Meta dayparting on daily budgets: how throttling actually works

You want overnight cheaper on a daily-budget Meta ad set. Ads Manager will not give you Ad scheduling. Throttling is the honest second mechanism: dim the daily budget in loss hours, walk it back to baseline when the hour is good, and do not touch it while Meta says LEARNING.

Dimmer, not switch

If you need true OFF (no delivery), switch that one set to lifetime and use the calendar. How-to: Meta dayparting. If you need to stay on daily because that is how the account thinks, throttle is the mechanism. Do not tell yourself it is the same control.

Worked numbers (so this is not poetry)

Prospecting ad set. Daily budget £100. You have flagged 01:00-06:00 as loss hours from 60 days of hourly data. Floor 20% = £20. Max step 20% of current. Throttling spend only helps if you know your break-even ROAS in the first place, which is the whole subject of true profit from Shopify ads.

Walk down (if every hour is still a loss hour and you tick once an hour):

• £100 → £80 → £64 → £51 → £41 → £33 → £26 → £21 → £20

That is the point of the step. You do not teleport to £20 at 01:01. You arrive later in the trough. Walk back up the same way when the hour is no longer flagged. If the ad set enters LEARNING after a creative swap, the next ticks are holds. Budget stays put until learning ends.

Projected overnight "save" is not "we printed money." It is (baseline hourly rate minus floor hourly rate) times flagged hours, and it stays labelled projected until you have a week of real ticks. Do not put a fake ROI on a dashboard.

How to do it yourself (the version that does not rot)

Meta Automated Rules can change daily budget by time. If you insist on DIY:

1. Write baseline £X on a note the buyer and the founder both see.
2. Rule A: in loss hours, decrease budget toward the floor in modest chunks. One giant "set to £20" is how learning gets twitchy.
3. Rule B: in good hours, increase back toward £X. If you only ship Rule A, you will spend next week on the floor.
4. Exclude ad sets in learning if the rule language lets you. If it does not, do not run the rule on new sets.
5. Timezone = ad account, unless you truly know you are using store time. Pick one. Write it down.
6. Review the rule list monthly. Dead rules are how accounts get haunted.

If that list already feels like a second job, that is the desk job. Not a course. A worker that already knows the baseline, the floor, the step, and the learning hold.

What Ralph actually does (no extra magic)

• Reads your Meta hourly history. Suggests loss hours only when the block is confidently below average. Empty data = no suggestion.
• On a daily-budget ad set, the suggestion is marked mechanism: throttle, not native. The UI should say so. applicable: false on native schedule means "this will not be written as adset_schedule."
• After you approve, a tick reconciles current daily budget to either the floor (loss hour) or the stored baseline (good hour), stepped, unless learning says hold.
• Timezone is the store timezone the rest of the desk already uses, not a third clock hiding in a rule.
• Propose → approve → apply. Automation flag off by default. Autopilot starts on suggest and does not write.

Ad Studio builds today ship as daily budget, so creation-time suggestions are evidence plus a throttle label, not a lie that the new ad set has a native calendar. Docs: Ad Studio.

When throttle is the wrong tool

• You need hours fully OFF (legal, brand, or "we do not sell at 3am"). Use lifetime + native.
• The ad set is in learning every other day because you keep rewriting creative. Hold means you will barely move. Fix the launch habit first.
• Loss hours have no spend. There is nothing to dim.
• The leak is an empty size at 20:00. Stock, not the dimmer.
• You want weather to twitch the budget every shower. No. Weather vs the clock.

Throttle that forgets the baseline is just a budget cut you will notice on Thursday.

Fashion, Plus, and noisy accounts

Fashion evenings can look strong and still hide a dead UK 10. Prove hours, then prove the size run. Fashion dayparting. Plus accounts with six locations should not invent a third timezone for the worker. Use the store TZ you already trust. Plus.

If you want the "why hire a desk" page after you understand both mechanisms: why Ralph for dayparting.

Throttle a daily-budget Meta ad set without wrecking learning

1. Prove the loss hours. 28-90 days. Cells need real spend. Write the hours down in the ad account timezone, then confirm store timezone if you will automate.
2. Store the baseline. Today's daily budget is the number you must restore to. Write it somewhere that is not your head.
3. Pick a floor, not zero. 20% of baseline is a grown-up start. Not £1 unless the baseline is tiny. You are dimming, not killing.
4. Step, do not slam. Move about 20% of the current budget per hour toward the floor, then the same on the way back up. Skip writes while the ad set shows LEARNING.
5. Restore in good hours. If you only have a 'set to £20 at 1am' rule and no restore, you will spend Tuesday on the floor. That is how rules rot.

If the hours are the problem: Ralph reads sixty days of your hourly results, proposes the switch-off with floors, and applies it on Meta after you approve. On Google he recommends the schedule. See Ralph · How Ralph dayparts · Docs