Most DSP owners get this decision wrong. They treat it as a binary — agency or in-house, pick one. The real market has three options, not two. And the right answer for most DSPs sits in the third option — the one most owners don’t know exists when they start the comparison.
The frame also skips a question that decides whether either choice will work: is the recruiting system repeatable, or does it live in one person’s head? That question — the repeatability test — is the difference between a $40,000 hire that scales the operation and a $40,000 hire who walks out the door with the whole system.
This article runs the math the way DSPs that get this decision right actually run it. Cost-per-hire by hiring volume. Speed to first placement. DSP-specific knowledge. Audit liability. Retention accountability. The numbers usually point to a different answer than salary versus fee.
Why This Choice Costs More Than It Looks
A bad recruiting choice doesn’t show up in the recruiter’s budget line. It shows up somewhere else. Open routes pull overtime up. Rookie drivers drag DCR and DAR scores down. Audit files get sloppy and Amazon contract risk goes up.
The typical 30-route DSP loses five drivers a month. That costs $9,000–$12,500 every month to replace them. Annualized, the bill runs $108,000–$150,000 — the equivalent of several routes’ worth of margin disappearing into a hiring loop. The recruiting choice sits upstream of all of it. Pick the wrong model and the cost compounds for the whole year.
That’s why this is a math problem, not a preference. The DSPs that get it wrong compare a salary to a fee. The DSPs that get it right compare cost-per-hire to hiring volume — and ask the repeatability question before they sign anything.
The Three Categories DSP Owners Actually Have
The binary frame leaves out the model most DSPs running 1 to 100 routes should be using. The real market breaks into three categories.
Contingency agencies. Pay-per-hire pricing, usually $1,500–$5,000 per driver placed. Generalist recruiters with no DSP-specific workflow. Useful for the occasional single-role fill. Useless for the volume a DSP actually needs to keep 25 to 30 routes running against 100% to 150% annual turnover. Most DSP owners try a contingency agency once and never go back. The model isn’t built for the cadence.
Software + done-for-you services — the third category. DSP-specific recruiting platforms with the recruiter labor included. Subscription pricing, usually $30,000–$50,000 a year. Cost scales with hiring volume instead of being locked at a full salary. Built for the way an Amazon DSP actually hires — automated screening, SMS pre-qualification, integration with Amazon’s onboarding flow, audit-ready file structure. This is the category most DSPs running 1 to 100 routes should look at first.
Staffing agencies. Full headcount provision. The agency carries the W-2. The agency manages the driver. The DSP pays a markup on driver wages, usually 25–40%. Used almost only for surge demand, peak season, or new station ramp. Not a steady-state recruiting solution at standard DSP economics. The math breaks fast at full-fleet utilization.
DSP owners who treat the question as a binary are missing the option that actually fits their operation.
How the Three Categories Compare
| Dimension | In-House Recruiter | Contingency Agency | Software + Done-for-You | Staffing Agency |
|---|---|---|---|---|
| Annual cost | $40,000–$75,000 fixed | $1,500–$5,000 per hire | $30,000–$50,000 subscription | 25–40% wage markup |
| Cost predictability | Fully fixed | Variable, scales with hires | Variable, scales with volume | Variable, scales with active drivers |
| Speed to first placement | 30–60 days (hire + ramp) | 1–2 weeks | 1–2 weeks | 2–4 weeks |
| DSP-specific knowledge | Depends on hire — most are HR generalists | Minimal | High — built for DSP context | Minimal |
| Retention accountability | Full — owned by DSP | Until placement only | Ongoing — incentive aligned | Vendor-managed, not DSP |
| Audit liability ownership | DSP owns, recruiter prepares files | DSP owns, agency hands off | Shared — varies by vendor | Split — agency on W-2, DSP on Amazon compliance |
| Scaling flexibility | Low | High | Medium-high | High |
| Best fit | 100+ routes, predictable volume | Occasional single-role fills | 1–100 routes, growth phase | Surge, peak season, ramp |
The Break-Even Route
The route count where in-house starts to make math sense is the question DSP owners actually need to answer. Generic recruiting industry sources cite 60 routes as the threshold. DSP-specific sources cite 100. Both are wrong as universal answers. Neither one accounts for the variables that actually move the math.
The honest framing is a 60 to 100 route gray zone with two questions inside it.
Below 60 routes. Outsourcing wins almost every time. A DSP this size hires somewhere between 15 and 35 drivers a year. Divide that into a $40,000–$75,000 in-house recruiter cost and the recruiter alone runs $1,200–$5,000 per hire — before adding Indeed spend, background checks, or the rest of the funnel. The cost-per-hire math doesn’t compete with a software + services model running $30,000–$50,000 a year with the recruiter labor built in.
100+ routes. In-house starts to make math sense if two things are true. Hiring volume is steady enough to keep a recruiter busy year-round. And the recruiter can absorb HR work beyond hiring — compliance, file maintenance, training coordination. When both are true, $40,000–$75,000 in fixed cost spread across 50+ hires a year drops cost-per-hire below what most outsourced models charge.
The gray zone — 60 to 100 routes. This is where the break-even route sits. It’s where most DSP owners get the decision wrong. Two questions resolve it.
First: is hiring volume predictable, or does it spike with new station launches and seasonal pressure? Predictable volume favors in-house. Spiky volume punishes the fixed cost.
Second: will the recruiter own work beyond hiring? An in-house recruiter who only recruits is underutilized at 30 to 50 hires a year. An in-house recruiter who also owns onboarding, compliance, and training coordination justifies the salary.
Run those two questions before defaulting to a route-count rule. The answer for most DSPs in the 60 to 100 route range is software + services — with the option to layer in-house when growth becomes predictable. The answer is rarely “60 routes, hire a recruiter.”
The Hybrid Model: When It Works, When It Breaks
The hybrid model is the operational pattern most multi-station DSPs eventually land on. In-house recruiter handles baseline hiring. Agency or staffing partner handles surge. DSP-specific software runs the workflow underneath. The pattern is real and sometimes correct. It’s also the model that breaks most often when ownership boundaries aren’t explicit.
When it works. The DSP has steady baseline hiring volume that justifies an in-house recruiter, plus occasional spikes — new station launch, peak season, sudden turnover wave — that exceed in-house capacity. The agency or staffing partner is contracted for surge only. The software handles the funnel for both. Applicant tracking, screening criteria, scheduling, file structure all live in one system. Handoffs between the in-house recruiter and the agency don’t drop candidates. Done well, this model holds cost-per-hire steady through volatility while keeping retention accountability inside the DSP.
When it breaks. Two recruiters, one funnel, and no one owns the candidate. The in-house recruiter assumes the agency is screening agency-sourced candidates. The agency assumes the DSP is doing the final compliance check. Candidates fall through the gap. No-show rates climb. The DSP starts paying full agency placement fees on hires that should have come through the in-house pipeline at a fraction of the cost. The handoff problem is the dominant failure mode.
The fix is documentary. Written ownership boundaries on every funnel stage. A single source of truth for the candidate file. A regular reconciliation between the in-house recruiter and the agency. A hybrid model with clean handoffs is usually the right answer for DSPs at 100+ routes or multi-station operations. A hybrid model without clean handoffs costs more than either pure approach and produces worse retention.
The Repeatability Test
The agency-versus-in-house frame skips the question that actually decides whether either choice will hold up: is the recruiting system repeatable, or is it institutional knowledge living in one person’s head?
The repeatability test runs in two directions.
If the in-house recruiter leaves tomorrow, can the next person pick up where they left off? If the answer is no — if the screening criteria are in someone’s head, the candidate pipeline is in their personal Indeed account, the compliance file structure is whatever folder system they invented — then the $40,000–$75,000 recruiter isn’t a recruiter. It’s a single point of failure with a salary attached. Most DSPs find this out the hard way when their first in-house hire leaves after 18 months and recruiting collapses for a quarter.
If the agency contract ends tomorrow, is anything left behind? A contingency agency on a per-hire model leaves nothing. No documented screening process. No candidate pipeline. No operational knowledge transferred. The DSP starts from zero the day the contract ends. A software + services model in the third category usually passes this test better. The workflow, the candidate database, and the screening criteria live in the platform the DSP keeps using regardless of who runs it.
The repeatability test separates DSPs that scale from DSPs that stay stuck at one station. Pick the recruiting model that produces a system, not the model that produces the cheapest individual hire. The cheapest hire stops mattering the moment the person running it walks out the door.
What To Do, By DSP Size
The decision framework that actually works once cost, speed, repeatability, and audit liability are all factored in.
1 to 2 stations, fewer than 60 routes. Outsource. Almost always. The volume doesn’t justify a fixed in-house salary. The third category — software + done-for-you services — is the right starting point. Cost-per-hire stays in the $200–$350 range in well-run setups. The system stays repeatable because it lives in the platform, not in a single person.
60 to 100 routes, single station. Run the gray-zone math. If hiring volume is steady and the recruiter can own HR work beyond hiring, in-house starts to make sense. If volume is spiky or the role is recruiting-only, stay with software + services and revisit at 100 routes. Don’t hire a recruiter to fill a hypothetical hiring forecast.
100+ routes, single or multi-station. In-house begins to win the math — but only if the operation passes the repeatability test before the offer letter goes out. Hire someone who documents their workflow into the system, not someone whose value is the workflow inside their head. Layer in agency or staffing partners for surge. Keep DSP-specific software running underneath.
Multi-station, 200+ routes. Hybrid is almost always correct, with handoff boundaries documented in writing. Centralized software, in-house recruiter for baseline, agency or staffing for surge, audit files reconciled monthly. The cost predictability problem disappears at this scale because hiring volume smooths across stations.
Closing Argument
The agency-versus-in-house question is a math problem disguised as a strategy question. The DSPs that get it right solve for cost-per-hire by hiring volume, not for salary versus fee. They evaluate three categories, not two. They name the gray zone between 60 and 100 routes instead of pretending one threshold applies. And they ask the repeatability question — whether the recruiting system survives the recruiter — before they sign anything.
In the work I have done across multiple Amazon DSP stations, DSPs that move from a binary frame to a three-category frame consistently bring cost-per-hire into the $200–$350 well-run range. They keep the audit readiness and retention accountability inside the DSP rather than at a vendor. The reframe alone changes the answer. The math does the rest.