In-House Programmatic: Guides, Comparisons and Tools

In-house programmatic means taking responsibility for selected parts of media buying. Teams may own strategy, data, bidding logic or infrastructure in different combinations. Start with the ownership model and its operating cost.

Start with the business decision and the data flow. Identify who owns the customer relationship, who receives each signal, who runs the decisioning system, and what evidence supports the result. A vendor feature list alone cannot settle those questions.

This guide brings together the reference material for this topic. Comparisons describe different operating models, rather than declaring a universal winner. The linked pages retain source dates, vendor-claim labels and the limits of each example. Read those limits before applying a figure or architecture to your own organization.

1. Percentage of media spend. The platform fee is a cut of every dollar that passes through. The Trade Desk states that it charges "a platform fee that is generally based on a percentage of our clients' total spend on the platform" and reports revenue net of media (The Trade Desk 10-K, FY2025). Most self-service enterprise DSPs work this way. The rate is negotiated per contract and almost never published.

2. Flat software fee. The vendor charges a monthly license, sometimes with a volume cap (queries per second, impressions) and sometimes with a floor expressed as a percentage of spend. This is the white-label and bidder-as-a-service model. Epom's two listed tiers are "5% of your ad spend or $250 per month, whichever is higher" and "5% of your ad spend or $2000 per month, whichever is higher" (Epom DSP pricing, accessed 17 October 2026). Beeswax, now part of FreeWheel, was described at acquisition as software that "sits in buyers' own clouds and uses a software-as-a-service pricing model" (Beet.TV, 18 December 2020).

Read the full dsp pricing explained 2026: take rates, fees and minimums guide for context, sources and limitations.

Rates used, with their status:

Results (all dollar figures calculated from the rates above):

Read the full dsp pricing explained 2026: take rates, fees and minimums guide for context, sources and limitations.

Two features of the timeline matter for planning. The first is the ownership churn: a buyer who signed with AppNexus before the 2018 sale dealt with three owners and two renamings before the product closed, and the destination it was pointed to was itself renamed within seven months. The second is that Microsoft described the exit as a strategy change, not a failure: "the future of digital engagement is conversational, personalized, and agentic" and the DSP "no longer aligns with our investment in this future" (AdExchanger, 14 May 2025). A rented platform can be closed by a healthy owner for reasons that have nothing to do with you.

Read the full microsoft invest shutdown: dsp migration and exit playbook guide for context, sources and limitations.

Microsoft said it was "committed to supporting our clients during this transition" and would "continue to support access to Microsoft and partner inventory through third-party DSPs" (Microsoft Advertising, 14 May 2025). Inventory access was never the problem. The assets below were.

Creatives. Creative assets can be re-uploaded, but audit status, creative IDs used in reporting, macro configurations and third-party tag wrappers cannot. Each new platform re-audits, and until it does, delivery is throttled or blocked.

Read the full microsoft invest shutdown: dsp migration and exit playbook guide for context, sources and limitations.

Each question is scored 1 to 5. Each group has a weight, and the ten weights sum to 100. The group score is the mean of its six question scores divided by 5 and multiplied by the group weight; the total is the sum of group scores, out of 100. Six questions are pass/fail gates: a "fail" on any of them removes the vendor regardless of total. The threshold for a shortlist is 70; the threshold for a contract is 75 plus a clean proof test.

Scale. 5: answered fully, with evidence, and the term is offered in writing. 4: answered fully, evidence provided, term negotiable. 3: answered, no evidence, or term available at extra cost. 2: partial or evasive answer. 1: refused, or the red-flag answer given.

Weights.

Pass/fail gates: Q7 (data ownership), Q13 (log-level data available), Q49 (current SOC 2 Type II or ISO 27001), Q50 (DPA and subprocessor list), Q56 (sunset clause), Q57 (export at exit in documented formats).

Read the full dsp rfp template: 60 questions and a scoring sheet guide for context, sources and limitations.

Paste the tables into your RFP as they stand, or download the CSV. Each row has the question, why it matters and the answer that should worry you.

Read the full dsp rfp template: 60 questions and a scoring sheet guide for context, sources and limitations.

• Own means the company built or operates the layer itself, with public evidence.
• Rent means a third party provides the layer under contract.
• Hybrid means both, with the split described in the cell.
• n/d means not disclosed in any public source we could find. We do not guess.
• Netflix is a media owner selling ads; Lego and P&G are advertisers buying them. The layers are the same, but the direction of money differs, so read Netflix's "DSP" row as "which DSPs it sells through". The layers follow the chain described in RTB architecture explained.

Read the full netflix, lego and p&g in-house ad tech: what they built guide for context, sources and limitations.

Netflix is the clearest case because it reports to shareholders every quarter. The sequence is dated and public.

Three things stand out.

First, the thing Netflix built is the ad server and the data layer, not a bidder. The Ads Suite decides which ad plays in a Netflix stream and exposes Netflix data through APIs. Demand still arrives through other companies' DSPs. Netflix never announced its own DSP and, as a seller, has no reason to.

Read the full netflix, lego and p&g in-house ad tech: what they built guide for context, sources and limitations.

Every line below is a company claim unless a third-party source is given. Prices change; check the linked page.

Beeswax is the only vendor we found that markets the category by that name. Other platforms offer custom-bidding hooks, but as a feature of a managed DSP rather than as a single-tenant product.

Two platforms that searchers put in this group do not belong in it. Basis Technologies sells a unified buying platform with its own DSP to agencies and in-house teams, and backs its in-housing pitch with a Forrester study it commissioned, reporting a 48% ROI (Basis, 10 September 2021); that is a managed, multi-tenant platform, not a white label. Viant's Adelphic is a "subscription-based omnichannel demand-side platform" for self-service buyers (Viant, 4 May 2017); we found no white-label program. A vendor called "DSP14" appears in search suggestions, but no live site resolved for that name on 17 October 2026, so it is not listed.

The self-hosted route is mostly custom development today, which is why how to build a DSP platform is a separate page.

Read the full bidder as a service vs white label dsp vs self-hosted dsp guide for context, sources and limitations.

Three mechanics drive the cost difference.

Read the full bidder as a service vs white label dsp vs self-hosted dsp guide for context, sources and limitations.

The three models are often combined.

Read the full bidder as a service vs white label dsp vs self-hosted dsp guide for context, sources and limitations.

A DSP has two halves that run at different speeds. The hot path handles every bid request in milliseconds: parse, filter, match campaigns, price, respond, log. The warm path runs in seconds to hours: ingest wins and events, update budgets and pacing, refresh audiences, build reports, reconcile against exchange invoices. Most build plans staff the warm path, which looks like ordinary software, and discover the hot path, which looks like a trading system, after the first exchange integration.

Read the full how to build a dsp platform in 2026: qps, team, cost guide for context, sources and limitations.

Every exchange you connect to is a separate project: an onboarding form, a test endpoint, a QPS quota, a certification pass and an account manager who decides how much traffic you see. BidsCube scopes an MVP at "5-10 SSP integrations" and names Google AdX, Magnite, PubMatic and OpenX among the first (BidsCube, 17 July 2026). Geomotiv adds "$50k+" of "integration debt" for identity integration, supply-path logic and "SSP testing or certification" (Geomotiv, 5 May 2026).

Read the full how to build a dsp platform in 2026: qps, team, cost guide for context, sources and limitations.

The published ranges are below, with dates. They do not agree because they scope different things: an MVP versus a full platform, a vendor-built versus an in-house team, and whether salaries are included.

Read the full how to build a dsp platform in 2026: qps, team, cost guide for context, sources and limitations.

Three different things are called working media, and the number changes by ten to thirty points depending on which one a study reports.

Money that reaches the sell side. This is advertiser spend minus buy-side fees (agency, DSP, data, verification) and minus the SSP fee: what the publisher's ad server receives. The ANA calls it "seller revenue" and measured it at 76.2% of spend in Q3 2025 (ANA Benchmark, Q3 2025). The CMA's 65% and ISBA's 51% are the same concept measured on different data (CMA, Appendix R, 1 July 2020; ISBA, May 2020).

Money that buys a quality impression. The ANA's headline metric, TrueAdSpend, is "the ad spending going to TrueImpressions", where a TrueImpression is one that is "non-IVT, measurable for viewability, and viewable" (ANA Benchmark, Q1 2025); the Q1 2026 release describes the index as measuring "programmatic spend delivering fraud-free, measurable, viewable, and MFA-free impressions" (ANA, 27 May 2026). That is the 45.1% figure. It is seller revenue with the wasted impressions removed.

Read the full ad tech tax explained 2026: every fee, with ana numbers guide for context, sources and limitations.

The studies disagree because they measure different things on different data. The table lists each dated public datapoint with its scope.

Read the full ad tech tax explained 2026: every fee, with ana numbers guide for context, sources and limitations.

The matrix is deliberately symmetrical: every model has every line, even where the default is zero, so that a vendor's "no fees" and a builder's "no licenses" are both visible as zeros rather than omissions.

Read the full ad tech tco calculator methodology: costs, formulas, sources guide for context, sources and limitations.

Use the interactive model above or the downloadable workbook to compare the operating models. The calculation includes the parallel-run cost entered for migration. Hosting is an editable annual budget; use the bidder capacity planner to estimate it from measured load.

The original draft’s static worked-example totals did not include all parallel-run costs described in its formulas. Those totals and the derived break-even claim are withheld. Compare the calculated results with your own contract, staffing plan and transition schedule. Data and supply-side fees remain visible in each model so the comparison uses the same cost boundary.

Read the full ad tech tco calculator methodology: costs, formulas, sources guide for context, sources and limitations.

The XLSX contains an Inputs sheet and a Model sheet. Inputs carry source links or editable-assumption labels. Results use native formulas for recurring costs, migration, an optional exit provision and the rented-versus-owned break-even calculation. Hosting is an editable annual budget informed by the bidder capacity planner.

Read the full ad tech tco calculator methodology: costs, formulas, sources guide for context, sources and limitations.

All figures are the ANA cost waterfall (sequential averages across contributing advertisers) unless marked median. "n/d" means the ANA has not published the figure outside its gated report.

The ANA began reporting higher- and lower-performing cohorts in late 2025. The gap is the most useful number on this page for an individual advertiser, because it shows what the same market yields to buyers with and without control of quality, price and supply.

These are not comparable with the ANA series (different scope, geography and method) and are listed for context. The method differences are explained in The Ad Tech Tax 2026.

Read the full working media benchmark: programmatic transparency tracker guide for context, sources and limitations.

What it leaves out. "Agency fees, ad serving fees, managed service fees, and brand safety metrics are not part of the Cost Waterfall" (ANA Benchmark, Q1 2025). Seller revenue in this series is therefore higher than the publisher share in ISBA, which included agency fees.

Third-party inputs. MFA classification comes from DeepSee.io and carbon from Good-Loop (ANA Benchmark, Q3 2025); a change in either, or in the MRC-based viewability standard the verification vendors apply, would break the series.

Read the full working media benchmark: programmatic transparency tracker guide for context, sources and limitations.

Platform data is the only input no one else has, and the one readers will trust least, so the rules are strict and public.

Read the full the cost of renting ad tech: annual report methodology guide for context, sources and limitations.

Every metric is reported with its source, date, country and sample, and metrics from different definitions are never averaged together.

Read the full the cost of renting ad tech: annual report methodology guide for context, sources and limitations.

1. Headline: the fee share per dollar this year, last year, and the five-year view.
2. Method and sources (this page, condensed).
3. The dollar's journey: a waterfall from advertiser spend to publisher receipt, US and UK, with each study's figure.
4. Working media over time: the ANA series, with the metric changes marked.
5. Take rates from filings: DSP, SSP and retail media platforms, with the calculation shown.
6. Fee variance: what the same inventory costs through different paths.
7. The unattributable gap and what closed it.
8. Retail media: gross billings to net revenue.
9. What owning costs: TCO at three budgets; infrastructure per billion requests; staff.
10. Platform data: the opt-in aggregates, with contributor counts.
11. Limitations.
12. Appendix: source register, data dictionary, aggregation code, "cite as".

Read the full the cost of renting ad tech: annual report methodology guide for context, sources and limitations.

How much does Amazon DSP cost? Amazon does not publish a rate card. Its product page says "Your current pricing structure is carried forward" into Amazon Ads Agent and refers pricing questions to an account executive (Amazon Ads, accessed 17 October 2026). Documentation summarised in July 2025 put the managed service at a $50,000 minimum, a 15% service fee and a 10% platform fee (PPC Land, 15 October 2026). Reports on the self-service minimum conflict: one dated guide says Amazon removed it in November 2025 (Nova Data, updated 8 October 2026), while Digiday reported on 11 November 2025 that the DSP kept its minimums (PPC Land, 15 October 2026). Ask for both in writing.

What is The Trade Desk's take rate? About one fifth of spend. Revenue of $2,896.3 million against gross spend of $13,394.7 million in 2025 implies 21.6%, and 20.3% in 2024 (The Trade Desk 10-K, FY2025). The company said in April 2025 that the rate "has been consistently around 20% of spend on our platform" (AdExchanger, 28 April 2025). Individual contracts vary; Needham has estimated a 12% to 15% range for the platform fee alone (PPC Land, 4 September 2026).

Does DV360 charge a platform fee? Yes, but Google does not publish it. The UK CMA, using 2019 data supplied by Google, estimated the DV360 take rate at 5% to 15% plus third-party fees (CMA, Appendix R, 1 July 2020). Google's own help pages are specific only about data: third-party lists cost "$0.50 to more than $2 CPM" (Google DV360 Help, accessed 17 October 2026).

How much does it cost to start a DSP? From $250 a month for a white-label front end with a 5% of spend floor (Epom DSP pricing, accessed 17 October 2026), to a usage-based flat fee for a bidder-as-a-service (PPC Land, 17 December 2020), to a build whose cost is staff and infrastructure rather than a license. No vendor publishes a bidder-as-a-service or self-hosted price; the figure comes from an RFP.

Are DSP fees negotiable? Yes. The Trade Desk's filing lists "pricing competition, volume discounts and shifts in media, client and channel mix" among the reasons its revenue may not grow in line with spend (The Trade Desk 10-K, FY2025). The CMA found DSP take rates ranging from 5% to 42% across platforms in 2019 (CMA, Appendix R, 1 July 2020). What is rarely negotiable without log-level data is visibility into the fees taken after the DSP.

• DSP Pricing Explained 2026: Take Rates, Fees and Minimums: What a DSP costs in 2026: Amazon, DV360, The Trade Desk, white-label and self-hosted fees, data and log fees, minimums, plus a $1M worked example.
• Netflix, Lego and P&G In-House Ad Tech: What They Built: Layer-by-layer table of what Netflix, Lego and P&G built in-house (ad server, DSP, data, clean room, measurement, creative), what they rent, and lessons.
• Ad Tech Tax Explained 2026: Every Fee, With ANA Numbers: Trace programmatic advertising fees from advertiser to publisher. Compare dated ANA, ISBA and Adalytics findings, cost boundaries and reporting limits.
• Working Media Benchmark: Programmatic Transparency Tracker: Quarterly tracker of how much programmatic spend reaches working media: ANA 36% (2023), 43.9% (Q4 2024), 45.1% (Q2 2026), plus ISBA, CMA and Adalytics.
• Ad Tech Cost Optimization: 12 Ways to Cut Spend: Ad tech cost optimization starts with a consistent cost boundary. Compare the workflow, operating decisions and required evidence.
• In-House Programmatic: Why Brands Like Lego Are Switching: In-House Programmatic: a practical guide to the definitions, operating decisions, measurement and evidence you need.
• Media Agency Rebates and Kickbacks: Executive Guide 2026: Media rebates and kickbacks create a governance question: does the buyer understand how its agency and intermediaries are compensated? Require clear.
• Programmatic Advertising 2020–2026: What Changed and Why: Programmatic advertising has expanded from automated media buying into a broader question of control over data, decisioning and the supply chain.
• Cloud DSP Costs: Fees, Data Ownership and Exit Terms: Cloud DSP Costs: a practical guide to the definitions, operating decisions, measurement and evidence you need.
• White Label DSP: Platforms, Pricing and Deployment 2026: White Label DSP: a practical guide to the definitions, operating decisions, measurement and evidence you need.

• Microsoft Invest Shutdown: DSP Migration and Exit Playbook: Microsoft Invest closed 28 Feb 2026. What advertisers lost, a data-portability checklist, the mirror-campaign method, contract clauses, a decision table.
• Bidder as a Service vs White Label DSP vs Self-Hosted DSP: Three ways to run your own DSP without a full build: bidder-as-a-service, white-label DSP, self-hosted DSP. Ownership table, vendors, decision tree.
• How to Build a DSP Platform in 2026: QPS, Team, Cost: How to build a DSP: nine engineering steps from exchange integrations and QPS to pacing, logs and compliance, team and budget ranges, 12-month plan.
• Managed vs Self-Hosted Ad Tech: Which Model Fits You?: Managed vs Self-Hosted Ad Tech: a practical guide to the definitions, operating decisions, measurement and evidence you need.

• DSP RFP Template: 60 Questions and a Scoring Sheet: DSP RFP template: 60 questions on fees, data ownership, log-level data, identity, supply, brand safety, measurement, AI, security and exit, with scoring.
• Ad Tech TCO Calculator Methodology: Costs, Formulas, Sources: How the TCO calculator prices rented, white-label and in-house ad tech: seven cost categories, a formula per line, dated defaults, worked examples.

• The Cost of Renting Ad Tech: Annual Report Methodology: How the annual Cost of Renting Ad Tech report is built: ANA, ISBA, Adalytics, 10-Ks, cloud and labor prices, anonymized platform data, metrics and limits.

1. The Trade Desk 10-K, FY2025
2. Epom DSP pricing, accessed 17 October 2026
3. Beet.TV, 18 December 2020
4. AdExchanger, 14 May 2025
5. Microsoft Advertising, 14 May 2025
6. Basis, 10 September 2021
7. Viant, 4 May 2017
8. BidsCube, 17 July 2026
9. Geomotiv, 5 May 2026
10. ANA Benchmark, Q3 2025
11. CMA, Appendix R, 1 July 2020
12. ISBA, May 2020
13. ANA Benchmark, Q1 2025
14. ANA, 27 May 2026

• October 18, 2026: aligned the search title and heading, expanded the source-linked guide and reviewed its page groups.
• October 17, 2026: assembled this guide from the approved reference pages, retaining their source links and limitations.