How to Build a Monthly Creative Production Calendar

Count live products, expected cold spend, and how fast frequency usually hits 2.5–3.5 in this account. For each product that spends meaningfully, plan at least one 6-creative batch per week (3 hooks × 2 avatars, one body). Smaller spend can live on a two-week cycle. Write the number of batches first. Themes come after. A calendar that starts with 'holiday vibes' and no batch count will miss inventory.

A simple default: one product, weekly 3 × 2, plus two statics pulled from the winning frames, and one iterate of last week's strong-hook/weak-body cell. That is eight to ten assets, not forty near-duplicates. If you run three products, you do not automatically owe 18 videos a week — you owe a ranked list. Put 70% of production on the product that can actually spend, 20% on the challenger, 10% on experiments. Calendars fail when every SKU gets a ceremonial ad and the hero product goes stale on Friday.

Work backwards from the test window. Each batch needs 48–72 hours of untouched delivery and ~1,000 impressions per variant. If your weekly media cannot fund that for six cells, drop to three hooks on one avatar rather than shipping a starved grid. Write the impression budget next to the batch on the calendar. Production without media is a render queue; media without production is a fatigue clock. The calendar has to show both or it is a content plan in costume, and costume plans miss the 2.5–3.5 band every time.

Give every week a single question: which hook family, which avatar, which proof beat, or which offer. Put that question in the Monday cell. A week named 'always-on plus promo plus UGC plus founder' is four tests colliding. The month can still have a sale week — that week tests offer, and you accept that you will learn less about hooks than you wanted.

A clean month for one hero product might look like: week 1 hook families on avatar A/B, week 2 more hooks on the winning face, week 3 new first frames / B-roll on the winning line, week 4 offer or bundle only if the body still holds. That follows the iteration ladder instead of reinventing the ad every Monday. Seasonal moments get a row, not a takeover, unless the business actually depends on that week. If Black Friday will eat the last week, pull production forward so you are not rendering on Wednesday of the sale.

Map each week to campaign setup: test campaign gets the new 6-pack; scaling campaign keeps last week's graduate; retargeting gets statics and a reminder cut. Write those destinations on the calendar so files do not all dump into one Advantage+ pile with no names. When a founder asks for 'more ads this week,' the theme is the filter: yes if it serves the question, no if it is a seventh near-duplicate. The calendar is a decision tool, not a yes machine. If two questions share a week, pick the one that replaces a tiring open — fatigue does not wait for the promo brief.

Put four named checkpoints on every batch: brief locked (hooks, avatars, kill rule), QC on a phone, launch into the test campaign, readout after 48–72 hours. Assign a person and a weekday to each gate even if it is the same person. Gates without owners slip, and the month becomes a pile of half-launched folders that never get a clean read. A gate with no weekday is a wish, and wishes slip.

A workable rhythm: Monday brief, Tuesday produce and QC, Wednesday launch, Friday readout for the batch that launched the previous Wednesday — because of the 48–72 hour wait, the readout is always looking at last wave, not the one you just shipped. Draw that lag on the calendar so nobody expects Friday numbers on Wednesday's ads. If a gate is missed, the batch moves a week; it does not sneak out un-QCed. Policy rejects at launch are a QC failure, not a platform surprise — add a claims check to the Tuesday gate.

Keep the readout artefacts tiny: a six-row table (hook, hold, CTR, spend, conversions, decision) and the three buckets — kill, iterate, graduate. Paste that table into the following week's brief. The calendar should link to those tables, not hide them in Ads Manager. When frequency on a graduate is climbing toward 2.5–3.5, the readout must spawn a refresh row next week even if CPA still looks polite. Polite CPA is how calendars go quiet until the crash. If the table is not filled on the readout day, treat the batch as slipped, not as 'still learning' by default.

Estimate hours per batch: brief, render, QC, launch, readout. Multiply by the number of batches from step one. If the total exceeds the team's real capacity, cut products or weeks before you cut the no-touch window. An overloaded calendar ships late, and late batches are how frequency cooks the only winner while you are still generating variant four.

Be rude about time. A 3 × 2 AI UGC batch from a locked brief can be a couple of hours including QC; a batch that also needs new handheld B-roll is an afternoon; a client round of 'can we see more options' can double it. Put client approval on the calendar as a gate with a deadline, or do not offer it. Capacity also includes the buyer's no-touch discipline — if they will spend Thursday 'optimising' the test, you do not have a 48–72 hour read, you have a hobby. Reduce batch count until the process is boringly keepable.

Build a 20% slack week into the month: one lighter production day for iterates, static frame pulls, and the emergency refresh when something dies early. Calendars that are 100% utilised break on the first reject or the first viral week that burns frequency faster than planned. Slack is not laziness; it is how you protect the ladder. If you never use the slack, add a challenger product. If you always blow it, you are still overcommitted. Cut a batch before you cut the 48–72 hour no-touch window — a late crowded week is just six noisy stories.

Put the calendar in the same doc as briefs and readout tables — one URL. Colour is optional; columns that matter are week, product, test question, 6-pack status, launch date, readout date, graduate. Review it every Monday for 15 minutes and move anything that slipped. A pretty board nobody opens is not a calendar, and it will not stop frequency from cooking the only winner.

Status values should be ugly and true: briefed, in render, QCd, live, readout done, graduated, slipped. Slipped is a first-class state. Hiding slip by silently stacking two batches in one week is how QC dies. If a week slips, the next week's question moves; you do not 'catch up' by launching 12 cells into the same ad set. Twelve near-simultaneous ads on a modest budget will not reach a clean read. Protect the grid size you set in the volume math, even when someone wants to 'make up for lost time' with a dump of files.

Share a one-screen version with anyone who can derail the month — founder, client, media lead. The point of sharing is to pre-answer 'can we just drop a new offer in.' The calendar says when offer week is, and what you will not mix. Klip Kanvas fits the Tuesday produce gate: one brief, 3 × 2 render, statics from the same product data. Put that sitting on the calendar as a block, not as leftover time after meetings. Production that is not on the calendar does not happen. If the block is not on the week view, assume it will lose to meetings.