Start your holiday ads now

Meta claims 49% of shoppers on its apps plan to spend more this holiday than last year. Read that number for what it is: a self-reported survey result from the platform that sells the ads. It is directional, not a promise. But directionally it points one way. Shoppers intend to spend, advertisers believe them, and auction prices will reflect that belief. Advantage+ needs its learning phase locked in before Black Friday. Switch it on in November and you are paying peak CPMs while the algorithm is still figuring out who to serve.

What the 49% actually tells you

Half of Meta's shoppers planning to spend more is an intent signal, not a receipt. People overstate their future spending the way they overstate their future gym attendance. But intent signals move markets anyway, because advertisers act on them. When every media buyer walks into Q4 believing demand will be strong, budgets follow, auctions get crowded, and CPMs climb before a single dollar of that extra spending actually materializes.

That is the first reason the timing argument in this post matters more than the number itself. The 49% figure is Meta's case for why you should be there. The learning-phase argument is the case for why you should already be there. One is marketing; the other is mechanics. The mechanics do not care whether shoppers spend 49% more or 9% more. They care whether your campaigns are trained when the auction gets expensive.

The second read of the number is competitive. If roughly half of shoppers plan to spend more, your competitors have seen the same stat. The brands that capture that spending are the ones whose targeting, creative, and bidding are already dialed in when the surge hits. Nobody figures out their holiday strategy during the holiday. They execute the one they built in October.

October is the learning period

Learning phases need stable budgets and conversion volume. Starting in October gives the algorithm weeks of signal before the Black Friday auction spike. Starting in November means learning during the most expensive auction of the year. That sentence deserves to sit with you for a moment, because it is the entire post in one line.

Think about what "stable budgets and conversion volume" actually requires. Stable budgets means you are not doubling and halving spend, not pausing and restarting, not rewriting the campaign structure every few days. Every significant edit sends the system back to re-learn. Conversion volume means the algorithm needs a steady stream of purchase events to separate signal from noise. Both of those things take calendar time. You cannot compress October into the first week of November by spending more, because the volume has to arrive at a pace the system can learn from, and the auction prices it out from under you anyway.

October traffic is also cheaper traffic with a second job to do. The shoppers clicking in October are the seed of your November remarketing pools. Every site visitor, video viewer, and engager you collect now becomes someone you can reach again during the peak with a warm message instead of a cold introduction. Starting in November means building that pool at peak prices, from zero.

What the learning phase actually needs

The learning phase is not a button you press; it is a state your campaigns settle into when three conditions hold. First, budget consistency. Pick the spend level you intend to hold and hold it, because erratic budgets teach the system nothing except volatility. Second, structural patience. Resist the urge to restructure ad sets, rewrite creative wholesale, or chase daily performance swings. Let the system accumulate observations. Third, conversion events it can trust. Your pixel, conversions API, and event quality need to be clean before October, not fixed during it. A learning phase built on broken tracking is just expensive noise.

Black Friday is the harvest

The brands that win Q4 are the ones that planted in Q3. Creative testing, audience seeding, and learning-phase completion all happen before the holiday, not during it. The harvest metaphor holds up because farming and media buying share the same rule: you cannot rush the growing season.

Creative testing in October means you walk into November knowing which hooks, offers, and formats actually convert, instead of A/B testing at $40 CPMs. Audience seeding means your remarketing lists, lookalike seeds, and engaged-shopper pools are full before the peak. Learning-phase completion means your bidding is optimizing on real buyer behavior instead of exploring the auction space. Each of these is a Q3 and October activity. Each of them is nearly impossible to do well in November, when every test costs peak prices and every day of learning is a day of the short selling season burned.

The November launch tax

Consider two identical advertisers with identical holiday budgets. The first launches in October. By Black Friday, the learning phase is complete, winning creative is identified, remarketing pools are full, and every dollar of peak-season spend goes toward exploiting a trained system. The second launches in November. Its first weeks of signal arrive exactly when CPMs are at their annual peak, so it pays the highest prices of the year for the least informed impressions of the year. The learning phase completes, if it completes, somewhere around the time the season is winding down.

This is the November launch tax, and it is paid twice. You pay it once in inflated CPMs on every impression, including the exploratory ones that teach the system nothing about your best buyers. You pay it again in opportunity cost, because the days spent learning are days your competitors spend harvesting. The original post said starting in November means learning during the most expensive auction of the year. The worked example just makes the cost concrete: same budget, trained system versus learning system, and the difference is the entire holiday.

What to do this week

1. Confirm your Q4 campaigns are live now, with the budgets you intend to hold. If they are not live, today is the day, not next week.
2. Lock the structure. No more restructures, no more rebuilds. From here to Black Friday, changes should be creative swaps and budget holds, not architecture.
3. Run your creative tests in October. Test hooks, offers, and formats while impressions are cheap, and kill losers fast so winners get the volume.
4. Seed your audiences deliberately. Site traffic, video viewers, social engagers, past purchasers. Every pool you fill in October is a warm audience in November.
5. Verify your tracking end to end. Pixel firing, conversions API deduplicated, purchase events clean. Fix it now, while the data it produces is cheap.
6. Set the measurement standard for the season: revenue and ROAS, not clicks and impressions. Decide the scoreboard before the game starts.

The honest caveats

Three things this post does not claim. First, the 49% is Meta's number, from Meta's survey, on Meta's apps. Treat it as the platform's sales pitch with a kernel of directional truth, not as independent research. Second, an early start does not fix weak creative or a bad offer. A trained algorithm serving a message nobody wants is just an efficient way to lose money. The learning phase optimizes delivery; it does not create demand. Third, auction pressure hits everyone. Starting early gives you an efficiency edge, not immunity. Your competitors read the same playbook. The advantage goes to whoever executes it best, not whoever heard it first.

Are your holiday campaigns live yet, or are you gambling on a last-minute launch?

Data over opinions.