A 25-, 100-, or 450-location platform that centralizes too much breaks the local equity that customers buy from. A platform that centralizes too little burns the margin advantage that justified the roll-up. The 2026 decision rule for trades platforms — Wrench Group, Authority Brands, Apex Service Partners, Sila Services, Path Light Pro, Redwood Services, and ARS-Rescue Rooter — has converged on a specific architecture: phone systems and Avoca-style call answering centralize at the platform; tech scorecards and dispatch logic centralize; AI tooling spend and vendor relationships centralize; the prompt libraries and the QA layer centralize. Brand voice, local SEO, named-tech customer relationships, CSR scripts at the moment of customer contact, financing tier offers tied to local market dynamics, and the customer-facing IVR script stay local. This lesson is the decision rules — when to centralize, when to keep local, and the operating cost of getting it wrong in either direction. The board-defendable version of the platform/local architecture. The version a PE board reads on slide 5 of the quarterly review next to the EBITDA waterfall.

Why the Centralize-vs-Local Decision Defines Platform Margin

Centralization buys procurement leverage. Per-seat AI tool discounts of 30-50% versus independent retail. Shared technical integration plumbing built once at HQ and reused across 40-280 locations. Centralized prompt libraries maintained by a Prompt Librarian role rather than reinvented at each site. Cross-portfolio benchmarking on common metrics that surface the bottom-quartile sites for targeted remediation. Centralized vendor governance that holds Avoca, Rilla, Hatch, and ServiceTitan accountable to platform-level deployment success metrics rather than site-by-site contract management. The procurement and operating leverage of centralization is what makes the platform's EBITDA-per-location target achievable.

Localization protects revenue. The customer who calls "Smith Plumbing" — a brand that has served three generations in their metro — answers the phone to "Smith Plumbing." The CSR script references the neighborhood, the local landmark, the seasonal weather pattern. The technician who arrives wears the Smith Plumbing logo, drives a Smith Plumbing truck. The review response references the specific Smith Plumbing tech the customer met. Strip too much local equity in pursuit of centralization, and the customer experience reads as a national chain that bought the local shop — and customer trust drops, review velocity slows, and the local market's preference for the established brand erodes within 12-24 months.

The 2026 platforms that get this right run a decision matrix at the workflow level. Each AI workflow, each customer-facing surface, each operational layer gets assessed: does centralization produce more value than the local equity it costs, or does the local equity outweigh the centralization gain? The decision is not "more central" or "more local" as a default; it is workflow-specific. The platforms that run the decision matrix workflow-by-workflow produce 2-4 points of EBITDA margin over the platforms that default to either extreme.

What Centralizes — Phone Systems and Voice AI

The phone system centralizes. A platform's inbound call routing, IVR architecture, missed-call recovery, after-hours overflow, and voice AI deployment all run from a central infrastructure regardless of which brand the call originates against. Avoca deployed at the platform level handles inbound across all locations through a unified Avoca tenant; CallRail integration plumbs to the platform's call tracking infrastructure; ServiceTitan or HCP integration plumbs to each location's instance. The customer hears the right brand answering the right line — the routing is brand-aware — but the underlying voice AI, call recording, transcription, sentiment analysis, and missed-call recovery operate as platform infrastructure.

The reason is integration economics. Avoca deployed independently at 100 locations is 100 separate vendor relationships, 100 separate integrations into 100 separate ServiceTitan instances (or worse, mixed instances of ServiceTitan, HCP, Sera, and Jobber), 100 separate IVR configurations to maintain, and 100 separate weekly QA reviews. Avoca deployed at the platform level is one vendor relationship, one integration plumbed once into a unified call routing architecture, one set of platform-level QA review cadence, and one prompt library that adapts brand-by-brand without reinventing the core. The cost differential at 100 locations is $40K-$120K per year in deployment, support, and QA overhead alone — before procurement discount.

The same logic applies to ServiceTitan Voice, Jobber AI Receptionist, Housecall Pro AI Agents, and any other voice AI deployment the platform standardizes on. The customer-facing surface (IVR greeting, hold music, brand-specific phrases) is brand-local; the AI infrastructure is platform-central. The platform's IVR routing logic reads the inbound line, identifies the brand, and applies the correct brand-specific greeting and call-flow before handing off to the AI. The brand experience preserves; the operating leverage centralizes.

What Centralizes — Tech Scorecards and Cross-Location Benchmarks

Tech scorecards centralize. A platform tracks the same set of metrics for every technician across every brand across every location: revenue per visit, average ticket, repair-vs-replace rate, MPR enrollment rate, financing close rate on eligible jobs, recall rate, customer review velocity per tech. The scorecard format is standardized; the metric definitions are standardized; the dashboard view is standardized. The Wrench Group tech in Phoenix, the Apex Service Partners tech in Tampa, and the Authority Brands franchisee's tech in Cleveland all show up on a comparable dashboard.

The reason is benchmarking. Without standardized tech scorecards, the platform cannot identify the top-quartile tech (whose practices should be studied and replicated) or the bottom-quartile tech (who needs targeted coaching). With standardized scorecards, the platform's training and development function can identify the techs whose MPR enrollment is 60%+ and capture their pitch sequence; can identify the techs whose recall rate is 7%+ and run the repair-vs-replace coaching cycle. Cross-location benchmarking is the platform's coaching engine — the engine that lifts median performance toward top-quartile performance across hundreds of techs.

Tech scorecards also standardize the comp-and-incentive structure. Platform-level comp programs tie to scorecard metrics; high-performing techs across the portfolio are recruited into platform-wide accelerator tiers, cross-location mobility programs, or trainer-track roles. The scorecard becomes the platform's talent operating system. Without standardization, talent identification stays local, the platform's compounding advantage on tech development doesn't accumulate, and the median wage cost rises as good techs leak to competitors who offer career paths.

What Centralizes — AI Tooling Spend, Vendor Governance, and the Prompt Library

AI tooling spend centralizes. The platform negotiates master agreements with Avoca, Rilla, Hatch, ServiceTitan, Podium, Birdeye, NiceJob, and Ryze AI on behalf of the entire portfolio. Per-seat pricing reflects volume; deployment success clauses reflect framework discipline; data pooling agreements (and the governance discipline that comes with them — covered in L5 Ch3) reflect platform-level data architecture. Locations do not negotiate individual contracts; they consume the platform's vendor agreements. Procurement leverage compounds quarterly as the platform's footprint grows and vendor competition for the platform's business intensifies.

Vendor governance centralizes. A platform-level Director of AI Operations (the role L5 Ch4 covers) owns the vendor relationships, the quarterly business reviews with each vendor, the contract renewals, the deployment success measurement, and the rollback playbook ownership. The vendor's customer success manager works through the platform's central governance rather than calling individual GMs. The vendor's roadmap commitments tie to the platform's stage-gate cadence (the framework from Lesson 1). Vendor-level escalations route through the platform; site-level operating exceptions route through regional directors.

The prompt library centralizes. A platform-level Prompt Librarian (the role from L5 Ch4) maintains the master prompts for each AI workflow — the CallRail Conversation Intelligence prompts, the Hatch nurture sequence prompts, the GLSA AI bidding negative-lead-dispute prompts, the AEO publishing prompts, the review response drafting prompts, the Rilla coaching summary prompts. Each prompt is versioned, tested, and tied to the workflow's measured lift. Brand-specific overlays sit on top of the master prompts — Mister Sparky's electrical voice differs from Benjamin Franklin's plumbing voice; the overlay encodes the difference without forking the master. Centralized prompts produce centralized lift; brand-local overlays produce brand-specific customer experience.

What Stays Local — Brand Voice, CSR Scripts at the Customer Moment, and the Named Tech Relationship

Brand voice stays local. Every customer-facing surface — the IVR greeting, the post-job follow-up text, the review response, the AEO-published service-area page, the email drip campaign — reads in the brand's voice. Smith Plumbing reads like Smith Plumbing; Mister Sparky reads like Mister Sparky; Benjamin Franklin reads like Benjamin Franklin. The brand voice is the customer's relationship with the brand; centralized voice flattens to a generic corporate tone that breaks trust within a quarter.

The mechanism is the brand-specific prompt overlay. The master prompt for review response drafting at the platform level produces a generic professional response. The brand-specific overlay adds the brand's tone (warm and conversational for Smith Plumbing, technical and confident for Mister Sparky, premium and white-glove for a Wrench Group commercial brand), the brand's signature phrases ("we treat your home like our home" or "100% satisfaction or you don't pay"), and the brand's specific tech-named or service-named references. The customer reads the brand-specific response; the platform's QA reviews the brand-specific output; the master prompt evolves at the platform level without breaking brand voice.

CSR scripts at the moment of customer contact stay local. The platform standardizes CSR training, scorecard metrics, and the overall CSR floor management. The script the CSR reads when the customer says "my heat went out and the baby is crying in the background" is brand-and-location-local: the CSR references the local dispatcher, the local time-window expectations, the brand's specific service-area and after-hours coverage, the local financing options. Centralized scripts at the moment of customer contact produce robotic interactions; local scripts produce the trust-building moments that convert calls to bookings.

The named-tech customer relationship stays local. When a homeowner calls back six months after a service visit and asks for "Jake who fixed my furnace last spring," the platform's response is "Jake is on our team — let me check his schedule." Jake is the local tech, with the local brand truck, working out of the local shop. The centralized platform infrastructure (dispatch, scheduling, scorecards, voice AI) supports Jake's relationship with the customer; it does not replace it. The named-tech relationship is the highest-value customer-acquisition engine in trades; platforms that centralize away from named-tech relationships sacrifice the moat.

What Stays Local — Local SEO, Financing Tier Offers, and Named Service-Area Equity

Local SEO stays local. Each location's Google Business Profile, named service-area pages, local citation density, neighborhood-specific FAQ schema, and AEO publishing cadence are managed at the local level (or the regional level for clusters of nearby locations). The reason is mechanical: Google's local algorithm rewards location-specific signals — local phone number, local address, local citations, local-tech-named reviews, neighborhood-specific content. Centralizing local SEO into a national template flattens these signals and the platform loses local pack rankings to independent competitors with disciplined local SEO.

The platform's role in local SEO is governance and infrastructure, not execution. Platform provides the AEO publishing framework, the structured data templates, the FAQ schema library, the citation-building partnership network (industry trade press, local Better Business Bureau, local trade association membership). Each location applies the framework to its specific market with its specific neighborhood references and named techs. The platform's quarterly review measures local SEO performance across locations and surfaces the bottom-quartile sites for targeted support, but execution stays local because the algorithm rewards local execution.

Financing tier offers stay local because local market dynamics shape what works. A Pacific Northwest market with 18% average household income above the metro median supports premium financing tiers (high credit, low APR, longer terms); a Phoenix outer-suburb market with 28% household income below metro median needs deep-credit financing options (Wisetack-style fast-service or Synchrony-style branded revolving). A platform that mandates a single financing tier across all locations loses the close-rate on the tier mismatch. Centralizing the financing vendor relationships (Wisetack, GreenSky, Synchrony master agreements) is platform leverage; mandating the tier offer at the customer-moment is platform mistake.

Named service-area equity stays local because the brand-in-neighborhood relationship is a 5-30 year accumulation that doesn't transfer. Smith Plumbing's reputation in the Walnut Creek market is built over 35 years of techs, reviews, neighborhood referrals, BBB ratings, and Yelp accumulation. If the platform rebrands Smith Plumbing to a national brand, the local equity drains quarterly — Yelp reviews and Google rankings on the old brand fade as the new brand catches up at a fraction of the velocity. The roll-up AEO problem (Lesson 5 in this chapter covers this in depth) is the platform-level version of this trade-off: rebrand for centralized citation gain or preserve local equity. The decision is workflow-specific and market-specific, not a default.

The Decision Rules and the Platform Architecture Matrix

The 2026 platform architecture matrix captures the workflow-by-workflow decision. Each AI workflow, each customer-facing surface, each operational layer is plotted on two axes: customer-perceived value of localization, and operational economics of centralization. Workflows where customer-perceived value of localization is high and centralization economics are modest stay local (brand voice, CSR scripts at customer moment, named tech relationships, local SEO, financing tier offers, named service-area equity). Workflows where centralization economics dominate and customer perception is unaffected centralize (phone system infrastructure, voice AI core, tech scorecards, AI tooling spend and vendor governance, prompt library architecture, cross-location benchmarking). Workflows where both axes matter get hybrid architectures with platform-level infrastructure and brand-local overlays (review response drafting, AEO publishing framework, financing vendor relationships with local tier mandate, IVR routing with brand-aware greeting).

The hybrid architectures are where most operating sophistication concentrates. Review response drafting: platform owns the master prompt and the QA cadence; brand owns the voice overlay and the local-tech references; location owns the 60-second skim before posting. AEO publishing: platform owns the structured data templates and the FAQ schema library; brand owns the voice; location owns the neighborhood-specific content and the named-tech features. Financing vendor relationships: platform owns the master agreement with Wisetack/GreenSky/Synchrony; location owns the tier offer at the customer moment. IVR routing: platform owns the routing infrastructure and the AI deployment; brand owns the greeting script and the call-flow tone; location owns the after-hours coverage policy and the local-emergency dispatch rules.

The discipline that protects the matrix is governance. Without explicit governance, centralization creep happens — the platform adds one more standardization layer per quarter, each of which seems small, until the brand voice has drained and the named-tech relationship has commoditized. The Director of AI Operations role owns the matrix and the quarterly review of every centralization decision against the customer-perceived localization value. The role's job is not to maximize centralization; it is to optimize the matrix workflow-by-workflow.

How the Eight Platforms Operate This in 2026

Wrench Group runs the matrix across its HVAC and plumbing portfolio with ServiceTitan as platform standard, Avoca as centralized voice AI infrastructure, tech scorecards, AI tooling spend, and prompt library centralized. Brand voice, local SEO, financing tier offers, named-tech relationships, and CSR scripts at customer moment stay brand-local.

Authority Brands runs the matrix across One Hour, Benjamin Franklin, and Mister Sparky franchise networks. The complication is franchise autonomy. The platform centralizes what it has operational authority over (franchise-system master vendor agreements, franchise-standard tech stack, national MAP spend). Franchisees opt in voluntarily on most centralization (most opt into Avoca and tech scorecards because lift is documented; many keep CSR scripts and financing tiers local). The override request memo from Lesson 3 governs deviations.

Apex Service Partners runs the matrix across acquired brands with ServiceTitan + Rilla + CallRail as standard stack. Acquired brands retain name and voice (preserving local equity) while operating infrastructure centralizes within 90 days of close. The post-acquisition integration plan explicitly maps centralize-vs-local for each operational layer.

Sila Services, Path Light Pro, Redwood Services, and ARS-Rescue Rooter each run portfolio-specific variations. Sila's owned-location structure simplifies the matrix; Path Light Pro's electrical specialization shifts financing tier dynamics; Redwood and ARS run residential matrices similar to Wrench. Framework is consistent; calibrations differ.

For the 5-15 location multi-shop independent, the matrix still applies. Phone system and voice AI centralize at operator level (one Avoca tenant). Tech scorecards centralize. AI tooling spend negotiates at operator level. Brand voice and CSR scripts at customer moment stay local. Local SEO stays per-location. Architecture discipline same; scale different.

Key Takeaways

• The platform/local architecture decision is workflow-by-workflow, not a default. The 2-axis matrix: customer-perceived value of localization vs. operational economics of centralization. Plot each workflow; decide independently.
• Centralizes: phone system infrastructure and voice AI deployment (Avoca, ServiceTitan Voice, Jobber AI Receptionist), tech scorecards and cross-location benchmarking, AI tooling spend and vendor governance, prompt library architecture, dispatch logic, master vendor agreements. The procurement and operating leverage is what makes platform EBITDA-per-location achievable.
• Stays local: brand voice, CSR scripts at the customer moment, named-tech customer relationships, local SEO and AEO execution, financing tier offers tied to local market income dynamics, named service-area equity (the 5-30 year brand-in-neighborhood accumulation).
• Hybrid architectures are where most sophistication concentrates: review response drafting (platform master prompt + brand voice overlay + location 60-second skim), AEO publishing (platform framework + brand voice + location neighborhood content), financing vendor relationships (platform master agreement + local tier mandate), IVR routing (platform infrastructure + brand greeting + local after-hours coverage).
• The brand-specific prompt overlay is the mechanism that preserves brand voice while centralizing prompt architecture. Master prompt at the platform level produces generic professional output; brand-specific overlay adds tone, signature phrases, and tech-named references. Customer reads brand-specific; platform governs master.
• Tech scorecards centralize because cross-location benchmarking is the platform's coaching and talent operating system. Top-quartile techs get studied and replicated; bottom-quartile techs get targeted coaching. Standardized scorecards enable platform-wide comp programs and career paths that retain talent.
• Local SEO stays local because Google's local algorithm rewards location-specific signals. Centralizing flattens the signals and the platform loses local pack rankings to disciplined independent competitors. Platform provides framework and templates; location applies them.
• The named-tech relationship is the highest-value customer-acquisition engine in trades. Platforms that centralize away from named-tech relationships sacrifice the moat. The centralized infrastructure supports the named tech; it does not replace the relationship.
• Governance is the discipline that protects the matrix. Without explicit governance, centralization creep accumulates — the Director of AI Operations role owns the matrix and reviews every centralization decision quarterly against customer-perceived localization value. The role's job is matrix optimization, not centralization maximization.
• The eight platforms run the same matrix with portfolio-specific variations: Wrench (owned HVAC/plumbing), Authority Brands (franchise with autonomy constraints), Apex Service Partners (acquired brands with name preservation), Sila Services (owned Northeast HVAC), Path Light Pro (commercial electrical), Redwood Services, Leap Partners, ARS-Rescue Rooter. The framework is consistent; the calibrations differ.