Introduction

Demand-side platforms (DSPs) let advertisers bid on open-web and app inventory through exchanges. In practice, “using a DSP” can mean anything from a self-serve web console with a credit card to a multi-year partner seat on an enterprise stack. Teams that already buy search and social need a clear map of which path matches their budget, ops capacity, and compliance surface.

Three layers of the landscape

Walled gardens (search and social)

Google Ads, Meta, Reddit, TikTok, LinkedIn, and similar platforms combine identity, placement, and conversion APIs in one product. They usually offer the strongest intent (search) or behavioral (social) targeting. They also concentrate policy and payment risk: one disapproval or restriction can block a large share of spend. Portfolio guidance often uses manager structures (for example Google MCC plus child accounts per domain or tight brand group, Meta Business Portfolio with multiple pages and ad accounts) to isolate blast radius rather than pretending one ad account can safely mix many unrelated destinations.

Self-serve open-web DSPs

Mid-market and SMB-oriented DSPs (examples commonly evaluated: StackAdapt, Choozle, Quantcast, Simpli.fi, AdLib-style aggregators) emphasize direct signup, credit-card funding, and campaigns keyed by domain or line item. They diversify where impressions run relative to walled gardens; they do not replicate Meta’s social graph or Google’s search query stream. Industry commentary describes a long-running pattern of high historical minimums on classic enterprise DSPs while lower-friction self-serve products expand access for smaller advertisers; exact floors remain vendor-specific and time-varying.

Seat-based enterprise DSPs

The Trade Desk, Google Display & Video 360, Yahoo DSP, Basis, and similar platforms are frequently reached through agencies or certified partners. Economics often include platform fees, data fees, and minimum media that make tiny card-funded experiments difficult unless spend is consolidated monthly. These tools can be correct at scale; they are a poor default for sub-thousand-dollar learning loops unless a partner agrees to aggregate tests.

Operational implications for multi-product portfolios

  1. One DSP seat versus many domains — A single self-serve account often supports multiple insertion orders or line items tagged by landing URL and UTM. That differs from “one walled-garden ad account for 26 unrelated products,” which can confuse reviewers and reporting.
  2. Attribution — Open-web conversions are typically noisier than pixel-native social. Teams plan first-party leads, geo holdouts, or disciplined UTM hygiene when scaling.
  3. Risks to budget — Invalid traffic, unsuitable placements, and creative policy still apply off-Facebook. Budget for filters, allowlists, and appeals.
  4. Verification discipline — Internal research logs repeatedly note that web research contradicts itself on minimums (for example Choozle-style commitments). Treat every dollar figure in a third-party article as unverified until confirmed in the vendor’s current pricing or contract flow.

Conclusion

Choosing a DSP is less about picking a “best brand” and more about matching access model (self-serve versus partner seat), minimum economics, and measurement reality to the portfolio’s stage. Walled gardens remain central for sharp targeting; open-web DSPs add reach diversification and different policy surfaces—not a duplicate stack.

References

Internal portfolio synthesis (paths and matrices, not reproduced here in full): private business-operations playbooks on DSP evaluation and multi-product resilience.