TTD AE cheat sheet

The Trade Desk platform organizes accounts hierarchically: partner/advertiser account, then campaigns, then ad groups (line items), then creative and targeting at the ad group level.

Each advertiser account holds its own seat, users, permissions, and billing; a seat can access multiple advertiser accounts but data does not automatically pool across advertisers.

Onboarding requires advertiser name, currency, time zone, and billing setup; currency and time zone drive reporting and budget interpretation, so they must match the insertion order.

User roles separate admin, campaign manager, and reporting-only access; least-privilege assignment is the default for new users.

Data ownership: first-party data (CRM, site tags, offline conversions) belongs to the advertiser and stays within that advertiser account unless explicitly shared.

Agency/partner hierarchy: a partner seat manages multiple advertiser accounts; cross-advertiser reporting requires the partner-level view, not a single advertiser seat.

Pixel and universal pixel deployment is a prerequisite for site-based retargeting and conversion tracking; verify tag firing before launching retargeting campaigns.

Naming conventions applied at account, campaign, and ad group level drive downstream reporting filters; inconsistent naming breaks automated reports and QA.

Campaign holds budget and flight dates; ad groups (line items) hold bid, targeting, creative, and inventory. Budget can be set at campaign or ad group level, not both meaningfully.

Ad group bid types: CPM (fixed), CPC, CPA/CPCV (conversion-optimized), and vCPM; the bid type determines what the bid value represents and how pacing behaves.

Flight dates and dayparting restrict delivery windows; dayparting uses the ad group's time zone, so confirm time zone before scheduling.

Frequency caps are set per ad group or campaign and apply per user per time window; overly tight caps throttle delivery and can cause underdelivery.

Creative assignment: multiple creatives in an ad group rotate; rotation is optimized by the platform unless fixed rotation is selected.

Budget type: daily vs. total (lifetime). Daily budgets can flex within platform rules; total budgets spread across the flight and stop when exhausted.

Conversion tracking must be attached at the ad group level for CPA/optimization bidding; without a conversion pixel the optimizer has no signal.

QA checklist: bid type matches goal, budget covers flight, targeting is not over-restrictive, creatives approved, tracking verified before launch.

First-party segments: CRM onboarding (hashed email/phone), site retargeting (pixel-based), and offline conversion segments; each has different match rates and refresh cadence.

Third-party data segments are licensed and mapped to cookies/IDs; availability varies by region and identity environment, so validate scale before committing budget.

Lookalike modeling expands a seed first-party segment; seed size and quality determine reach and performance, and lookalikes refresh as the seed updates.

Cross-device identity graph links devices to a household or user; match rate depends on deterministic vs. probabilistic signals.

Exclusions: suppress converters from prospecting, and suppress existing customers from acquisition where appropriate; exclusion logic overrides inclusion.

Segment overlap and frequency: layering multiple segments can shrink reach; check estimated reach before launch to avoid underdelivery.

Data onboarding requires hashed identifiers (SHA-256 typical) and consent compliance; unhashed PII should never be uploaded.

Audience refresh: CRM segments update on upload cadence, pixel segments update near real time; stale segments degrade retargeting performance.

Deal types: open exchange (RTB), private marketplace (PMP, invitation-based), preferred deal (fixed price, non-guaranteed), and programmatic guaranteed (fixed price, guaranteed volume).

PMP deals carry a deal ID; the deal ID must be attached to the ad group or bid request to access that inventory.

SPO (supply path optimization) reduces duplicated auction paths to the same impression, lowering take rates and improving win efficiency.

Supply path transparency: prefer direct or short paths to publishers; check reseller chains and fees before enabling a path.

Viewability and brand safety controls (pre-bid filters, allow/block lists, IVT filtering) shape available inventory and should be set before bidding.

Auction types: first-price and second-price; most modern exchanges run first-price, so bid shading or bid strategy matters for efficiency.

Deal priority: guaranteed > preferred > PMP > open, generally; a higher-priority deal can preempt open auction bids for the same impression.

Inventory forecasting: estimate available impressions for a deal or exchange before setting budgets; low forecast risks underdelivery.

Pacing modes: even (spread evenly across flight), ASAP (spend as fast as possible), and front-loaded; pacing controls how budget is released over time.

Bid types map to optimization goal: CPM for reach, CPC for clicks, CPA/CPCV for conversions; the optimizer adjusts bids within the strategy.

Effective CPM (eCPM) = total spend / impressions × 1000; compare eCPM across ad groups to spot inefficiency.

CPA = spend / conversions; target CPA bidding aims to hit the target while spending the budget, and may underdeliver if the target is too low.

Budget pacing formula: expected spend to date = daily budget × elapsed days; variance from actual spend signals over- or under-pacing.

Bid floor: the minimum bid to win an impression; bids below floor never win, so a floor above the max bid causes zero delivery.

Frequency and recency influence pacing: tight caps or short recency windows reduce available impressions and can slow spend.

Optimization requires sufficient conversion volume; low-volume ad groups should use CPM/CPC until data accumulates before switching to CPA.

Attribution models: last touch, first touch, linear, time decay, and position-based; the model determines which touchpoint gets conversion credit.

Conversion window: the lookback period after an impression or click during which a conversion is attributed; windows differ by channel and must be aligned.

View-through conversions (VTC) credit impressions without a click; click-through conversions (CTC) credit clicks; report both separately to avoid double counting.

Reporting dimensions: date, campaign, ad group, creative, audience, inventory source, geography, device; cross-tab carefully to avoid inflated totals.

Metrics hierarchy: impressions → clicks → CTR → conversions → CVR → CPA; each stage exposes where the funnel leaks.

CTR = clicks / impressions; CVR = conversions / clicks; CPA = spend / conversions; ROAS = revenue / spend.

Deduplication: a user may see multiple impressions and click once; attribution rules decide which event counts, so reconcile platform vs. analytics numbers.

Incrementality testing (holdout/control groups) measures lift beyond baseline; platform-reported conversions are correlational, not causal, without a control.

No delivery: check flight dates, budget exhausted, bid below floor, targeting too narrow, creative rejected, or tracking not firing.

Underdelivery: widen audience, raise bid, relax frequency cap, extend flight, or add inventory sources; verify pacing mode is not restricting spend.

Overdelivery: daily budget caps and pacing limits; check for budget set at both campaign and ad group level causing double counting.

Tracking issues: verify pixel fires, conversion tag placement, deduplication settings, and that the conversion event matches the optimization goal.

Creative QA: size, format, landing page URL, and policy compliance; rejected creatives stop delivery for that ad group.

Data discrepancy: platform vs. analytics differences come from attribution windows, cookie/ID loss, and deduplication; document the reconciliation method.

Change log discipline: record bid, budget, targeting, and creative changes with timestamps; unlogged changes make performance shifts unexplainable.

Pre-launch QA: confirm bid type, budget/flight, targeting and exclusions, creatives, tracking, and naming; run a small test before scaling.