How It WorksPricingIntegrationsAffiliatesDocsBlog
Log inGet Started
Back to Docslearn

Optimizing Ad Spend

Which conversion event to optimize each platform toward, when to move deeper in the funnel, and the one metric that keeps ad spend honest for lead generation.

Table of contents

The metric that matters

Cost per lead rewards forms, and forms are cheap to collect and easy to fake. The number that tracks the business is cost per qualified lead, and behind it cost per booked appointment and cost per closed deal.

Once DealOracle sends qualified-lead and closed-deal events back to the platforms (Lead to close), two things become possible: you can read those costs per campaign in Oracle, and you can ask the platforms to optimize toward them.

Choosing the optimization event

Every platform optimizes toward one event per ad set or campaign. Choose the deepest event that still produces enough volume for the algorithm to learn:

  • Meta: the algorithm wants roughly 50 optimization events per ad set per week. A new account starts on Lead, moves to Schedule or Qualified once those pass the threshold, and to Purchase only at real volume. Below the threshold, optimize one level up and let the deeper event feed value and match data.
  • Google Ads: import Qualified and Converted as separate conversion actions, mark the outcome as primary for bidding, and use Maximize conversion value or target ROAS once values are flowing. Enhanced Conversions for Leads improves the match on all of them.
  • TikTok: same principle; SubmitForm for volume, Lead or CompletePayment for quality once the account has history.

Send every event regardless of which one you optimize toward. The platforms use the others for reporting and matching.

Reading Oracle

Oracle joins platform spend with DealOracle's lead outcomes, so each campaign shows spend, leads, qualified leads, appointments, closed deals and revenue in one row, in the account's currency and timezone.

Three views to check weekly:

  1. Cost per qualified lead by campaign. The campaign with the cheapest forms is often the most expensive here.
  2. Revenue by first touch versus last touch. If prospecting looks worthless on last touch and valuable on first touch, it is doing its job.
  3. Leads without a click ID by landing page. Those pages are leaking attribution; fix them before judging the campaigns that send traffic to them.

Common traps

  • Judging a campaign a day after changing its event. Give the platform a full learning period, usually a week and 50 events.
  • Optimizing toward Purchase with 12 purchases a month. The algorithm cannot learn from that. Optimize for Qualified and let Purchase carry the value.
  • Comparing platform ROAS across platforms. Each platform claims overlapping conversions. Compare inside DealOracle, where a lead is counted once.
  • Forgetting the sales team. If Qualified is a manual stage, its quality depends on the people moving cards. Agree the definition before you optimize toward it.