The in-house programmatic vs agency decision is usually framed as a cost question and almost never modeled as one. The business case a CMO receives typically compares a single line - the agency's percentage-of-spend fee - against the fully loaded salary of a programmatic trader, finds a favorable difference, and recommends in-housing. That comparison omits DSP platform fees, seat minimums, data and audience costs, verification and brand safety vendors, creative operations, reporting infrastructure, and the two or three additional roles a functioning trading operation actually requires.
Run properly, the model produces a specific and useful number: the annual working media level at which in-housing becomes cheaper. For most luxury and high-consideration advertisers, that threshold sits far higher than the business case assumes - and cost is not the variable that should decide it anyway.
In-house programmatic vs agency: what each model actually costs
Start with the agency side, because it is the simpler stack. Fee structures fall into four families, and the distinction between them matters more than the headline percentage.
| Fee model | Typical range | How it behaves | Best suited to |
|---|---|---|---|
| Percentage of media spend | 8%–18% of working media | Scales with budget; agency economics improve as spend grows | Brands with stable, growing budgets |
| Flat monthly retainer | $12,000–$60,000/month | Decoupled from spend; predictable | Brands with volatile or seasonal spend |
| Hybrid retainer + performance | $8,000–$30,000 base + 3%–8% incentive | Aligns upside; requires a clean metric | Brands with defensible incrementality measurement |
| Cost-plus / disclosed | Documented costs + 15%–25% margin | Full transparency into every dollar | Procurement-led organizations, large advertisers |
The percentage-of-spend model dominates the middle market and is the source of most in-housing pressure, because at $8M in annual working media a 12% fee is $960,000 - a number that reads like an obvious savings opportunity when set beside a $180,000 trader salary.
Now the in-house stack, which is where the analysis usually breaks down.
| Cost component | Annual range | Notes |
|---|---|---|
| DSP platform fee | 6%–15% of media | Charged regardless of operating model; not saved by in-housing |
| DSP seat minimum / license | $60,000–$250,000 | Most enterprise DSPs require committed annual spend or a license floor |
| Programmatic trader (1–2 FTE) | $110,000–$185,000 each, fully loaded ~1.35x | Two are needed for coverage, vacation, and departure risk |
| Media analyst / reporting | $85,000–$130,000 fully loaded | Required unless the DSP's native reporting is accepted as sufficient |
| Ad ops / campaign management | $75,000–$115,000 fully loaded | Trafficking, tagging, QA, discrepancy resolution |
| Third-party data and audience fees | $1.00–$4.50 CPM incremental | Wealth and intent segments carry the highest rates |
| Verification and brand safety (IAS, DoubleVerify) | $0.15–$0.45 CPM | Non-negotiable for luxury inventory standards |
| Reporting infrastructure / BI | $25,000–$120,000 | Warehouse, ETL, dashboarding, log-level data storage |
| Training, certification, turnover | $20,000–$60,000 | Programmatic roles turn over every 18–26 months on average |
Two components deserve emphasis because they are almost always missed. First, the DSP platform fee is not saved by in-housing - it is charged to the seat holder either way, so a brand comparing a 12% agency fee against zero is comparing against a number that is really 6% to 15% before any labor. Second, one trader is not an operation. A single-threaded trading function means campaigns pause during vacation, and a departure creates a three-to-five-month capability gap in a discipline where the platform configuration lives largely in one person's head.
Where the in-house programmatic vs agency break-even actually falls
Modeling the two stacks against annual working media produces a crossover point. Using midpoint assumptions - a 12% agency fee, a two-trader in-house team with one analyst and one ad ops resource, standard data and verification costs, and a mid-tier reporting stack - the picture looks like this.
| Annual working media | Agency cost at 12% | In-house fully loaded cost | Effective difference |
|---|---|---|---|
| $2,000,000 | $240,000 | $780,000–$950,000 | In-house costs 3–4x more |
| $5,000,000 | $600,000 | $820,000–$1,000,000 | In-house costs 35–65% more |
| $10,000,000 | $1,200,000 | $900,000–$1,120,000 | Roughly at parity |
| $20,000,000 | $2,400,000 | $1,050,000–$1,350,000 | In-house saves 45–55% |
| $50,000,000 | $6,000,000 | $1,400,000–$1,900,000 | In-house saves 70%+ |
The crossover sits between $9M and $14M in annual working media for most advertisers under these assumptions. Below roughly $8M, in-housing is a cost increase presented as a cost reduction. Above $20M, the savings become large enough that the operating-model question is genuinely open.
This is the number most in-housing decks never produce, and it explains a pattern worth naming: a meaningful share of brands that in-house programmatic return to external management within two to three years, usually citing talent turnover and inventory access rather than cost.
The variables that matter more than cost
Cost determines whether in-housing is affordable. It does not determine whether it is correct. Four factors carry more weight for luxury advertisers.
Inventory access. This is the decisive one in premium environments. Direct private marketplace deals with premium streaming and publisher inventory are relationship-mediated. A brand spending $4M annually is a small account to a major streaming platform's programmatic sales team; an agency aggregating $60M across a portfolio of luxury clients is not. That difference shows up as deal priority, access to scarce premium placements, first look at new formats, favorable floor pricing, and responsiveness when a deal underdelivers. In-housed brands consistently report weaker PMP access than they anticipated, and it is the hardest gap to close because it is structural rather than skill-based.
Measurement capability. Running a DSP is a mechanical skill that can be hired. Designing a geographic holdout, running a properly powered incrementality test, or building a media mix model is a specialized analytical discipline. Brands that in-house the buying but not the measurement often end up optimizing to platform-reported conversions, which is precisely the failure mode in-housing was supposed to eliminate. Platform-attributed conversions in luxury categories routinely overstate incremental conversions by 30% to 70%.
Spend volatility. In-house teams are a fixed cost. Agency fees on a percentage model are variable. For brands with pronounced seasonality - Q4-weighted retail, spring-weighted residential, event-driven hospitality - a fixed team sized for the peak is expensive during the trough. This dynamic alone disqualifies in-housing for a number of otherwise well-suited advertisers.
Category velocity. Premium CTV, curation, clean rooms, retail media, and identity infrastructure change materially every twelve to eighteen months. An agency amortizes that learning across a client portfolio. A two-person in-house team amortizes it across one brand and learns primarily from its own mistakes.
Hybrid models: what most sophisticated advertisers actually run
The framing of in-house programmatic vs agency as a binary is itself the error. The structures that hold up over time are hybrids, and there are four common patterns.
- Brand-owned seat, agency-operated. The advertiser holds the DSP contract and owns the data, log-level files, and audience assets; the agency operates the platform. This preserves transparency and portability without absorbing the staffing burden. It is the most common structure among sophisticated mid-market luxury advertisers and, in most cases, the right default.
- Split by channel. In-house the channels that are operationally simple and continuously running - paid search, brand social, retargeting - and keep external management for premium CTV, PMP negotiation, DOOH, and audio, where relationships and specialist knowledge dominate.
- Split by function. External buying with in-house strategy, or in-house buying with external measurement. The second variant is increasingly common: the brand runs the platform, an independent partner designs and adjudicates the incrementality work so the party buying the media is not the party grading it.
- Transitional in-housing. A defined two-to-three-year program with agency-led knowledge transfer, documented playbooks, and staged handover of specific channels. Materially more likely to succeed than an abrupt cutover, and it preserves inventory relationships through the transition.
Auditing your current arrangement before deciding
Most brands considering in-housing are responding to a transparency concern rather than a cost concern. That concern is usually addressable without restructuring. Before making an operating-model decision, ask for five things in writing:
- A full disclosure of every fee and take rate in the chain: agency fee, DSP platform fee, data costs, verification, any curation or marketplace fees, and any rebate or media-value arrangement.
- Log-level or impression-level data access, or a clear statement of what is withheld and why.
- The working media ratio - the percentage of your gross budget that actually clears as a bid into an auction. Below 60% warrants a hard conversation; competitive arrangements typically land between 65% and 80% after all platform, data, and verification costs.
- Supply path documentation showing which SSPs and resellers your spend traverses, per our guidance on supply path optimization.
- Ownership terms for audiences, deal IDs, and creative assets if the relationship ends. Portability is the single most valuable protection an advertiser can negotiate, and it is far easier to secure at signing than at separation.
An agency that answers all five without friction is delivering most of what in-housing promises. One that resists is the argument for in-housing, and the case is about control, not cost.
Five common mistakes in the in-house programmatic vs agency decision
- Comparing the agency fee to a salary. The complete in-house stack runs 2.5 to 4 times the headline salary line once platform, data, verification, infrastructure, and coverage roles are included.
- Assuming the DSP fee disappears. It does not. It is charged to the seat, whoever holds it.
- Staffing a single trader. One trader is a capability, not an operation; the departure risk alone justifies the second seat.
- In-housing buying while outsourcing nothing else - including judgment. Brands that in-house execution but not measurement typically end up more dependent on platform-reported numbers than they were before.
- Underestimating inventory access loss. This is the most frequently cited reason for reversal, and it takes twelve to eighteen months to become visible - long after the decision has been declared a success.
The talent problem nobody models
The in-housing business case treats staffing as a solved problem once the salary line is funded. In practice, hiring is where most in-house programmatic operations struggle first. Experienced traders with genuine premium video and PMP negotiation experience are concentrated in agencies, holding companies, and platform sales organizations, and they are expensive to pull out of those environments - a competent senior trader with five or more years of hands-on DSP experience commands $150,000 to $210,000 fully loaded in most markets, and considerably more in New York, Los Angeles, and Chicago.
The retention problem is worse than the hiring problem. Programmatic roles turn over roughly every eighteen to twenty-six months industry-wide, and the dynamic is structurally worse inside a brand: a trader at an agency works across a dozen accounts, several verticals, and multiple platforms, which is intrinsically better career development than running one advertiser's account for three years. Brands that in-house frequently find themselves hiring from the middle of the talent market, training the person on their own budget, and losing them to an agency or platform two years later - at which point the accumulated platform configuration, deal relationships, and campaign history walk out with them unless the operation has been rigorously documented, which it rarely has.
Budget realistically for this: an annual allowance of $20,000 to $60,000 covering certification, conference attendance, platform training, and the recruiting cost of one replacement hire is not padding, it is the actual run rate of keeping a two-person trading function staffed.
How to think about it if you are a luxury advertiser
Below roughly $8M in annual working media, in-housing is very unlikely to be economically justified, and the premium inventory access you would forfeit is disproportionately valuable in luxury categories where brand-adjacent placement is part of the product. Between $10M and $20M, a brand-owned seat with external operation captures nearly all the transparency benefit at a fraction of the organizational cost. Above $20M with stable, non-seasonal spend and a genuine ability to recruit and retain trading talent, full in-housing becomes a defensible strategic choice - though even then, most such advertisers retain external partners for premium video and measurement.
Stillwater Media works with a limited number of luxury and high-consideration brands each quarter, and we operate under whichever structure serves the client - including brand-owned DSP seats with full log-level transparency and documented supply paths. If you are evaluating your media operating model and want an honest read on where your working media ratio actually sits, apply to work with us.
Frequently Asked Questions
At what spend level does in-house programmatic become cheaper than an agency?
Under midpoint assumptions - a 12% agency fee against a two-trader in-house team with an analyst, ad ops support, standard data and verification costs, and a mid-tier reporting stack - the crossover falls between $9M and $14M in annual working media. Below roughly $8M, in-housing typically costs 35% to 300% more than agency management, and above $20M it can save 45% or more.
Do you still pay DSP fees if you bring programmatic in-house?
Yes. The DSP platform fee of roughly 6% to 15% of media is charged to whoever holds the seat, so it is incurred under both operating models and is not a saving created by in-housing. Enterprise DSPs also commonly impose a seat minimum or license floor in the $60,000 to $250,000 range, which a brand absorbs directly when it holds its own contract.
What does a brand lose by in-housing programmatic?
The most frequently cited loss is premium inventory access: private marketplace deals with major streaming platforms and premium publishers are relationship-mediated, and a single brand spending $4M annually carries far less priority with a platform's programmatic sales team than an agency aggregating tens of millions across a portfolio. Brands also commonly underestimate the loss of measurement specialization and the cross-client pattern recognition that keeps pace with a category that changes materially every twelve to eighteen months.
What is a hybrid programmatic operating model?
A hybrid model divides ownership rather than choosing between full in-housing and full outsourcing. The most durable pattern is a brand-owned DSP seat operated by an external partner, which preserves data ownership, log-level transparency, and portability without the staffing burden; other common splits allocate simple always-on channels in-house while keeping premium CTV, PMP negotiation, and measurement external, or keep buying in-house while an independent partner designs and adjudicates incrementality testing.
How do I tell whether my current agency arrangement is fair before deciding to in-house?
Request five disclosures in writing: every fee and take rate in the chain, log-level or impression-level data access, your working media ratio, supply path documentation showing which SSPs and resellers your spend traverses, and ownership terms for audiences, deal IDs, and creative if the relationship ends. A competitive working media ratio typically lands between 65% and 80% of gross budget after platform, data, and verification costs, and anything below 60% warrants a direct conversation.
