> [Portrait Research index](https://portraitresearch.com/llms.txt) [Canonical page](https://portraitresearch.com/mosaic/) Mosaic # Build your shortlist on deeper research. Investigate companies against your investment criteria. Research agents build the evidence behind a ranked shortlist you can review in depth. [Book a demo](https://portraitresearch.com/demo/) [See how it works](https://portraitresearch.com/mosaic/#mosaic-how-it-works) The Mosaic creation page inside Portrait. Example research requests, not completed results. US-listed software companies above $2B with gross margins over 65% — keep the ones whose filings and earnings calls show net revenue retention above 110%, then rank the survivors by how durable that growth looks. ## From investment criteria to company-level analysis. 1. ### Define what a company must demonstrate. Choose your universe and financial filters. Specify what Mosaic should investigate, what evidence would qualify or rule out a company, and which priorities should guide the comparison. #### Research section v4 The research task every agent was given for this screen, including any pass/fail gates authored into it. ##### US Industrial-Capex Compounders — framework v4 US specialty industrial / A&D / specialty-materials companies at $10-100B market cap that are direct beneficiaries of AI datacenter, semiconductor fab, US reshoring, defense industrial base, and IIJA/grid capex mega-cycles. Modeled on EMCOR (EME), Comfort Systems (FIX), Sterling Infrastructure (STRL), and Carpenter Technology (CRS). Qualitative context Screen aims to identify US specialty industrial / aerospace-defense / specialty-materials companies that are direct beneficiaries of the current wave of US structural capex — AI datacenter buildout, semiconductor fab construction, reshoring, defense industrial base rebuild, and IIJA/grid infrastructure. Modeled on EMCOR (EME), Comfort Systems USA (FIX), Sterling Infrastructure (STRL), and Carpenter Technology (CRS). This framework runs as FIVE research tasks, no separate filter section. Tasks 1-4 each demand a PASS/FAIL verdict with cited evidence from the company's own filings (10-K/10-Q/earnings-call/investor presentation); a company must PASS all four to qualify. Task 5 is the peer-valuation construction, independent of pass/fail. Ground every judgment in the Portrait document corpus; do not infer end-market exposure from industry classification alone. View quantitative filters #### Quant filters v1 The deterministic numeric, geography, and industry filters that decided which companies were eligible. ##### Quantitative filters — v1 1. **United States** — geographies: US 2. Specialty industrial / A&D / materials tag set 3. Market cap $10-100B 4. **Revenue YoY > 8%** — total\_revenue\_yoy\_LTM > 8 5. **Operating margin > 5%** — operating\_margin\_LTM > 5 6. **Positive FCF** — fcf\_LTM > 0 7. **Net debt / EBITDA < 3x** — net\_debt\_ebitda\_ttm\_LTM < 3 Saved US Industrial-Capex Compounders case: current research v4 instructions excerpt. 2. ### Review the research before you scale. Start with a small sample of companies. Examine the research, check how your criteria are applied and assess the cost before a larger run. Adjust the instructions where needed. #### Trial Review a small sample before the full run. Inspect the research, then refine your framework or ranking criteria with Easel. Full-run estimate: waiting on the trial's final cost. **HEI.US** HEICO Corporation Passed summary.md ##### HEICO Corporation — framework v4 screening summary **Overall:** PASS on Tasks 1–4. HEICO is a US aerospace/defense and electronics compounder with direct, material exposure to the defense-industrial-base rebuild, strong specialty differentiation, and growing contractual visibility. Read the full assessment Its principal contrary point is that this is not a capital-intensive infrastructure contractor: FY2025 capex was only $72.9 million on $4,485.0 million of sales (about 1.6%), so the thesis fit is through mission-critical defense/space components and aftermarket supply rather than direct ownership of datacenter, fab, or grid construction assets. HEI 10-K FY 2025 ###### 2\. Specialty differentiation — PASS HEICO’s differentiation is unusually well evidenced. Its FSG uses proprietary technology to design FAA-approved replacement parts, has approximately 20,000 FAA PMA part approvals, and adds approximately 400–550 PMAs annually. The filing says the FAA approval process creates a significant technical and regulatory barrier to entry. FSG also performs proprietary DER repairs, while ETG designs mission-critical, high-reliability electronic, microwave, power, sensing, and defense/space components for harsh environments. FY2025 R&D expense was $120.9 million, supporting an engineering- and qualification-led rather than commodity model. HEI 10-K FY 2025 ###### 3\. Backlog / forward visibility — PASS At April 30, 2026, HEICO reported $2,622.6 million of remaining performance obligations associated with firm contracts, of which $1,177.0 million was expected to be recognized in the remainder of FY2026 and $1,445.6 million thereafter (the majority in FY2027). This was up from $2,107.6 million at October 31, 2025, a 24.4% increase in six months, and up 36.3% from October 31, 2024. The balance therefore provides substantially more than six months of forward visibility. HEI 10-Q Q2 2026 HEICO dossier from the saved five-company trial. Retained research can include work added after the trial. The cost estimate is still pending. 3. ### Investigate your universe, company by company. Mosaic researches the candidates that meet your financial criteria using the approach you’ve reviewed. Detailed company dossiers bring together the findings and supporting analysis, including where a business meets your requirements or falls short. #### Company results The outcome the agents recorded for each company. 26/59 passed screens - **AEIS.US** Advanced Energy Industries, Inc. Not passed - **AA.US** Alcoa Corporation Not passed - **ALLE.US** Allegion plc Not passed - **ALSN.US** Allison Transmission Holdings, Inc. Not passed - **AU.US** AngloGold Ashanti plc Not passed - **APG.US** APi Group Corporation Passed - **ARXS.US** Arxis, Inc. Passed First seven company results from the saved full research run. These are example outcomes, not a product-wide success rate. 4. ### Find the strongest matches—and understand why. Mosaic ranks the researched candidates against your investment priorities. Compare the results alongside the company dossiers to understand what distinguishes the candidates and decide which names deserve further diligence. #### Ranked companies (26) Ordered by the ELO score the judge comparisons produced. | Rank | Company | ELO | | --- | --- | --- | | 1 | **STRL.US** Sterling Infrastructure, Inc. | 1,355.5 | | 2 | **MTZ.US** MasTec, Inc. | 1,347.9 | | 3 | **NVT.US** nVent Electric plc | 1,346.2 | | 4 | **FIX.US** Comfort Systems USA, Inc. | 1,341.9 | | 5 | **DY.US** Dycom Industries, Inc. | 1,324.4 | | 6 | **PWR.US** Quanta Services, Inc. | 1,276.5 | Read company research STRL.US · Company research #### summary.md ##### Sterling Infrastructure, Inc. (STRL) Screening conclusion: QUALIFIES Sterling passes all four qualitative framework gates. Its strongest fit is direct exposure to AI/data-center and semiconductor/manufacturing construction through E-Infrastructure, supported by rapidly expanding backlog and a differentiated mission-critical execution niche. The principal caution is that data-center revenue is not separately reported and the business retains fixed-price contract, customer-concentration, and project-cancellation risks. ###### 1\. Capex-megacycle exposure — PASS Sterling’s FY2025 10-K says E-Infrastructure serves “data centers, semiconductor fabrication, manufacturing, distribution centers, warehousing, power generation and more” (STRL 10-K FY 2025). First six of 26 ranked companies in the saved case. ELO expresses relative fit to the rubric, not expected investment returns. ## Frequently asked questions. How is Mosaic different from Screening? Screening applies explicit quantitative and qualitative filters and returns company-level results. Mosaic applies a reusable research brief to each candidate, retains the company dossiers and, for screening frameworks, can rank the researched companies against your criteria. Use Mosaic when the condition itself needs investigation before you decide where to focus. [Compare research workflows](https://portraitresearch.com/documentation/reference/compare-research-tools/) What kinds of criteria can Mosaic investigate? Define questions that require company-level evidence—for example, the durability of revenue, customer dependence, sources of margin improvement or whether exposure to a theme is direct and material. State what would qualify or rule out a company and what evidence should be examined. Coverage varies by company and period, and source access depends on your firm’s entitlements. Can I use my own company universe? Yes. Start with a defined company list or use quantitative filters to form the universe. Mosaic runs Quant before Research to resolve which companies are eligible. Coverage and available documents vary; an excluded company or missing result is not an investment verdict. What can I review before a full run? Start with a paid Research trial of up to 20 companies. Inspect the dossiers, outcomes and measured usage; for screening work, you can also rank the trial. A completed trial under the current Research instructions is required before full Research, and changing those instructions requires a new trial. The 20-company limit applies to the trial, not the full research universe. Estimates are not fixed prices. [Review the Mosaic workflow](https://portraitresearch.com/documentation/mosaic/research-companies/) How are companies ranked? For screening frameworks, Mosaic compares researched companies against the ranking criteria you define. The order and Elo scores express relative fit to that rubric—not probabilities, valuations, expected returns or an objective recommendation. Research-only tasks do not rank companies. How do access and pricing work? Mosaic is available in the web app for enabled accounts with Easel access, relevant data entitlements and a separate Mosaic credit balance. Research trials and ranking consume Mosaic credits; the trial informs an estimate for a larger run, not a fixed price. Book a demo to discuss access, your research brief and expected scope. [Read about usage and estimates](https://portraitresearch.com/documentation/mosaic/usage/)