The free Florida bid channels that beat ConstructConnect.

The bottleneck.

A small civil or structural firm in Florida, one to five people doing engineering services rather than trade subcontracts, tries to set up a bid pipeline. The first vendor name that surfaces is ConstructConnect. Its Project Intelligence product advertises a starting price of $129 per month (Starter) to $199 per month (Professional) per market when billed annually, then goes quote-based above that; users report real quotes from roughly $2,150 per year up to $25,000 per year depending on seats, market coverage, and modules (pricing as of June 2026). The entry price is lower than a sales call makes it feel, but the all-in cost for useful coverage climbs fast and stays opaque. ConstructConnect's 2026 materials cite a project database of more than one million public and private projects across 400-plus markets. The marketing implies that subscribing is the obvious first move.

For a subcontractor hunting bid invitations on a defined trade scope (concrete, drywall, MEP, structural steel), the implication is roughly correct. ConstructConnect is built for subs and for general contractors sending subs invitations on MasterFormat-coded packages. For an engineering consultancy selling design, forensic investigation, inspection, or AI integration work, the implication is wrong. The sales motion is different. The filtering taxonomy does not match. The price is not the issue. The fit is.

The same firm has an underused alternative directly available: the free public-procurement channels in Florida, designed for the buyers (state agencies, counties, water management districts, federal civilian agencies) who actually advertise engineering scope. Subscribing costs zero. The output is delivered as email, which is the right shape for an autonomous ingest pipeline. The matchup is not subtle once it is laid out side by side.

What ConstructConnect is built for, and where the mismatch shows up.

ConstructConnect's public marketing positions the product around three claims: more than a million projects, MasterFormat trade-scope filtering, and integrations with Salesforce and SmartBid. The primary user is a subcontractor estimator who needs to know which projects in a region are advertising bid packages in the estimator's specific trade. The mismatch for an engineering consultancy shows up in four places.

• The audience. Engineering consultancies sit on the design side, not the subcontractor side. The buyer who advertises a structural design RFP or a bridge inspection RFQ is an agency procurement officer, not a general contractor chasing trade coverage. The channels those buyers use are public-procurement portals, not ConstructConnect.
• The taxonomy. MasterFormat is a construction-trade taxonomy. A "bridge structural design" scope or a "post-storm forensic investigation" scope does not map cleanly to a MasterFormat division. The filters work for trade subcontracts and degrade for professional-services scope.
• The export model. ConstructConnect offers Excel export and integrations with Salesforce and SmartBid. There is no public REST API. An autonomous filter has to ingest through email digests or scrape the Bid Center dashboard. The email-digest path works, but only as well as ConstructConnect's chosen email format on a given week.
• The noise floor. More than a million projects is a lot of noise to filter for a three-person Florida-focused firm. The price would be defensible if the alternative were nothing. The alternative is not nothing.

Where ConstructConnect would help: large multi-trade vertical projects where the firm wants to be the engineer of record on a scope an architect is putting out for sub-bidding, or large structural rehab packages where the firm could be a sub-consultant to a general contractor. A real but narrow slice. Not worth a quote-based subscription in Phase 1. Dodge sits in adjacent territory with more planning-phase intelligence (advance notice months before bids advertise), which fits selling design services better than ConstructConnect does; it is also paid and custom-quoted, worth revisiting in Phase 3 against measured pipeline data, not before. PlanHub is even more subcontractor-shaped than ConstructConnect. Skip.

The free Florida channels worth the time.

Seven account registrations cover the entire useful inventory for a Florida engineering consultancy.

• Florida Vendor Information Portal (VIP) at vendor.myfloridamarketplace.com. Replaces the legacy VBS. One subscribe path catches every FDEP Bureau of Design and Construction solicitation (the office that procures all State Parks engineering and construction) and every FDEP Office of Resilience and Coastal Protection solicitation. Commodity codes are UNSPSC 8-digit, not the older NIGP 6-digit family some runbooks still reference. Codes worth selecting for a structural or civil practice: 81101500 Civil, 81101502 Technical drawing, 81101505 Structural, 81101510 Highway, 81101512 GIS, 81101514 Geotechnical, 81101522 Earthworks, 81101528 Stormwater, 81101606 Marine. Add bridge-engineering codes by in-app search.
• FDOT Contact Mailer (CMA) at cma.fdot.gov. The email-only signup the older runbooks describe is gone; CMA is now an account-gated app. Sign in, register, and subscribe inside the app to Bid Solicitation Notification plus any relevant Professional Services groups. CMA covers FDOT lettings statewide and the design-consultant Professional Services channel. Bidding on Professional Services contracts requires FDOT prequalification under §287.055 F.S., a separate longer process best treated as a later task, not Phase 1 ingestion.
• SAM.gov at sam.gov/opportunities. Authentication is through Login.gov; if the account is not already verified, build in extra time because ID verification can stall a day if it goes to manual review. Inside SAM, create multiple narrow saved searches rather than one mega-search. A starting set for a Florida engineering consultancy: NAICS 541330 in Florida, plus keyword alerts for boardwalk, bridge inspection, pier repair, and forensic engineering. Set each alert to daily.
• OpenGov at procurement.opengov.com. One free vendor account covers Pinellas, Hernando, and Sarasota counties. Follow each county portal inside OpenGov and set commodity categories for civil engineering, structural inspection, construction inspection, and environmental. Hernando includes Weeki Wachee Springs State Park, a direct fit for the state-parks engineering vertical.
• DemandStar, reached through the Manatee County portal. Vendor registration is free, and the free tier covers notifications for a single county (one agency). Additional counties run $35 per year each (verified against DemandStar's 2026 supplier pricing). A firm watching both Manatee and Pasco picks one as the free county and pays $35 per year for the second, not one registration covering both. Set product categories to engineering, construction inspection, bridge, and structural.
• Hillsborough County Bonfire at hillsboroughcounty.bonfirehub.com/portal. Not the look-alike hcfl.bonfirehub.com, which is a different tenant. Register and select engineering, construction, and inspection commodity categories.
• SWFWMD at swfwmd.state.fl.us/business/procurement. Complete the Vendor Registration and Payment Information form and opt in to bid notifications. Two cautions from a live 2026 registration: the district returns the actual forms through a Kiteworks secure-email portal rather than as a plain reply (check spam), and completion is gated on E-Verify enrollment, mandatory under FS 448.095, before the district will register the vendor. The E-Verify MOU is worth storing alongside the W-9 and certificate of insurance, because the same gate recurs across Florida public-agency vendor registration. Once the registration clears, alerts arrive from both swfwmd.state.fl.us and watermatters.org.

Seven accounts, seven sender families, eleven distinct channels of inventory. Zero subscription cost.

What the autonomous loop looks like.

The seven subscriptions produce email. Email is a tractable input. The pipeline shape that actually works for a one-person firm runs in six stages.

Mail capture. An Apple Mail rule fires on the seven sender families and routes inbound bid notifications to a Business Development mailbox. A second rule runs an AppleScript handler that calls a Python worker. The mailbox lives on iCloud rather than On My Mac, so the inbox is visible from phone and tablet.

Lead parse. The worker parses each notification into a structured lead record: agency, project title, scope text, location (county, FDOT district, state), deadline, value band, MasterFormat codes, NAICS codes, positive and negative keyword hits, raw email path. A small Claude call extracts the unstructured fields (scope text, value band) that the agency formats do not present consistently. Cost per parsed lead lands in pennies, carried as firm overhead, not client-billable.

Score. A second worker scores the lead against a versioned Experience Matrix: project-type weight (40 percent), geography weight (15 percent), agency weight (15 percent), scope-size weight (15 percent), keyword signals (15 percent, additive and capped at ten hits), then a negative-keyword multiplier that halves the total when a disqualifying term appears (MEP-only, architectural-only, single-family residential). The range is 0.0 to 1.0: a strong match at 0.75 and up, a watch-tier match from 0.50 to 0.74, and a skip below 0.50.

Notify. Strong matches fire a notification immediately. Watch-tier and skip leads write silently to a weekly scored log. A scheduled job sends a 6:00 AM digest summarizing the last 24 hours of scored leads.

Promote. Strong matches that warrant pursuit get promoted from a lead record to a pursuit folder with a canonical template: a state file, a directory for the RFP or RFQ as posted, a directory for the firm's draft response, and a directory for Q&A with the agency.

Log outcomes. Every pursuit that reaches at least the pursuing status appends one row to a win-loss log: pursuit ID, captured date, submitted date, outcome, winner, winning bid, the firm's bid, notes. The ledger is what makes the system measurable: conversion rate by channel, conversion rate by project type, average time from capture to submission, win rate against named competitors. None of those numbers are guessable without it.

The whole loop is small: roughly 60 to 90 minutes of subscription setup once the mail rules and the scheduled digest are in place, plus four to six hours of focused engineering for the scoring and digest workers. The scripts run as user-level workers. No servers, no cloud bill.

The Experience Matrix is the actual asset.

The scoring weights are where the system stops being generic. The matrix encodes one firm's profile: which project types fit, which geographies are reachable, which agencies are the right buyer, what scope size matches the team's turnaround, and which keywords signal a strong match. For a Florida structural, civil, forensic, and inspection consultancy, the high-fit categories sit at or near 1.0: bridge structural design and rehabilitation, bridge inspection, forensic structural investigation, boardwalks, marine and coastal structures, underwater structural inspection, UAS drone inspection under Part 107, and prestressed and post-tensioned concrete repair. Geographies cluster around FDOT District 7 (Hillsborough, Pinellas, Pasco, Hernando, Citrus) with surrounding districts at reduced weight. The agency families that matter most: FDOT, FDEP and State Parks, county park-rec and public works, USACE, FEMA, and NPS.

Scope size carries its own weight band, and the threshold there is a business decision, not a technical one. The matrix sets a value floor at $10K and puts the sweet spot at $10K to $250K, because a small-package response runs roughly ten hours end to end (qualification, partner-PE coordination, drafting, QA, portal submission). Below $10K a perfect-fit win still loses money once the partner-PE stamp split comes out; above $500K the small-firm subconsultant role lands in the watch tier rather than a strong-match alert.

The matrix is a plain text file, versioned in the same workspace as the leads. The weights are starting weights, not settled ones: the first cut is a best guess, and the second cut is the one informed by 30 days of actual capture volume and conversion.

What to evaluate before building this.

Five questions matter more than the channel list.

• Actual technical capacity to maintain a small pipeline. A firm without anyone who can read and modify the code should not build this. The pipeline is small, but it is real software. The right answer for a non-technical firm is to subscribe to the same channels and triage by hand for 30 days, then build the pipeline later, when the manual triage cost makes the engineering pay back.
• Pursuit cadence. A firm that pursues one bid a quarter does not need real-time alerts; a daily digest is fine. A firm chasing five pursuits a month wants real-time notifications on strong matches and a same-day review block.
• Partner-PE coverage. Many high-fit scopes (bridge structural, forensic, geotechnical, marine and coastal) require a PE stamp. A small firm without an in-house PE needs a confirmed roster of partner PEs willing to stamp the relevant disciplines per project. The matrix scores fit; the partner roster determines whether the firm can actually submit. A small registry of partner coverage lets the daily digest flag leads where no confirmed partner covers the discipline.
• Geography cap. A Florida-only firm flags out-of-state leads at near-zero weight. A firm with Southeast US forensic reach keeps a modest weight on the Georgia, Alabama, and South Carolina slice. The matrix has to reflect what the firm will actually pursue.
• Which channels send through which aggregator. Several channels have moved off direct county email onto centralized procurement platforms. Bid alerts now arrive from opengov.com, bonfirehub.com, demandstar.com, or myfloridamarketplace.com, not from the agency's own domain. Without the aggregators in the mail rule, those leads silently fail capture. This is the most common silent-failure mode in a pipeline of this shape.

When paid sources start to earn their keep.

After 30 days of running the free-channel pipeline, the firm has real numbers: leads per week per channel, strong-match rate, false-positive rate (strong matches the firm would actually skip), and false-negative rate (skipped leads the firm would actually pursue, surfaced through manual spot-checks).

The question to ask of ConstructConnect, or Dodge, or any paid layer at that point is concrete: would a subscription (entry $129 to $199 per month per market, more for real coverage) add pursuit volume the free channels are missing, in scopes the firm would actually pursue, at a rate that pays back the cost? If the free-channel inventory already exceeds the firm's pursuit capacity, paying for more lead volume does not help; the bottleneck has moved to bid-response throughput. If the free channels systematically miss a category the firm cares about (large multi-trade vertical projects, sub-consultant scope on large structural rehab), the paid layer earns its keep in that specific category. The decision becomes a data question rather than a sales pitch.

How Rarefied Earth thinks about this work.

The firm's own pipeline is the one described above. ConstructConnect was evaluated and skipped in Phase 1. The seven free channels are subscribed and live. The capture path runs through an Apple Mail rule architecture that routes notifications to a Business Development mailbox on iCloud, visible from phone and tablet. The Experience Matrix, the pursuit-folder template, the outcomes ledger, and the channel registry all live in the same workspace as the rest of the firm's operating substrate.

The point of building it this way is not engineering elegance. The point is a measurable pipeline. After the first 30 days, the matrix gets tuned against real lead volume rather than against a best guess. The decision to subscribe or not subscribe to a paid layer gets made against numbers. The firm's own pursuit conversion is in a ledger that survives the next workspace migration. For a small Florida engineering consultancy reading this from a similar starting position: do not subscribe to ConstructConnect first. Subscribe to the seven free channels first. Build the capture pipeline if there is anyone in the building who can maintain it; triage manually for 30 days if not. Re-evaluate the paid layer after the first month of measured volume. The free-channel inventory is wider than the marketing implies, and the data the firm needs to make the paid-layer decision does not exist until the firm has 30 days of its own captured leads.

Sources and further reading.

Related work.

This piece is the front-office companion to the firm's field guide on how AI takeoffs work for general contractors: the takeoffs guide covers what happens after a contractor wins a bid, and this one covers what happens before the engineer of record ever sees the RFP. The same structural-engineering background runs through the firm's PCI Journal pile-bent restoration research.