Steering automation: Advantage+, Smart+, Accelerate and Performance+

0:00 Steering automation

If you learned paid social a few years ago, your job was a bit like a sound engineer at a huge mixing desk. Dozens of sliders: ages, interests, placements, bids, budgets. Today most of those sliders have been replaced by a single switch labelled automation. And a lot of marketers feel like they've lost control. In this lecture you'll learn what automated campaigns actually do behind the scenes, the five inputs you still control, how Meta, TikTok, LinkedIn, Pinterest and Snapchat automate differently, a launch checklist you can use tomorrow, and how to test automation properly instead of trusting it blindly.

0:44 Operator → pilot

Here's the mental model. You've moved from operator to pilot. A pilot doesn't flap the wings. A pilot sets the destination, checks the fuel, sets a few hard limits like altitude, and monitors the instruments independently. With automation, the destination is your optimisation goal. The fuel is your conversion data. The limits are a handful of constraints like location and minimum age. And your independent instruments are your store data and tests. Here's the key idea: automation is very good at getting you exactly where you told it to go. So if the destination is wrong, or the fuel is dirty, it gets you to the wrong place very efficiently.

1:32 Three loops

What does the automation actually do? Three loops, for every single impression. First, retrieval: out of millions of ads, it shortlists which ones could be shown to this person. The more genuinely different your creative, the more ways it can match different people. Second, prediction: it estimates how likely this person is to complete your event, and with value optimisation, how much that's worth. Third, auction and pacing: it bids and spreads your budget across placements and time. And it learns from the conversions you send. So if your purchase event includes refused cash on delivery orders or test orders, it optimises towards those.

2:17 Five inputs you control

So what do you still control? Five inputs. One, the goal and event: the deepest event with enough volume, and value optimisation when your purchase values are accurate. Two, conversion data: pixel plus conversions API, deduplicated, with good match quality, plus CRM events for leads. Three, creative: eight to fifteen genuinely different concepts, refreshed regularly. Four, hard constraints: location, minimum age, language only if truly needed, brand safety settings, and caps on spend reaching existing customers. And five, budget and bid: enough budget to learn, and cost or ROAS goals only once you know what's realistic.

2:59 Platforms (Sept 2026)

Now platform by platform, as of September twenty twenty-six. Meta's Advantage plus sales, leads and app campaigns switch on automated audience, placements and campaign budget by default. Extra audience inputs act as suggestions, and you can define existing customers and cap their budget. TikTok's Smart plus automates targeting, bidding, placements and budget, with module-by-module controls and Symphony creative tools built in, while GMV Max optimises TikTok Shop sales. LinkedIn's Accelerate builds campaigns from your website and goal. Pinterest's Performance plus automates bidding, budget, targeting and formats. And Snapchat's Smart campaign solutions cover bidding, budget and creative. Features change often, so check each help centre.

3:44 Example 1: Karachi electronics (illustrative)

Let's do a simple worked example with illustrative numbers. A Karachi electronics store ran eight interest-based ad sets on Meta. Cost per order hovered around two thousand four hundred rupees, and every ad set showed a learning limited warning. They moved to one Advantage plus sales campaign with twelve genuinely different concepts, and capped spend on existing customers at fifteen percent. Crucially, they started sending a delivered event from their server, so the system learned from paid cash on delivery orders rather than refused ones. Three weeks later, cost per delivered order was lower and stable, and the team spent its time making new concepts instead of fiddling with audiences.

4:32 Example 2: UK beauty on TikTok (illustrative)

Now a realistic scenario on TikTok. A UK beauty brand sells through TikTok Shop and its own website. It runs GMV Max for shop sales and a separate Smart plus campaign sending traffic to its website. Why separate? So it can see which channel produces profitable customers. Both campaigns are fed with creator content, often as Spark Ads. And here's the discipline. It judges GMV Max on contribution after affiliate commissions, platform fees and discounts, not on gross merchandise value, because gross sales can look fantastic while margin quietly disappears. That one habit changes which campaign gets more budget.

5:15 Launch checklist

Before any launch, use a checklist. Pixel and conversions API live and deduplicated, with correct purchase value and currency. Match quality checked and a test purchase visible exactly once. Catalogue healthy, if you use one. Objective set to sales, optimising for purchase. Location and minimum age set. Existing customers defined, with a budget cap decided. Eight to fifteen distinct concepts. Creative enhancements reviewed one by one, switching off anything you can't approve. A weekly budget of roughly fifty times your target cost per acquisition. UTMs and naming in place. And decision rules written down. The full checklist is in the lesson text.

5:59 Test it, don't trust it

Don't choose automation on faith. Test it. Run the automated campaign against your manual setup with the same objective, the same creative, similar budgets and the same dates, using the platform's A/B test feature so audiences don't overlap. Then judge on store-verified cost per new customer or contribution, not platform ROAS alone. Why? Because automated campaigns can drift towards existing customers, who look like cheap conversions but would have bought anyway. And watch the classic mistakes: broken tracking at launch, too few similar creatives, treating suggestions as hard targeting, judging after two days, and editing during the learning phase.

6:42 Measure success

How do you measure success with automation? Three things. Store-verified cost per new customer and contribution per order, week over week. The share of spend and results going to new versus existing customers. And creative diversity: how many distinct concepts are live, and what share of spend the top ad takes. If one ad has eaten nearly all the spend for weeks, the system is telling you it's running out of options, so add new concepts. For advanced control, like value rules, margin-based bidding and incrementality testing, continue to the AI Performance Marketing course once you finish this one.

7:25 Recap and try this now

Let's recap. Automated campaigns retrieve, predict and bid for you, and they learn from whatever you send them. You're the pilot: set the right goal, feed clean conversion data, supply genuinely different creative, set a few hard limits, and check results against your own store data. Meta, TikTok, LinkedIn, Pinterest and Snapchat automate different parts, so check the current controls. Here's your try this now. Take the launch checklist from the lesson text and run it against an existing account or a campaign you're planning. List every unchecked item, the fix, and who owns it.