receivables, and there is no accompanying rise in the allowance or disclosure of collection trouble. Nevertheless, the prior claim that FCF consistently exceeds net income is no longer true on a TTM basis. Q3 collections should convert the growth into cash; failure to do so would make the 12% print lower quality than it appears.

6.5 Balance sheet and return on capital

At July 3, cash was $66.6M, current assets $312.0M, current liabilities $157.8M, and there was no funded debt. The approximately $80.7M balance that data vendors may label debt consists of operating-lease liabilities. Deferred-compensation assets of roughly $142M are nearly offset by related liabilities; they are not freely distributable excess cash and remain subject to creditors. The current ratio is approximately 2.0x.

Cash declined $155.3M from year-end because the company returned substantially more than first-half FCF. The balance sheet remains safe, but “net cash” has changed from a major valuation cushion into ordinary operating liquidity. Management identifies $50M–$70M as a comfortable cash range, placing the quarter-end balance near the bottom of that range.

ROIC.ai reports annual ROIC between 17.7% and 28.3% over FY2020–FY2025 and roughly 20.8% annualized for the latest quarter. A ten-year series averages about 20%. The exact calculation is definition-sensitive for an asset-light company with cash, leases, and a deferred-comp trust, but the Greenwald economic test clearly passes: returns stayed far above a reasonable cost of capital through both growth and utilization cycles. ROE reached roughly 31% in Q2, though buyback-driven equity contraction mechanically lifts it.

6.6 Financial-quality verdict

INTERPRETATION — high quality, with cash conversion newly on watch. The company has a debt-free balance sheet, high and persistent returns, low capital intensity, pricing power, and EBITDA margins that expanded with utilization. The Q2 volume bridge is strong. The qualifications are rising SBC, working-capital absorption, a now-smaller cash reserve, and a project whose reimbursement intensity temporarily distorted gross growth. None implies solvency or accounting risk; each matters to per-share value and valuation.

7. Capital Allocation

7.1 First-half scorecard

The Q2 monthly detail is unusually useful: 109,000 shares at $64.96, 313,000 at $60.10, and 703,000 at $59.00. Only $0.342M remained under the prior authorization at quarter-end. The board added $50M on July 30.

INTERPRETATION. Capital allocation passes on timing and strategic coherence. Management bought little during the most expensive years, increased activity during the de-rating, and concentrated purchases near the trough. The action created genuine per-share accretion: July 31 shares outstanding were 47.546M versus 48.610M at April 8, and weighted diluted shares declined 4.9% year over year in Q2.

Sizing is more debatable. First-half FCF was only $25.7M against $177.4M of dividends and repurchases, so the company deliberately liquidated much of its excess cash. Lower cash also reduced Q2 interest income by $1.63M year over year. Management expects more than $100M of FCF in the second half and said repurchases should be slower than during the prior twelve months. That cadence is sensible: another $146M half-year would require borrowing or push liquidity below the stated comfort range.

7.2 Reinvestment, dividend, and incentives

Exponent reinvests mainly through recruiting, training, laboratories, and offices rather than acquisitions. Goodwill remains only $8.6M. In a talent-constrained business, headcount growth is the capital-cycle variable: Q2 hours grew faster than FTE, so current hiring is productive. The risk appears when headcount growth persists after demand slows and utilization falls.

The dividend has risen for 13 consecutive years and consumes roughly half of normalized earnings. It is a durable component of capital return, but not a downside guarantee. The annualized yield is 1.77% at the current price.

Firm-wide bonuses equal 33% of pre-bonus pre-tax profit, helping variable compensation absorb revenue volatility. Executive compensation is modest relative to company size and a large portion is deferred into equity. The weak point remains the absence of a clear per-share, ROIC, or relative-TSR measure in the CEO scorecard. Rising SBC—5.2% of first-half net revenue—also deserves more attention now that excess cash is smaller.

7.3 Insider and registration update

There were no code-P open-market purchases and no discretionary code-S sales after July 8. CEO Corrigan executed planned sell-to-cover transactions in July and August, retaining a net 642 and 672 shares, respectively. On July 28 she exercised 12,665 options at $37.45 under no 10b5-1 plan and sold none; her direct ownership rose to 128,436 shares, then to 129,108 after the August planned transaction. This is mildly positive texture but not equivalent to an open-market purchase.

Seven S-8 post-effective amendments filed August 11 were one administrative deregistration of unsold shares under obsolete equity plans. They are not a new share registration or incremental dilution.

7.4 Capital-allocation verdict

INTERPRETATION — disciplined timing, aggressive sizing, reduced flexibility. The board acted when valuation was favorable and materially reduced the share base. That is good stewardship. The next test is restraint: repurchases should be funded by restored cash conversion and executed on weakness, not used automatically because authorization exists.

8. Changes and Headwinds — Last Two Years

8.1 What the recovery resolved

The operating sequence is now clear. EXPO over-hired into 2022–2023, utilization fell to 69%, and the company right-sized FTE in 2024. Proactive consumer-electronics demand then softened, producing approximately flat Q1 and Q2 2025 net revenue. Q3 2025 grew 10%, Q1 2026 grew 10%, and Q2 2026 grew 12%. Q2 is the first post-baseline quarter to combine a raised guide, higher hours, higher headcount, better utilization, and margin expansion.

That sequence breaks the “Q1 was a one-quarter rebound” claim. It also makes the February–May 2026 de-rating look primarily like a multiple/category event rather than an earnings warning. The stock’s reaction after Q2—3.5% the next session on 2.3x normal volume and a sustained 200-day-average reclaim—shows the market recognized the operating information.

8.2 What remains unresolved

The recovery has a concentration component. A large user-research study was roughly 4% of Q2 net revenue versus the prior 2% expectation. Removing the entire study produces approximately 7.6% growth; removing only its unexpected two-point expansion produces approximately 9.8%. Thus the quarter was not wholly project-driven, but the 12% headline is not the cleanest run-rate measure. Q3 guidance of 8%–10% while the study normalizes is the next breadth test.

Full-year utilization remains guided at 72.5%–73.0%, unchanged from FY2025’s 72.5%, because Q4 is seasonally weaker. This is the best evidence for the surviving bear case: EXPO can post excellent individual quarters without establishing a structurally higher annual utilization ceiling.

Environmental & Health also remains mixed. Q2 hours grew 3%, but first-half hours fell 2%. Standardized regulatory, survey, and data-collection work is the part of the portfolio most plausibly exposed to automation or client efficiency demands. Weakness has not spread to Engineering, which grew first-half hours 10%.

8.3 Run-rate hygiene

• The large study should fall from approximately 4% to 2% of net revenue.
• Reimbursements are zero-margin and should not be capitalized as organic growth.
• Deferred-comp expense and trust-asset gains offset pre-tax but distort operating and miscellaneous lines.
• Lower cash reduces interest income.
• The FY2025 biennial managers’ meeting does not recur in 2026; the next is expected in 2027.
• Arizona land-lease cash rent rises from roughly $1M to $6.2M annually in January 2028; straight-line expense already recognizes part of the burden.
• FY2026 capex guidance of $12M–$14M implies heavier second-half spending after only $4.2M in H1.
• FY2026 SBC guidance of $27.9M–$28.4M implies roughly 4.7%–4.9% of guided net revenue.

8.4 Upcoming events

The September 29 Investor Day at the Phoenix Test & Engineering Center is the next qualitative catalyst. It should be judged on quantitative disclosure around proactive/reactive mix, client concentration, AI-related revenue, hiring capacity, utilization, and data-center infrastructure opportunity—not presentation polish. Q3 results are the next hard operating test. Q4 then completes the prior bear-falsification window and provides the seasonal utilization comparison.

9. Risk Analysis

The highest-conviction risk is valuation interacting with an operating miss. At a lower multiple, a one-quarter project normalization would be tolerable; at the current multiple, it could trigger another material de-rate. The highest-consequence franchise risk is reputation, but there is no current evidence of impairment. The most actionable accounting risk is collections, not recognition: receivables need to turn into cash in H2.

AI deserves a two-sided treatment. Direct substitution is least plausible in bespoke physical experiments, cross-disciplinary failure analysis, regulated-device studies, and expert testimony. Efficiency deflation is more plausible in document review, standardized testing, surveys, and regulatory dossiers. Conversely, AI embedded in hardware, data centers, energy systems, transportation, and medical devices creates more failure modes and human-factor questions. Q2 shows the benefit side in reported hours. It does not settle the long-run net effect.

10. Valuation Discussion — Embedded Expectations, No Price Target

10.1 Current cross-check

Using the July 31 share count of 47.546M and the August 21 close of $70.21 gives market capitalization of approximately $3.338B. Subtracting $66.6M of cash and adding no funded debt gives indicative enterprise value of approximately $3.272B. Operating leases are excluded from this enterprise-value presentation and treated as operating obligations.

AZI’s standardized feed reports $2.1557 of TTM EPS and therefore 32.57x P/E, 5.89x P/S, 10.09x P/B, and a 24.4th-percentile composite valuation versus company history. Its components are P/E 8.9th percentile, P/S 17.9th, and P/B 46.4th. The filing-derived $2.23 EPS is preferred for absolute P/E; the feed is retained for internally consistent historical percentiles.

INTERPRETATION. The own-history percentile can mislead if read alone because EXPO spent much of the last decade at very high multiples. The stock is below its own historical median and expensive in absolute terms. Both statements are true. The correct question is what growth and durability the current price requires.

10.2 Embedded expectations

At approximately 31.5x earnings, an investor’s return depends on some combination of earnings growth, dividends, and terminal multiple. Illustrative three-year arithmetic, not forecasts:

The table shows why quality alone is insufficient at the present price. A mid-single-digit grower that normalizes to a high-20s multiple produces limited return; a satisfactory outcome requires upper-single-digit per-share growth with a premium multiple preserved. The dividend adds less than two percentage points annually. Buybacks can add per-share growth, but future pace is constrained by the reduced cash balance and should not be extrapolated from H1.

A reverse-DCF framed on approximately $109M of TTM FCF, a 2.5% terminal growth rate, and a ten-year first stage implies roughly mid-to-high-single-digit long-run FCF growth at an 8%–9% cost of equity, with the precise answer highly sensitive to discount rate and working-capital normalization. That demand is close to the historical algorithm rather than the depressed FY2024–2025 rate. The July setup offered recovery optionality at a lower requirement; the current setup prices in a meaningful portion of the proof.

10.3 Peer and quality context

CRA International and FTI are imperfect but informative. CRA’s Q2 revenue grew 12.8%, utilization was 77%, and Forensic Services grew double digits. FTI’s Forensic and Litigation Consulting revenue grew 4.1%, with higher rates but utilization falling to 54%. EXPO’s 28.7% EBITDA margin on net revenue remains substantially above CRA’s consolidated 12.7% non-GAAP EBITDA margin and FTI Forensic’s 16.1% segment margin, though definitions and corporate allocations differ.

The premium is therefore economically justified by better margins, no funded debt, and a narrower reputation moat. The valuation risk is not that EXPO should trade exactly at peer multiples; it is that the premium can shrink if hours normalize or cash conversion disappoints. February–May demonstrated that premium compression can occur without a solvency or moat event.

10.4 Valuation verdict

INTERPRETATION — improved fundamentals, fuller expectations. Current multiples require durable volume growth after the large study normalizes. Own-history percentiles remain below median but do not provide a large absolute margin of safety. The valuation debate has shifted from “does the market wrongly price a structural decline?” to “how much of the recovery is already capitalized?”

11. Variant Perception

11.1 Consensus and strongest bull case

The apparent consensus is that EXPO is an unusually high-quality consultancy whose 2023–2025 slowdown was temporary, but whose people-based model deserves a lower multiple than in 2021. Q2 moved consensus toward recovery: volume, price, capacity, and margin all improved, and guidance rose.

The strongest bull case is more specific. The 2023 downturn reflected one large consumer-electronics product cycle plus self-inflicted over-hiring, not moat erosion. AI is not merely benign; its migration into physical products and infrastructure expands demand for user research, biomechanics, safety, reliability, human factors, and failure analysis. The reactive core remains protected by the need for independent, defensible human expertise. Q3 2025, Q1 2026, and Q2 2026 now form a multi-print recovery, with Q2 hours up 9%. The large study’s unexpected contribution explains only about two growth points. If Q3 holds 8%–10% as it normalizes, the recovered algorithm is broader than one project.

11.2 Strongest bear case

The strongest bear is no longer that AI is already shrinking firm-wide hours; Q2 directly contradicted that mechanism. The surviving bear argues that EXPO’s model remains headcount × utilization × rate, annual utilization cannot escape roughly 73%, and current strength is a lumpy AI-device development cycle. A 4% study temporarily lifts utilization; Q4 seasonality restores the annual ceiling; standardized E&H work remains weak; and clients eventually use AI efficiency to reduce paid hours. At approximately 31.5x earnings, even a high-quality mid-single-digit grower can de-rate materially.

This version is stronger because it accommodates the evidence rather than denying it. It concedes Q2 was good but disputes persistence and valuation. Its empirical markers are Q3 growth below guide as the study normalizes, hours growth falling below FTE growth, annual utilization around 72%–73%, E&H weakness spreading, and multiple compression despite stable earnings.

11.3 Adjudication

Factor positioning adds a useful but non-fundamental observation. EXPO retains a negative Momentum loading (-0.47), near-zero Value loading, and only a small quantitative Quality loading despite the recent rally. Approximately three quarters of return variance remains idiosyncratic. The stock is therefore not obviously crowded in a classic momentum factor, and company results can dominate the next move. This is context, not thesis evidence.

11.4 Variant-perception verdict

INTERPRETATION. The differentiated insight in July was that AI-driven volume could falsify the rate-only bear. Q2 validated that insight. The new variant must move one step forward: reported breadth and cash conversion after the exceptional project normalize. The market has already paid for the first discovery.

12. Fact vs. Interpretation Table

13. Open Questions

1. When the large study falls from approximately 4% to 2% of net revenue, do total hours still grow faster than technical FTE?
2. How many clients, projects, and device categories constitute the AI-enabled user-research book, and how concentrated is revenue among the top three?
3. Can consolidated utilization hold 74%–75% outside Q1 without degrading consultant development or retention?
4. Does Q3 receivables collection restore FCF conversion above net income?
5. Why did billed AR grow roughly 35% year over year, and what portion relates to the exceptional reimbursable study?
6. How does management distinguish commodity data collection from differentiated experimental design in revenue and staffing terms?
7. Is E&H’s Q2 hours improvement durable after first-half hours fell 2%?
8. What is the long-run organic hiring capacity for high-quality consultants, and where does recruiting become dilutive to utilization?
9. Will the September Investor Day quantify proactive/reactive revenue and AI-related revenue rather than use qualitative examples?
10. How much of more than $100M expected H2 FCF is normal seasonality versus collection of the H1 receivables build?
11. What cash balance will management protect before using the new $50M authorization?
12. Will executive incentives add a per-share return or ROIC measure as excess cash becomes scarcer?
13. How should investors separate recurring margin expansion from deferred-comp presentation volatility?
14. Does reactive demand remain strong across consumer products, chemicals, and transportation when large proactive studies normalize?
15. Can the company offset the January 2028 Arizona rent step-up through utilization, rate, or other facility savings?

14. What Must Be True

Bull case — required evidence

• Q3 net revenue lands within or above the 8%–10% guide after the study normalizes.
• Hours continue to grow at least as fast as FTE; growth remains volume-plus-rate.
• Consolidated utilization holds around 74%–75%, with Engineering in the mid-70s.
• E&H avoids a return to persistent negative hours.
• Receivables convert, restoring full-year FCF conversion near or above net income.
• AI-enabled work broadens across clients and devices rather than depending on one anchor program.
• Hiring remains absorbed and does not recreate the 2023 utilization trough.
• Capital return slows to a cash-funded, price-sensitive cadence.

Bear case — required evidence

• The exceptional study explains most incremental hours and its normalization pulls growth below guide.
• Full-year utilization stays near 72%–73% because the annual ceiling is structural, not seasonal.
• Hours flatten while rate becomes the primary growth source over multiple quarters.
• E&H weakness spreads into Engineering or client budgets tighten across proactive categories.
• AI efficiency reduces paid hours faster than new physical-world use cases create work.
• AR remains elevated and FCF conversion fails to recover.
• The premium multiple compresses toward slower consulting peers even if earnings stay stable.

Literal score of the July tests

The pre-commitment matters. Calling the bull trigger a complete pass would move the goalposts because consolidated utilization missed 75%. Calling Q2 neutral would ignore that hours, FTE, utilization, margins, and guidance all improved. The correct score is partial pass with material positive thesis movement.

15. Source Appendix

Primary SEC and company sources

1. Exponent FY2025 Form 10-K, filed 2026-02-27: https://www.sec.gov/Archives/edgar/data/851520/000119312526082508/expo-20260102.htm.
2. Exponent Q1 FY2026 Form 10-Q, filed 2026-05-08: https://www.sec.gov/Archives/edgar/data/851520/000119312526214572/expo-20260403.htm.
3. Exponent Q2 FY2026 Form 10-Q, filed 2026-08-07: https://www.sec.gov/Archives/edgar/data/851520/000119312526340211/expo-20260703.htm.
4. Exponent Q2 Form 8-K, filed 2026-07-30: https://www.sec.gov/Archives/edgar/data/851520/000117184326005081/f8k_073026.htm.
5. Exhibit 99.1, Q2 earnings release: https://www.sec.gov/Archives/edgar/data/851520/000117184326005081/exh_991.htm.
6. Exhibit 99.2, dividend and repurchase authorization: https://www.sec.gov/Archives/edgar/data/851520/000117184326005081/exh_992.htm.
7. Company investor-relations Q2 release: https://investors.exponent.com/investors/news-events/news-details/2026/Exponent-Reports-Second-Quarter-2026-Financial-Results/.
8. Q1 2026 release for prior guidance: https://investors.exponent.com/investors/news-events/news-details/2026/Exponent-Reports-First-Quarter-2026-Financial-Results/default.aspx.
9. 2026 proxy statement, filed 2026-04-20: https://www.sec.gov/Archives/edgar/data/851520/000119312526163945/expo-20260416.htm.
10. CEO Form 4, filed 2026-07-17: https://www.sec.gov/Archives/edgar/data/851520/000085152026000092/edgardoc.xml.
11. CEO Form 4, filed 2026-07-30: https://www.sec.gov/Archives/edgar/data/851520/000085152026000094/edgardoc.xml.
12. CEO Form 4, filed 2026-08-18: https://www.sec.gov/Archives/edgar/data/851520/000085152026000096/edgardoc.xml.
13. Representative S-8 POS administrative deregistration, filed 2026-08-11: https://www.sec.gov/Archives/edgar/data/851520/000119312526344516/expo-s-8_pos_333_67806.htm.

Transcripts, peer sources, and market data

14. Q2 FY2026 earnings-call transcript, call held 2026-07-30; secondary transcript of management statements: https://www.fool.com/earnings/call-transcripts/2026/08/07/exponent-expo-q2-2026-earnings-call-transcript/.
15. CRA International Q2 FY2026 results, 2026-08-06: https://ir.crai.com/news-releases/news-release-details/charles-river-associates-cra-reports-financial-results-second-6.
16. FTI Consulting Q2 2026 results, 2026-07-30: https://ir.fticonsulting.com/news-releases/news-release-details/fti-consulting-reports-second-quarter-2026-financial-results/.
17. AZI adjusted daily price data, pulled 2026-08-23: https://azitrading.com/controls/download-data.php?t=EXPO.
18. AZI valuation index, generated 2026-08-23 through the repository integration; latest market date 2026-08-21.
19. FactorsToday factor and risk data, pulled 2026-08-23: https://www.factorstoday.com/api.
20. ROIC.ai EXPO financial cross-check, pulled 2026-08-23: https://www.roic.ai/quote/EXPO.

APPENDIX A — Standard Diligence Questionnaire

General

What does the company do? Exponent supplies independent engineering and scientific consulting for failures, disputes, safety, product development, human factors, reliability, and regulatory questions. It sells expert time and multidisciplinary project execution, not a software subscription.

How does it make money? Net revenue is hours × realized billing rate, influenced by technical FTE and utilization. Reimbursements are pass-through. Profitability rises materially when utilization improves because much of compensation and facility cost is fixed or semi-variable within a quarter.

Who pays? Corporations, law firms, insurers, government agencies, and product developers. The work spans consumer products, utilities/energy, transportation, chemicals, life sciences, and construction.

Is revenue recurring? Not contractually. Recurrence comes from repeat clients, continuous failures and disputes, regulation, and successive product-development programs. That is economically durable but project-lumpy.

What is the core KPI? Billable-hours growth relative to FTE growth, followed by utilization and rate. Revenue without hours can be rate-only; gross revenue can be distorted by reimbursements.

Cyclicality and Earnings Nature

How cyclical is it? Low-to-moderate overall. Reactive work is less discretionary and can be counter-cyclical; proactive work follows R&D, product, infrastructure, and regulatory budgets. Client product cycles can create material quarterly swings.

Where are earnings in the cycle? Recovery, not proven peak. Q3 2025, Q1 2026, and Q2 2026 establish a multi-print rebound. Full-year utilization guidance near 73% and the Q2 large-study contribution argue against calling it a new high-utilization regime.

What would a recession do? It could slow proactive product-development work and hiring. Reactive disputes, failures, recalls, and insurance matters provide partial defense. A broad client-budget freeze would still pressure utilization.

What caused the 2023–2025 slowdown? Over-hiring lowered utilization, followed by proactive consumer-electronics weakness and project timing. Management reduced FTE in 2024, and volume subsequently recovered.

Business Quality and Competitive Moat

What is the moat? A reputation intangible accumulated over nearly six decades, reinforced by multidisciplinary scale, independence, laboratories, recruiting reputation, and client selection risk in high-stakes matters.

Is there a network effect? No. More clients do not mechanically increase value to every other client. Knowledge accumulation and brand are scale/intangible advantages.

Are there switching costs? Selection-based rather than contractual. Clients can switch between matters, but choosing an untested expert or fragmented team creates litigation, regulatory, and execution risk.

What proves pricing power? Q2 hours rose 9% while realized rate rose approximately 4%; price and volume advanced together. FY2025 also grew revenue despite fewer hours, demonstrating the ability to raise rate during softness.

What could destroy the moat? A major independence, credibility, or testimony failure; inability to retain elite consultants; commoditization of differentiated work; or a cultural shift toward sales over objective science.

How attractive is the industry? Structurally mediocre in the broad market because talent is mobile and entry at a single discipline is easy. The premium, high-stakes niche is more attractive because credibility and integrated depth matter.

Growth and AI

What are the main growth drivers? Technical FTE growth, utilization recovery, annual price, AI-enabled physical products, utility and digital infrastructure, complex regulation, recalls, disputes, and adjacent medical-device work.

Is AI a threat or benefit? Both, but benefit is visible today. It creates product/user research, data-center reliability, ADAS, wearables, and human-factor work. It can automate standardized surveys, review, and dossier preparation or reduce hours per engagement.

What did Q2 prove? AI-related demand coincided with company hours +9%, FTE +6%, and margin expansion. It did not prove multi-year breadth, because one study was 4% of quarterly net revenue.

What is the clean next test? Q3 performance after the study normalizes to approximately 2% of revenue: 8%–10% growth, 74%–75% utilization, and hours growth at least matching FTE.

Financial Condition and Balance Sheet

Is the balance sheet safe? Yes. Cash was $66.6M, current ratio about 2.0x, and funded debt zero. Operating leases are obligations but not funded borrowing.

Why did cash decline? First-half repurchases of $146.1M and dividends of $31.3M exceeded $25.7M of FCF, intentionally using excess cash.

Is earnings quality high? Historically yes; FCF exceeded net income in FY2023–FY2025. TTM conversion fell to approximately 97% because receivables absorbed cash. Q3 collections are the verification point.

What accounting items require care? Reimbursements, deferred-comp expense/investment offsets, stock-comp tax effects, SBC, lease straight-lining, and changes in interest income as cash falls.

Is ROIC real? Yes directionally. Ten-year ROIC averages around 20%, far above cost of capital. Exact values vary with cash, lease, and deferred-comp treatment.

Capital Allocation and Management

Was the buyback well timed? Q2 was: 1.125M shares at $59.88, concentrated near the lows. First-half average was approximately $64.05, better than FY2025’s approximately $72.

Can that pace continue? Not safely from current cash alone. Management expects H2 FCF above $100M and says future buybacks should be slower and pullback-sensitive.

Is the dividend safe? It appears well covered on normalized full-year earnings and cash flow, with 13 years of increases. No dividend is guaranteed.

Are incentives aligned? Profit-based bonuses and deferred equity help. The absence of explicit per-share/ROIC targets and rising SBC are shortcomings.

What do insiders signal? No open-market purchases, but no discretionary post-July sales either. The CEO exercised and held 12,665 shares outside a plan; direct ownership increased.

Valuation and Market Data

What is the current absolute valuation? Approximately 31.5x filing-derived TTM EPS, 30.7x TTM FCF, 20.5x EV/EBITDA, and 5.77x net revenue.

What does own history say? AZI’s composite is around the 24th percentile. The apparent cheapness is relative to an unusually expensive history; it does not make the absolute multiple low.

What expectations are embedded? A satisfactory return likely needs upper-single-digit per-share growth with a premium terminal multiple. Mid-single-digit growth plus normal multiple compression produces weak arithmetic.

What changed since July? Business evidence strengthened and price rose 15.2%. The valuation now capitalizes more of the recovery, while excess cash per share declined.

Risks and Downside

What is the most likely downside mechanism? Large-study normalization or utilization disappointment causes a multiple de-rate, even if the moat remains intact.

What is the most important operational risk? Hours stop growing while FTE and compensation continue, recreating utilization pressure.

What is the newest financial risk? Receivables fail to convert and buybacks leave too little cash flexibility.

Is permanent capital loss likely? Solvency-driven permanent loss is unlikely. Valuation-driven loss can be material, as the February–May decline demonstrated.

Recent News and Events

What was material after July 8? The July 30 Q2 release, raised guidance, dividend, and $50M authorization; the August 7 10-Q details; routine insider filings; and administrative S-8 plan cleanup.

What was not material? The seven S-8 POS entries do not represent seven issuance events. No other post-Q2 operating release was identified through August 23.

What is next? September 29 Investor Day, Q3 earnings, Q3 cash conversion, and completion of the Q2–Q4 utilization test.

APPENDIX B — Evidence Ledger

Research completeness. The report uses filings through August 18, market data through August 21, and a report date of August 23, 2026. No PDF was generated. No recommendation or directional valuation zone appears outside Claude’s Take.