Programmatic ad ops best practices: the discipline guide
Programmatic ad ops is what happens between "the media plan is approved" and "the report goes to the client." Done well, it's the difference between a campaign that delivers on plan and one that drifts into low-quality inventory, over-paces in week one, and runs out of budget in week three. The discipline is unglamorous and high-leverage.
The daily cadence
Every working day, an ops operator checks five things per active campaign:
- Pacing. Is each line item on track to deliver its planned volume by end date? Flag any line over- or under-delivering by more than 10% on a single day; over 15% by mid-week requires action.
- Spend. Has any line burnt through its daily budget in the first few hours? This usually indicates frequency-cap settings too loose or a runaway bid strategy.
- Quality alerts. Has the verification vendor (DV, IAS, Moat) flagged any IVT or brand-safety incidents? Investigate and exclude as needed.
- Creative. Are all creatives serving? Rejected creatives, expired tags, or DCO failures show up as zero-impression line items.
- Errors. DSP error log, ad-server discrepancies, attribution model failures — quick scan, escalate only if pattern repeats.
The weekly cadence
| Activity | Goal |
|---|---|
| Placement report | Identify long-tail placements eating budget; exclude or whitelist |
| Audience performance | Compare audiences against each other; reallocate budget to winners |
| Creative performance | Pause underperforming creatives; queue refreshed variants |
| Frequency analysis | Verify frequency caps holding; adjust if average freq exceeds plan |
| Pacing forecast | Project end-of-flight delivery; intervene if forecast misses plan |
| Reconciliation | Compare DSP-reported delivery to publisher-side delivery; resolve discrepancies >2% |
The quarterly cadence
- Renegotiate PMP and PG deal terms based on performance and pacing data.
- Re-evaluate supply-path optimization rules. SSP relationships shift; what was optimal six months ago may not be now.
- Audit verification vendor settings: IVT thresholds, viewability floors, brand-safety categories. Industry standards drift.
- Review fee load across the supply chain. ANA-style transparency reporting from agencies is increasingly standard; if you're not getting it, demand it.
- Document what worked and what didn't — the institutional knowledge that prevents repeating mistakes.
What separates elite ad ops teams
Three patterns we see in the best ops teams (and the absence of which we see in struggling ones):
1. The runbook discipline. Every campaign, every deal, every seat is documented with enough detail that a new operator could pick it up cold. Deal IDs, seat IDs, contact people, decision rationales, known gotchas. This sounds bureaucratic; in practice, the alternative is institutional knowledge that walks out the door when a key operator quits.
2. The reconciliation rigor. DSP reporting, SSP reporting, publisher reporting, and your ad server's reporting will disagree. Elite teams reconcile weekly, isolate the discrepancy sources, and escalate ones that exceed thresholds. Mediocre teams accept the DSP number and ignore the others.
3. The pre-bid configuration. Brand safety, IVT, viewability, and inclusion/exclusion lists are configured at launch and audited quarterly. Mediocre teams configure them once and forget. Elite teams treat them as living configurations.
The metrics that actually matter
| Metric | What it tells you | Healthy range |
|---|---|---|
| Viewability (MRC) | Was the ad rendered viewably? | 70%+ for display, 80%+ for video |
| IVT (Sophisticated) | Was the impression sourced from a bot or non-human? | Under 1% |
| Brand-safety pass rate | Did the inventory pass your suitability filters? | 95%+ |
| Working media ratio | How much of your spend reached publisher inventory? | 55-70% for filtered programmatic |
| Pacing index | Actual delivery / planned delivery, normalized to time elapsed | 0.95-1.05 |
| Frequency average | Impressions per unique user | 2-5 for prospecting, 5-10 for full-funnel |
| DSP-to-publisher reconciliation gap | Difference between DSP-reported and publisher-reported delivery | Under 2% |
Common ops failures
Frequency cap set at line-item level only. The cap fires per line; users see your ad through every line independently, multiplying the actual exposure. Set caps at campaign and audience level too.
No inclusion list. Exclusion lists are commonly maintained but inclusion lists aren't. The result: budget bleeds into long-tail inventory that you've never proven works. Build inclusion lists from the placement report and bid more aggressively on proven supply.
Treating the DSP as the truth. The DSP knows what it bid and won. It doesn't know what the publisher served (sometimes the creative is blocked at the ad server) or what verification flagged after the fact. Cross-reference at least monthly.
Letting deal IDs go stale. A PMP deal that hasn't delivered in 30 days usually has a configuration problem: floor too high, audience filter too narrow, or DSP-side line item disabled. Audit dormant deals.
Skipping incrementality testing. Last-click attribution in programmatic is systematically biased toward retargeting and against prospecting. Run quarterly geo-holdout or conversion-lift tests to know what's actually causing conversions.
How big a team do I need for ops?
For a single advertiser running $1-3M/month programmatically, one to two full-time ops people. For agencies running multiple advertisers, the ratio is usually 8-15 campaigns per ops person. Below that and you're under-utilizing; above and you're under-monitoring.
Should ops sit with the buyer or the agency?
Either works. In-house ops teams know the brand intimately and can move faster on quality decisions. Agency-side ops teams have broader benchmarks and platform expertise. Best is usually a hybrid: agency ops as the operating layer, in-house oversight on strategy and quality.
What tools does an ops team need?
DSP UI(s), verification platform (DV, IAS, Moat), ad server (CM360 or equivalent), reporting layer (data warehouse — Snowflake/BigQuery — plus a BI tool like Looker or Tableau), and ideally a workflow tool for trafficking checklists. The DSP's native reporting is rarely sufficient at scale.
How do I detect supply-chain fraud?
Cross-reference sellers.json (each SSP's authorized account list) and ads.txt (each publisher's authorized sellers). Discrepancies — paths your DSP took that don't match the publisher's authorized chain — are red flags. Most major DSPs run this verification automatically; for the long tail, manual spot-checks remain useful.
What's the right balance between automation and manual ops?
Automation for bidding, frequency capping, viewability filtering, and pacing alerts. Manual for placement decisions, creative rotation, deal negotiation, and incident response. Trying to automate everything produces a campaign that's nominally optimized but missing the judgment calls.
How does ops change in a server-side-tagging world?
The operational layer doesn't change much. The integration layer does — the ops team needs to coordinate with engineering on server-side tagging, conversions API setups, and clean-room integrations. Ops teams that absorb some of this integration work add disproportionate value.
Operating checklist
- Define campaign objective, KPI, and incrementality measurement plan before launch.
- Onboard first-party data and confirm match rates with platform partners.
- Configure pre-bid filters: IVT, viewability floor, brand-safety thresholds.
- Apply supply-path optimization to suppress duplicate auctions.
- Set frequency caps at user, campaign, and audience level.
- Reconcile DSP, SSP, and ad-server delivery weekly.
- Document seat IDs, deal IDs, contacts, and decision rationales in a runbook.