--- id: agencias/resultado-meta-por-objetivo industry: Marketing agencies usage: widely used effort: low grain: Ad x day --- # Meta Ads results by objective The cost per result of each campaign measured by its real objective, without double-counting leads and matching Ads Manager. ## Executive summary **For:** Agencies that report paid media results to their clients. **What it solves** - The CPL and cost per result in the report match Ads Manager. - Each campaign is evaluated by the objective it was created for, not by a single lead yardstick. - "Campaign with no results" alerts stop firing for no reason on reach, conversation or sales campaigns. **Why it matters:** A wrong CPL leads to the wrong budget reallocation and, when the client compares it with the CRM, to a loss of trust in the agency. It is also the foundation for every other media case: the funnel down to the sale, alerts and automated reports. **Business questions it answers** - How much did each result cost, by client and by campaign, this month? - Which campaigns are delivering the objective they were created for? - Which client has a cost per result above target? - How many leads came from Meta forms and how many from the website? - Does it make sense to move budget from reach to conversion for this client? **What you need** - Meta Ads connected to Nekt (one account per client, if multi-client) - The objective agreed with each client: lead, conversation, purchase or reach **What you get** - A table by ad and day with result, result type and cost - A client dashboard with a cost per result that matches Ads Manager - An AI agent that answers media questions with the right rule - Alerts only for campaigns that have a conversion to measure ## Detailed ### The problem For each ad and day, Meta returns a list of actions with dozens of types, and the same lead shows up in several of them. Anyone who adds up the types that look like leads reports a CPL at half the real value. Anyone who expects leads from a reach campaign or a WhatsApp conversation campaign fires false alerts about campaigns with no results. ### Business rules 1. **The result comes from the ad set objective.** Read the ad set's optimization_goal and, for website conversions, the optimized event (promoted_object.custom_event_type). Never classify by campaign name or by the presence of some action. (Avoid: Classifying the campaign by its name or by the first action that appears.) 2. **Use the objective to result map.** Each objective has a single action_type that counts as the result. Reach objectives have no conversion to measure. (Avoid: Using lead as a universal result.) 3. **A lead is the lead action_type, and only that.** lead is already the sum of onsite_conversion.lead_grouped (instant form) and offsite_conversion.fb_pixel_lead (website pixel). Use those two to break down by origin, never add all three. (Avoid: lead + lead_grouped + fb_pixel_lead) 4. **Ignore the mirror types.** offsite_search_add_meta_leads, offsite_content_view_add_meta_leads, offsite_complete_registration_add_meta_leads and onsite_web_lead repeat leads already counted. (Avoid: Filtering action_types that contain lead.) 5. **A click is the link_click from the actions.** The insight's native clicks column includes clicks outside the link and does not match Ads Manager. (Avoid: Using the insight's clicks column.) 6. **Each campaign is judged by its own objective.** A reach campaign has no CPL; a conversation campaign has a cost per conversation. Say reach objective, no conversion to measure, never zero leads. (Avoid: A zero leads alert on a reach campaign.) 7. **A conversation started is not a sale.** It is the top of the WhatsApp funnel. Qualification comes from the CRM or the customer service tool. (Avoid: Reporting a conversation as a qualified lead.) 8. **A time on page pixel is not a lead.** If the Lead event fires based on time on page (for example, 20 seconds), the valid result is the form's custom conversion. This is a per-client rule. (Avoid: Trusting the pixel's standard Lead event.) 9. **Declare the tax on the budget.** If the agency passes on tax on the Meta budget (for example, 12.15% or 13.83%), state whether the cost includes tax or not and keep the criterion consistent. (Avoid: Mixing cost with and without tax across reports.) ### What counts as a result Based on the ad set's optimization objective. - **LEAD_GENERATION, QUALITY_LEAD** → Lead (`lead`) - **OFFSITE_CONVERSIONS · LEAD** → Lead (`lead`) - **OFFSITE_CONVERSIONS · PURCHASE** → Purchase (`offsite_conversion.fb_pixel_purchase`) - **OFFSITE_CONVERSIONS · COMPLETE_REGISTRATION** → Registration (`offsite_conversion.fb_pixel_complete_registration`) - **OFFSITE_CONVERSIONS · custom** → Custom conversion (`offsite_conversion.custom.`) - **VALUE** → Purchase, with revenue (`offsite_conversion.fb_pixel_purchase`) - **CONVERSATIONS, REPLIES** → Conversation started (`onsite_conversion.messaging_conversation_started_7d`) - **LINK_CLICKS** → Link click (`link_click`) - **LANDING_PAGE_VIEWS** → Landing page view (`landing_page_view`) - **THRUPLAY** → ThruPlay (`video_thruplay_watched_actions`) - **REACH, IMPRESSIONS, VISIT_INSTAGRAM_PROFILE** → No conversion to measure (`no action_type`) ### Definitions | Metric | Definition | |---|---| | Spend | sum of spend | | Result | value of the action_type for the ad set's objective | | Cost per result | sum of spend / sum of results in the period, dividing only once | | Form leads | onsite_conversion.lead_grouped | | Website leads | offsite_conversion.fb_pixel_lead | | CPL | spend of lead objective campaigns / lead | | Cost per conversation | spend of conversation objective campaigns / conversations started | ### Pitfalls - **CPL at half the real value.** Adding lead, lead_grouped and fb_pixel_lead counts each lead twice. How to avoid: Use only lead for the total. - **Duplicated ad set.** The ad sets table stores more than one version per id and the join duplicates spend. How to avoid: Keep the most recent version of each ad set. - **Average of daily costs.** Averaging each day's cost distorts the monthly number. How to avoid: Sum spend and results and divide once. - **False dropped-to-zero alert.** Delivery, travel and e-commerce clients were measured by leads and triggered alerts every day. How to avoid: Evaluate by the campaign's objective. - **The conversion changes everything.** In a real account, CPA went from R$ 8.39 (lead, purchase and conversation added together) to R$ 90.57 (only the objective's conversion). How to avoid: Agree on the result with the client before the first report. - **Lead Ads without UTMs.** The instant form does not go through the website, so the lead reaches the CRM without UTMs. How to avoid: Join to the CRM by form id or ad id. Validated on real data: In Nekt's own Meta Ads account, on every day with leads, lead was exactly lead_grouped + fb_pixel_lead. Adding all three would double the leads. ### Acceptance test - [ ] For one client and one closed month, each campaign's result and cost per result match the Results column in Ads Manager (difference below 1%). - [ ] Total leads equal form leads + website leads. - [ ] No reach, conversation or traffic campaign shows a CPL. - [ ] The final table has one row per ad and day. ### Context document ```text Meta Ads results - Each campaign's result is the one for the ad set's optimization objective (optimization_goal and promoted_object.custom_event_type). Never classify by campaign name. - Lead = action_type 'lead'. Do not add it to onsite_conversion.lead_grouped or offsite_conversion.fb_pixel_lead; those are its components (form and website). - Ignore action_types *_add_meta_leads and onsite_web_lead. - Click = action_type 'link_click'. - Reach, impressions and profile visit campaigns have no conversion: evaluate them by CPM, reach and frequency. - Conversation campaigns: result = onsite_conversion.messaging_conversation_started_7d. A conversation is not a sale. - Cost per result = sum of spend / sum of results in the period. - If a client's objective is not confirmed, say "objective not confirmed" instead of assuming lead. Per-client exceptions: - : the valid result is the custom conversion ; the pixel's Lead fires based on time on page and does not count. - : cost reported including % tax on the budget. ``` ### Variations - **Multi-client:** One Meta Ads source per client, unified with the account to client mapping. - **Revenue and ROAS:** For VALUE or purchase objectives, use action_values for the same action_type. - **LinkedIn and Google:** LinkedIn: MAX_LEAD counts form leads, without view-through. Google: the primary conversion of the conversion action.