Primes rarely call us because a sheet looks unfinished. They call because nobody can name the file the sheet came from. A GIS centerline on NAD83 and field notes collected on a local assumed system can render the same on a monitor. They still disagree the moment a duct run is scaled off the wrong basis. Procurement wrote that outcome months earlier, when the subcontract never named a controlling survey.
The buyer here is the prime that purchases OSP design rather than staffing it. Draftech's engineering stays in-house, inside a boundary the subcontract draws, so packages come back reviewable while the prime keeps every owner-facing decision it already owns. Construction sits outside that boundary unless the prime also buys full turnkey delivery. Draftech runs that work through Draftech-managed subcontract crews under its own QA/QC and safety program. One procurement can hold both scopes. They belong on separate exhibits.
What an OSP design subcontract for broadband project must settle
An OSP design subcontract for broadband project work is a prime-controlled agreement that assigns a defined in-house OSP engineering scope to Draftech. The prime keeps owner-facing authority. The written interfaces run from 1 signed design-basis page through HLD and LLD releases, then permit response and native DWG 2018 file custody.
The party nobody signs with still controls the route
Page one names 4 parties, not two: the prime holding the main contract; the owner receiving the network; Draftech as engineering author for the described scope; the external authorities whose permissions no clause here can grant on their behalf. Pole owners sit in that fourth slot, and so do DOT district permit offices. Promise an alignment the pole owner never authorized and the prime inherits a commitment it has to honor, usually in the week the permit was due back. We will not write that promise.
Federal Highway Administration 23 CFR 645 Subpart B, Accommodation of Utilities, listed on the FHWA Utility Principal Documents page with the date 11/22/2000, is the federal accommodation regulation for utilities on qualifying highway right-of-way. A design subcontract cannot grant that occupancy. Subpart B still decides whether a state transportation department may accommodate the facility, and the prime remains the party that has to live with a denial. We name those reserved decisions on page one rather than hiding them in a shared-responsibility row.
Where an OSP design subcontract for broadband project stops
The fiber network design outsourcing guide covers whether to buy design at all. The subcontract is what comes next, and it fails in one recognizable way: a responsibility chart that assigns every row to both parties. That reads as shared accountability. It works as an unresolved interface. Consultation can stay as wide as the prime wants. Authorship cannot be shared. One engineer's name goes on the geometry before the first sheet leaves our office. A second name goes on the check. Both names travel with every later revision of that segment so a reviewer two years later is not left guessing who froze the alignment.
Where a row has no owner at all, the prime's design manager assigns it before production leans on it. A segment whose utility permission is unresolved either waits or advances on a written assumption that states what happens when the assumption fails, including whether quantities already issued at HLD have to be withdrawn. Skipping that assignment is how a second, undocumented design gets built inside the first, on inputs nobody froze and nobody can later defend to the owner. Production does not start on that row.
Freeze the files production may treat as true
A prime hands over a data room and assumes completeness is self-evident. Two files in there will disagree, and the disagreement stays invisible on screen because both render correctly. Field-located vaults collected against NAVD88 and a GIS surface still sitting on an assumed project elevation can plot on top of each other until a riser length is computed off the wrong vertical. We surface that conflict before drafting volume makes it expensive, then we show what the wrong winner does to coordinates and to quantities already sitting in the owner's review. Choosing the winner is the prime's call. The subcontract does not let us pick it silently.
The approved design basis runs a page, and on a few programs it runs two. It states service intent and route limits, the architecture, the survey basis down to the horizontal and vertical datum, the owner standards the design has to meet, plus what each release stage must contain before anyone calls it that stage. The prime's design manager signs the page. The date on it is the date production may start. We read the rest of the data room for context. We design from the page.
Tie every open gap to the segment it affects
Our OSP design package guide separates planning, low-level design, permit sheets and construction deliverables so a buyer can see what each one holds. The subcontract makes that separation binding by writing acceptance criteria in evidence terms: which calculations are complete, and which attributes are populated, before a stage name goes on the cover. A package named LLD and a package fit for construction pricing are not always the same package. Production stays on the frozen basis until the prime's design manager authorizes a newer survey in writing, because an export timestamp never answers accuracy class.
| Subcontract interface | Draftech role | Prime decision |
|---|---|---|
| Input control | Register and test the technical inputs | Approve the source hierarchy |
| HLD and LLD | Author and check the in-house engineering | Accept each stage release |
| Permit response | Assess the technical effect and revise | Authorize scope or commercial change |
| Construction question | Return an engineered disposition | Approve the business impact |
| Closeout | Deliver accepted native and published files | Take final custody |
So the register records what arrived and what superseded it. Every open gap is tied to the segment it touches rather than swept into a general assumptions sheet, which is how designers work around a contained gap without quietly rewriting the rest of the route. Sheets and quantities that cross the gap stay provisional until it closes. That gives the prime real progress instead of a percentage that treats a large data room as approved direction. Each missing item is named and dated on the register, and it stays there until the prime's design manager closes it.
One question decides the whole freeze. When a newer survey file appears mid-stage, who authorized it to supersede the frozen basis? An export timestamp answers neither the accuracy class of the new survey nor the right to rewrite quantities the owner has already reviewed. We ask the prime's design manager for that authorization in writing. Production keeps building on the frozen basis until it arrives. Half a day of waiting beats a week of reconciling two sets of quantities. We will wait.
Tie every route view to one decision
A split baseline looks like this from the prime's side. The permit sheet puts the crossing at the section line. The LLD puts it east of that line. The bill of materials was built off the LLD. The agency has already stamped the sheet. Nobody was careless anywhere in that chain. An agency comment landed on the permit set. A drafter honored it. The change never travelled back to the released geometry. By the time procurement priced materials, four views of one route had become four routes. One of them is buildable. We catch that on one segment before we scale.
Our discipline leads work against controlled route segments instead of a folder of similarly named files. Every representation of a segment carries the revision identity of the decision behind it. HLD sets architecture and major capacity, then LLD adds placement, splice logic, quantities and construction detail on top of that architecture without opening a second baseline. Permit sheets present that same geometry in whatever form the reviewing agency asks for. That coordination is the daily substance of our fiber network design work, and it is the part a prime cannot read off a proposal because it only shows up when two views of the same crossing stop matching.
One segment, checked four ways
The white-label OSP design model matters here because presentation and authorship come apart. A prime that needs its own delivery identity gets its title block on the published sheet. Inside the file, the technical record still names the engineer who authored the geometry and the engineer who checked it. Professional certification follows whatever the jurisdiction requires. A reviewer opening that sheet two years later should be able to reconstruct the decision without calling the original designer. That holds whether or not our name is on the cover. We keep the names inside the file.
Before a production cell scales, we run one segment through all four representations and compare them line by line for route, capacity, placement, quantity and revision status. Formats may differ. The decision underneath cannot. When they diverge, only the affected representations go back to the discipline lead, and release resumes once geometry and revision identity agree again. The prime hears about it as a schedule item with a segment number attached, not as a stream of drafting chatter. Isolating the break to that segment is the whole point of the cell.
The isolation is deliberate, because a day of rework on one segment is cheap compared with discovering the same break in a construction package already out for bid, which then becomes a bid addendum the prime has to explain. We take that day on the first segment rather than on the hundredth.
Close each review cycle before the next issue
Ask a reviewer one question at the end of a cycle. Can you read this sheet without last week's markup open beside it? If the answer is no, the package is not released, whatever the transmittal says. Comments get logged and answered, then they are incorporated or rejected with a stated reason so the cycle actually closes. The marked copy becomes evidence of how the sheet reached its current state. Email attachments and shared redlines are fine for discussion. Neither becomes controlling just because it was the last file somebody opened.
Classify a mark before it moves geometry
Move the line is not one instruction. It can be a drafting correction from a sheet that never matched the surveyed curb. A constructability preference from someone who has built in that soil is a different request, and so is an owner's stylistic direction. An authority requirement with a permit number behind it is something else again. We record who made the mark and why, then we work out what honoring it does before anything moves. A comment that would push an alignment past the survey basis is an engineering request even when it arrived as a casual note in the margin. Both go back through the formal path.
Quality control is scoped to the use. That is why a permit exhibit can be complete while the construction package built from the same geometry is not. The agency wanted the crossing detail and the traffic control plan. The crew needs splice logic and a bill of materials that reconciles to the sheet it came from. Write into the subcontract which check runs at which stage. Or the prime ends up editing CAD files it never agreed to own. We will not take that edit as our QA.
The prime accepts an issue once each comment's status is visible at a glance: incorporated; rejected; deferred; superseded. Anything touching scope or quantity splits into its own commercial track so a drafting fix does not sit in a queue behind a contract argument. That sorting is what lets a reroute get priced honestly. We describe the effect on survey coverage, sheet count, permit resubmittal and materials against the accepted basis. The number the prime takes to the owner is one it can defend line by line.
Limitation we own: the limitation of our preferred frozen-basis process is that we will not let a newer survey replace the signed design basis without written authorization from the prime's design manager, even when that wait costs a production day. That trade-off protects quantities already sitting in the owner's review. It also idles a cell that could otherwise keep drafting.
Award recommendations after the sample package
If you are a single-state ISP or CLEC prime buying design only: award on the native-file sample rather than a PDF export. Attach that sample to the subcontract as the receiving standard. A firm whose files land clean costs the prime nothing after delivery. A firm whose files need rebuilding charges once in the fee and again in every hour the prime's own staff spend making the package usable. Ask each finalist for one closed project's comment log with client details stripped out. Read it for rejections. A log with no rejection in it means the hard questions were deferred to the field.
If you are a multi-state prime on a 2026 BEAD schedule: freeze the signed design basis before any production cell scales, because the permit clock starts on the geometry and nothing downstream recovers those weeks. Run the sample against a pair of conflicting inputs plus an owner comment dated after the sheet was issued. What you get back is a small, honest sample of how that firm will behave on the two hundredth segment, when nobody is watching the way they are watching now. We will return that sample the same way we return a first release.
If you expect to add construction later: price that exhibit at award while both scopes are still competitive. With Draftech the path is fixed by how the company is built. Engineering stays in-house. Full turnkey construction, when the prime elects it, runs through Draftech-managed subcontract crews under Draftech's QA/QC and safety program. A crew observation lands with the in-house engineers before any revised geometry is issued. Writing the option into the design subcontract now costs a paragraph. Adding it during the build costs a schedule. Price that option at award.
Every bidder's proposal describes a process. A sample package is the only part of a procurement that can be measured. Give each shortlisted firm one route segment, a pair of inputs that genuinely conflict, an owner comment dated after the sheet was issued, plus a week to return it. Ask for the native file. What comes back either survives the prime's own workstation or it does not. Attach that sample.
- Native open: open the DWG 2018 file on a prime workstation carrying only the prime's standard template, then count what fails to draw.
- Attribute round-trip: export to the format the prime's asset system ingests, then import the result back and list which attributes did not survive the trip.
- Comment travel: give the sheet and its comment log to an engineer who has never seen the segment, then ask what changed at the last revision.
- Exclusion pricing: put survey reconciliation on its own line; pole loading, permit revision rounds, certification, plus file conversion each get a line too.
Draftech's partner-led in-house engineering model is what stands behind a sample like that. Full turnkey construction stays a separate election under the managed crew framework. If the receiving standard is still undefined on your program, settle it before the award rather than after. Send the one segment your shortlist keeps answering differently to the Draftech design desk and compare what comes back against the files you already have to live with. That comparison is cheap now. It is expensive to run after signature. We would rather run it now.
Have a segment and a shortlist already? Send us the segment and the owner comment. We will return the sample package the same way we would return a first release.
