Retail Media Build vs Buy: Compare an Intelligence Layer
Compare building, buying and augmenting your retail media stack. Use a three-year cost model covering integration, operations, fees and decision ownership.
Retail media: build, buy or add an intelligence layer?
The useful decision is which capabilities your organisation should own and which it should source. Compare the full operating cost and control of each option against the same requirements. Keep the current system as a credible baseline, and treat any forecast revenue improvement separately from the cost calculation.
Start with the required decisions
List the capabilities the business needs: campaign management, advertiser access, eligible product selection, auctioning, pacing, serving, measurement and reconciliation. For each, identify the existing owner and the problem that justifies change.
A broad platform replacement may be appropriate when several essential capabilities are missing. A focused addition may be better when the commercial workflow works but ranking or another bounded decision needs improvement. If the current system meets the requirements, keeping it remains an option.
Compare three operating models
Option — When it may fit — Responsibility that remains
Build — Distinctive requirements and an enduring engineering commitment — Product ownership, infrastructure, reliability, maintenance and operational tooling
Buy — A supplier can meet most requirements with an acceptable contract and integration — Vendor management, data quality, configuration and commercial operations
Augment — A specific gap can be addressed through stable interfaces — Cross-system ownership, reconciliation, integration maintenance and fallback
Industry procurement guides, including Kevel's build-versus-buy discussion, reflect different commercial perspectives. Assess the architecture and contract, rather than accepting the label as a complete description.
Put costs on the same basis
Use a three-year view with one-time implementation plus recurring costs for each year. Include engineering, hosting, licences or usage fees, AdOps, support, measurement, security review and expected change work. Avoid counting the same person's time in both an internal team line and an outsourced service charge.
For augmenting, show both total retained-stack cost and incremental project cost. The existing platform is not free simply because its invoice already exists. Conversely, a sunk historical build cost should not distort a forward decision.
PA describes modular, usage-metered pricing; a scoped quote is required for a financial comparison.
A synthetic cost comparison
Assume a fixed three-year horizon, no discounting and comparable required scope. An internal build costs £120,000 upfront plus £150,000 annually: £570,000. A purchase costs £40,000 upfront plus £75,000 vendor fees and £35,000 internal operating cost annually: £370,000.
An augmented stack costs £30,000 upfront plus £70,000 for the retained stack, £35,000 for the additional service and £25,000 for internal operations annually: £420,000. These are invented inputs to demonstrate the worksheet, not PA prices, competitor prices or an industry benchmark.
Compare value and risk separately
Model delayed launch, uncertain demand and future volume as sensitivities. Do not quietly assume that every option earns the same media revenue or that the highest forecast is proven. Ask what evidence would justify each difference.
Before selecting an option, agree who owns decisions, how data leaves each component, what happens during a failure and how the organisation can change direction later. PA's modular-adoption page provides an entry point for scoping a bounded deployment.
Illustrative example
SYNTHETIC GBP,3 years,no discounting/tax: build=120000+3*150000=570000; buy=40000+3*(75000+35000)=370000; augment=30000+3*(70000+35000+25000)=420000. No revenue, profit or PA/competitor pricing claim. Augment annual costs include retained stack, overlay and internal operations.
Downloads
- pa-retail-media-build-buy-augment-tco.xlsx (illustrative template)
Sources
- How Pricing Works, checked 2026-10-25
- Start With One Component, checked 2026-10-25
- Retail Media Build vs Buy, checked 2026-10-25
How Particular Audience does it
Frequently asked questions
Is buying always cheaper than building?
No. Cost depends on scope, traffic, internal capabilities, contracts and the required degree of control. Compare the same requirements over a realistic operating horizon, including maintenance and change work. A synthetic cost example illustrates the model; it does not determine the answer.
When is an intelligence layer a useful option?
When a specific decision can be improved without replacing valuable systems, and the necessary interfaces are available. The case weakens if integration creates unclear ownership, duplicate costs or unreliable measurement. Establish a bounded pilot and retained-stack responsibilities before committing.
Should forecast uplift be included in TCO?
Keep it separate. Total cost of ownership measures costs. A business case can then combine those costs with explicitly assumed benefits, ramp-up and uncertainty. Blending unproven uplift into a cost number makes competing options harder to assess.
Related guides
- Improve retail media relevance without replacing your ad server: Assess a retail media relevance layer around your current ad server. Define data inputs, decision ownership, fallback behaviour and a measurable pilot.