The Facebook pixel done right: events, conversions, lookalikes

Facebook retired the separate conversion tracking pixel on February 15 this year. There is now a single pixel per ad account (you can create more through Business Manager, but one is right for almost everyone) that handles conversion tracking, optimization, custom audiences and dynamic ads. It is installed as a base code snippet in the head of every page, and it reports actions through events: nine standard events Facebook understands natively, and custom events you name yourself. Custom conversions sit on top, letting you define a goal from a URL rule or an event without touching code again.

1. Create the pixel in Events Manager (Ads Manager menu, Pixels). Name it after the business, not the campaign.
2. Install the base code on every page, ideally through Google Tag Manager so marketing can manage events without a developer for each change. If you are on Shopify, paste the pixel ID into Online Store preferences and it will fire PageView, ViewContent, AddToCart, InitiateCheckout and Purchase automatically. WooCommerce and Magento have official and third-party plugins that do the same.
3. Install the Facebook Pixel Helper Chrome extension and browse your site. Every page should show a green PageView with your pixel ID. Red or yellow warnings mean the code is duplicated, firing before the page loads or missing parameters.
4. Check Events Manager after 24 hours. You should see PageView counts that roughly match your analytics sessions. A big gap usually means pages missing the code, usually checkout, blog or landing page templates built separately.
5. Set the pixel's sharing and permissions in Business Manager so an agency or partner can use it without owning it. Never let a vendor install their own pixel as your primary one; you will lose the data when the relationship ends.

Standard events are what let Facebook optimize delivery and build dynamic ads. Fire them with parameters wherever possible, especially value and currency.

The remaining three, Search, AddToWishlist and AddPaymentInfo, are worth adding for stores that want full-funnel reporting, but the six above cover most businesses. For a lead-generation site, fire Lead on the thank-you page or on form success, not on the button click, or you will count failed submissions.

Give Lead a value even if it is an estimate. If an average lead is worth $80 to you, say so. It lets you compare campaigns on return rather than on volume, and later this year Facebook is rolling out value-based lookalikes that depend on it.

A custom conversion is a rule: "count a conversion when the URL contains /thank-you-demo" or "when the Purchase event fires with value over 200". They are quick to create, need no code, and are the only way to track goals on pages where you cannot edit the template.

Rules we apply:

• Keep the list short. You get 40 per account and they cannot be deleted, only archived.
• Name them with the site, the action and the date: "Main site - Demo request - Mar17".
• Prefer an event rule over a URL rule when the event exists. URL rules break when the thank-you page changes.
• Assign a category and a value so the conversion appears correctly in reporting columns.
• Test each one by completing the action yourself and watching it appear in Events Manager.

The pixel is only useful if your ad sets use it. Under Conversions objective, pick the event closest to revenue that still gets roughly 50 conversions per ad set per week. Below that, Facebook's delivery system does not have enough signal and performance becomes erratic.

The usual ladder: optimize for Purchase if you have the volume; if not, for InitiateCheckout or AddToCart until Purchase volume supports it; for lead businesses, optimize for Lead, and only drop to Landing Page Views or Link Clicks for brand-new accounts with no history. Move up the ladder as volume grows. Many of the accounts we see are stuck optimizing for clicks years after they had enough purchases to do better.

Set the conversion window deliberately. The default attribution is 28-day click and 1-day view. For a considered purchase that is reasonable; for an impulse product, judge on 7-day click and compare.

Custom audiences are where the pixel earns its keep a second time.

Website custom audiences let you retarget by behavior: visited a product page but did not purchase in 14 days, started checkout in the last 7 days, visited pricing twice. Build those three and run creative that speaks to each stage. Keep frequency sensible; a 7- to 14-day window at a few impressions a day is enough.

Customer file audiences take a hashed upload of your customer list, with email, phone and address fields improving match rates to somewhere between 40 and 70 percent.

Lookalike audiences build an audience that resembles a seed. The seed is everything. A lookalike of all website visitors resembles people who click ads. A lookalike of purchasers resembles people who buy. Our order of preference for seeds:

1. Purchasers in the last 180 days (ideally 1,000 or more people; Facebook will work from 100 but quality suffers).
2. Top 25 percent of customers by order value, from a customer file.
3. Leads that became customers, uploaded from the CRM.
4. Add-to-cart or InitiateCheckout if purchase volume is too low.

Start with a 1 percent lookalike in your main country, then test 2 to 3 percent as you scale. Broader than 5 percent is close to broad targeting. Refresh seeds quarterly; they do not update themselves when built from a file.

Review these monthly:

• Pixel Helper on five random pages, including a product, checkout and a landing page.
• Events Manager event counts versus your platform's own numbers. Purchase events should be within roughly 10 to 15 percent of actual orders; a wider gap means missing templates or duplicated firing.
• Diagnostics tab warnings about missing parameters or invalid currency codes.
• Reconcile reported revenue against your store or CRM at least quarterly. Facebook's attribution is generous; your bank account is not.

Two pixels on the same page. Usually a theme setting plus a plugin plus GTM. Everything doubles.

Purchase fired on the checkout page, not the confirmation page. Conversions inflate and optimization learns from the wrong people.

No value parameter. You can report on purchases but not on return.

Lookalikes seeded from all visitors. Cheap reach, poor buyers.

Optimizing for clicks with thousands of purchases in the account. The most expensive mistake on the list.

Nobody owns the pixel. It sits in a former contractor's personal ad account.

Measurement is the foundation of every growth program we run, and the pixel is part of a connected setup with analytics, the CRM, the store platform and the brand's creative plan rather than a standalone tag. Discover audits the current installation page by page, reconciles reported conversions against real orders and reviews who owns the assets. Plan defines the event map, the custom conversions, the audience architecture and the optimization ladder for each campaign goal. Build implements it through Tag Manager or the platform, tests every event and documents it. Improve reviews data quality monthly, refreshes seeds, moves ad sets up the optimization ladder as volume grows and feeds the results back into creative and budget decisions.

The ecommerce growth engagement is a representative pattern. A store was running Facebook campaigns optimized for link clicks with a pixel that fired Purchase on the checkout page and carried no values. We reinstalled through Tag Manager, mapped the full standard event set with values and product IDs, built staged retargeting audiences and seeded lookalikes from six months of purchasers. With ad sets re-optimized for Purchase, cost per order fell substantially within two months, reported revenue finally reconciled to the store's own figures, and dynamic product ads became possible for the first time.