Buyer's Guide

Eight steps to buying a paid-media finance platform without regret

We don't expect every reader to buy us. We do expect every reader who has $20M+ of annual paid media to insist on this level of rigour from any vendor they consider. Including us.

Steps

Sequenced for a 6-week eval

Roles involved

CFO · CISO · Procurement

And one ops sponsor

Artefacts produced

Memo, RACI, runbook, others

Bias warning

We wrote this

Cross-check with two other vendors

Define the problem in money, not in tools

Owner: CFO + VP Finance

Ask

What is our annualised pacing variance in basis points?
How many hours per month does the close take on paid media?
What is our FX leakage as a share of cross-border spend?

Trap

Buying a dashboard because it looks nice; the unit of measure is money, not screens.

What good looks like

A single-page memo with a euro number next to every pain point. If you can't quantify, the project isn't ready.

Separate operations from finance from compliance

Owner: CFO + CISO + Head of Performance

Ask

Where does ad operations end and finance begin in our org?
Who is accountable when pacing variance becomes a write-off?
Which controls are SOX-relevant?

Trap

Treating it as a marketing tool; budgets bleed into finance the moment a controller signs.

What good looks like

A RACI showing finance as accountable for outcomes, ops as responsible for inputs, security as consulted.

Insist on read-only proof before any write access

Owner: Security + Procurement

Ask

Can we run the platform in monitoring-only mode for 30 days?
What artefacts will we see during that period?
How is monitoring different from active mode in your architecture?

Trap

Vendors that require fund custody on day one; that's a multi-year decision squeezed into a pilot.

What good looks like

A clear monitoring tier with no write authority, followed by a graduated activation plan with kill switches.

Inspect the AI, don't take the demo's word for it

Owner: CISO + Data Lead

Ask

May we see the model cards for every AI feature?
What is the headline metric vs. a stated baseline?
Where is the human-in-the-loop boundary, written down?

Trap

Demos that hide auto-execute behind a fancy chart. The chart is not the policy.

What good looks like

Published model cards (see /model-cards), materiality thresholds in writing, and a recent independent audit.

Verify exit before entry

Owner: Procurement + Counsel

Ask

What is the data export format, and what is the SLA?
What is the exit-assistance period and the cost?
Can we re-ingest exports into our own warehouse?

Trap

Proprietary formats that look open in a demo but are sparse in practice.

What good looks like

Open standards (Parquet, ISO 20022, DATEV), 24-hour export SLA, 90-day assistance at standard rates.

Stress-test independence and governance

Owner: Audit committee

Ask

Who audits the vendor, and who advises them?
Is there a published trust ledger of decisions?
Are kill switches drilled, with public results?

Trap

Self-attestation only. Real assurance is signed by parties with their licences on the line.

What good looks like

External audit firms named, ledger published, drill results disclosed quarterly.

Price for outcomes, cap for downside

Owner: CFO + Procurement

Ask

Is pricing tied to spend, savings, or seats — and why?
What's the cap? Is there a true minimum, and is there a true maximum?
What is never charged for?

Trap

Open-ended percentages with no cap; that's a tax, not a price.

What good looks like

A published pricing philosophy with explicit caps and an explicit list of what is never billed.

Run a 14-day implementation, not a 14-week one

Owner: Implementation lead

Ask

What does day 1, day 7 and day 14 look like in writing?
What is the kill criterion before day 14 if it isn't working?
Who owns the runbooks afterwards?

Trap

Long, opaque implementations that consume the political capital you need for adoption.

What good looks like

A written 14-day plan with go/no-go gates and customer-owned runbooks at the end.