There is a version of performance marketing that looks like this: check last month's numbers, apply a percentage, call it a forecast. It is fast, it is simple, and it is almost entirely useless.

And then there is the version that actually works.

As a growth team, our fundamental responsibility is to provide every brand we work with a forecast and budget anchored to the performance metrics that make customer acquisition and retention scaleable and sustainable. Not directionally accurate. Not in the right ballpark. Structurally sound - built from the data up, calibrated to how the business actually operates, and flexible enough to absorb the unexpected without falling apart.

That distinction matters more than almost anything else in this industry. What follows is the framework we use to build it.

The problem with most forecasts is that they are built backwards. Someone sets a revenue goal, works backwards to a spend number, and calls it a plan. There is no architecture underneath it - just a number dressed up as a strategy.

Good forecasting requires you to build from first principles. That means four inputs working together:

Historical analysis and current trajectory. Where has the brand actually been? Not just topline revenue, but MER, nMER, ACoS, CAC by channel, and retention rates by cohort. The trend is the truth. If CAC has been climbing for three months and nobody has addressed the structural reason, no forecast will paper over it.

Inventory position. The most consistently underweighted input in growth forecasting. On-hand stock, restock timing, and new product drops are not logistics details - they are demand levers. The inventory position directly sets the ceiling on how aggressively you can run paid during any given period, and ignoring that ceiling is one of the most expensive mistakes a brand can make.

Marketing moments. Launches, promotions, sale periods, gifting windows - each has a distinct demand shape that should be modelled separately. Every moment should carry its own day-by-day forecast informed by the prior year read of that specific event type.

The desired future state. Forecasting is not purely a reflection of the past. It has to account for where the business is trying to go, and whether the channels supporting it are structurally capable of getting there at the target efficiency. Forecasts that are ambitious on revenue but unrealistic on the MER or nMER required to get there are not forecasts - they are wishful thinking with a spreadsheet attached.

Most brands think about inventory as a fulfilment problem. The growth team should treat it as a media planning problem.

The principle is simple: the amount of paid spend you can productively deploy against a product is directly constrained by how much stock you have available to fulfil the demand you generate. A brand running $50,000 per day against a product launch that runs out of stock on day two has not spent $50,000 building the business - it has spent a large portion of that burning paid budget on a product it cannot sell.

Reading the demand signals before a moment opens

Before any significant spend period, three signal sources should inform both inventory planning and spend ceiling estimates:

Platform search and browse behaviour. Rising click-through rates on existing campaigns, increasing product page dwell time, and growing add-to-cart rates in the weeks before a moment are early indicators of building demand. These signals tell you whether the audience is primed or whether you will need to do more work to generate intent. If CTR is declining heading into a promotional period, that is a signal to check whether the creative is tired or the audience is saturated - both of which affect the spend ceiling.

Historical sell-through rates by moment type. If you know that your last three promotional events sold through 60-70% of the promoted SKU's available stock within 48 hours, that sell-through rate should directly anchor your inventory request for the next equivalent moment. You are not guessing - you are indexing. The goal is to arrive at a moment with enough stock to sustain the full planned spend window, not just the opening day.

Email and owned channel engagement. The click-through rate on pre-moment emails and the segment of engaged subscribers (opened in L30, clicked in L90) who are actively browsing tells you the quality of the warm audience you are about to amplify with paid. A large, highly-engaged email list heading into a drop is a multiplier on paid efficiency. A disengaged list means paid is doing heavier lifting, which should be reflected in a more conservative spend ramp.

Setting the spend ceiling from stock depth

The practical application is straightforward. Before each moment, the media plan should include an explicit stock depth check: how many units are available across the promoted SKUs, what is the expected conversion rate at the planned traffic volume, and does the available inventory match or exceed the demand the media plan will generate?

If it does not, one of two things needs to happen: either the inventory request goes back to the buying team with a data-backed case for a deeper stock position, or the media plan is trimmed to match the available stock. Running out of stock mid-campaign is not just a lost revenue problem - it is an algorithm disruption problem. Pulling back spend mid-flight because stock has run out forces the platform algorithm to re-learn from a smaller audience, and you pay for that inefficiency in the days that follow.

The single biggest forecasting edge most brands are sitting on and not using is their own historical moment data.

Not the total. The shape.

Every promotional moment - a sale, a product launch, a collab, a gifting period - has a day-by-day revenue curve that tells you how demand builds, peaks, and decays. That curve is specific to your brand, your audience, and your moment type. It is more valuable than any industry benchmark because it is built from your actual customer behaviour, and it repeats with surprising consistency year over year when the moment structure is comparable.

What to track for every marketing moment

For every significant moment you run, the post-event record should include:

• Day-by-day revenue vs. plan for the full window, including the 3-5 day pre-moment shoulder and the 5-7 day post-moment tail

• Spend by channel per day, and the MER, nMER, and ACoS achieved at each spend level

• New customer vs. returning customer revenue split, and the CAC achieved on new customers during the moment

• Audience segment performance: what did cold, engaged, and repeat audiences contribute individually, and at what efficiency?

• Creative performance by format and funnel stage: what drove volume, and what drove efficiency?

• Inventory position at open, at peak, and at close - including any stockout events and the revenue impact

• The variance between plan and actual, and the identified reason for any material gap

This is not a retrospective exercise. It is the raw material for next year's forecast.

Indexing forward from prior year shape

When you are building the forecast for an equivalent moment this year, the prior year curve is your starting point - not your ceiling or your floor. You index forward from it based on three variables: the current trajectory of the business (is the brand growing, flat, or declining relative to this time last year?), the media environment (have CPMs shifted materially, has platform efficiency changed?), and the moment structure (is this year's event comparable in depth, duration, and promotional mechanics?).

If last year's promotional event delivered $400K over 5 days with a 15% discount and the brand has grown 20% in the intervening period, the prior year shape gives you the day-by-day distribution and the prior year total gives you the baseline. The growth index and any structural changes to the moment layer on top.

This is also where the post-event variance analysis from the prior year becomes critical. If last year's event underperformed on day three because a key creative fatigued overnight, that is a production and scheduling insight for this year - not just a historical footnote. If it overperformed on the final day because of a last-chance email to the non-purchasers, that is a playbook element to build in by default.

Most brands treat the marketing calendar as a scheduling tool - a place to log what is happening when. The growth team should treat it as the primary strategic document for the year.

The distinction matters because a scheduling calendar is reactive. A strategic calendar is designed. The difference between the two is whether the calendar answers three questions:

• Does every moment on the calendar have a purpose beyond filling space - a specific commercial objective, a target customer segment, and a measurable outcome?

• Is the spacing between moments intentional, with explicit plans for how the whitespace periods are used to rebuild retention metrics and reduce audience fatigue?

• Does the paid media calendar align with the email and content calendar so that each channel is amplifying the others rather than running independently?

One of the most underused levers in fashion and apparel is the manufactured moment - a demand event that is created by the brand rather than dictated by the retail calendar.

Black Friday exists. You will be there. But you are competing with every other brand for the same audience attention, the same CPM pool, and the same promotional-discount-conditioned customer. The brands that grow most efficiently are not the ones that show up loudest in November. They are the ones that create three or four high-conversion moments through the year that are entirely theirs.

What makes a manufactured moment work

A manufactured moment needs two things to generate genuine demand: scarcity and social proof. The scarcity can be real (limited stock, genuine product drop) or temporal (a 72-hour window). The social proof needs to be earned before the moment opens, not assembled on the day.

The anatomy of a well-structured manufactured moment typically looks like this:

Pre-launch phase (Days -10 to -3). Email capture or waitlist campaign. Paid media focused on driving traffic to a landing page with a registration CTA - not a purchase CTA. The goal is to build a qualified, high-intent audience list before any promotional mechanics go live. Meta and TikTok cold traffic at this stage is cheap because you are not asking for a transaction. You are building the list you will convert when the moment opens.

Warm-up phase (Days -3 to -1). Shift paid allocation toward the engaged segment you have been building. Retargeting the pre-launch registrants, product browse remarketing, and email to the waitlist with preview content. The goal is to arrive at launch day with a warm audience that is primed to buy within the first few hours - because early conversion velocity is the single biggest driver of the algorithm learning window that follows.

Peak window (Days 1 to 3). Full spend activation across cold, engaged, and repeat. Creative shifts to conversion-optimized formats. Email goes to the full purchasable list in a structured sequence: launch send, 24-hour check-in, 48-hour social proof update, final hours urgency. Paid is pushing volume; email is driving conversion rate. The two should be coordinated to hit the customer at multiple touchpoints without being repetitive.

Close-out and tail (Days 4 to 7). Spend begins to pull back as conversion efficiency drops. Focus shifts to non-purchasers from the peak window - a final email to the list segment that opened but did not buy, retargeting of site visitors who added to cart. This phase often delivers 15-20% of the total moment revenue at a fraction of the peak-day CAC, because you are working a warm audience that has already been through the consideration cycle.

Spend flexibility is not the same as reactivity. Reactive spend changes are driven by single data points - a bad day, a good day, a competitor promotion. Flexible spend changes are driven by a read of structural signals across multiple inputs, executed against a pre-agreed framework.

The framework for spend flexes has two sides: the signals that tell you to push harder, and the signals that tell you to pull back.

Signals that support flexing up

Spend should increase when the efficiency metrics are holding or improving at current spend levels. The specific signals to watch are:

• Conversion rate is stable or rising across the past 3-5 days - not just on a single day

• CTR is holding on cold traffic, indicating the audience pool is not yet saturated at current volume

• ACoS is within target and nMER is above floor on the new customer segment specifically

• Inventory depth supports increased demand - there is stock available to fulfil the volume the spend increase will generate

• The algorithm has had sufficient learning time at the current budget level (typically 7-10 days post any significant change) before the next increase

When all of these conditions are met, the order of operations for a spend flex up is: first expand audience depth within existing campaigns before increasing budget, then increase budget incrementally (15-25% at a time to avoid resetting algorithm learning), then monitor the efficiency read for 3-5 days before the next increment.

Signals that support pulling back

Spend should decrease - or hold rather than increase - when:

• ACoS has been above target for 5 or more consecutive days across multiple ad sets, not just one

• CTR is declining on cold traffic despite creative refresh - indicating audience saturation at the current volume

• nMER on new customer campaigns is below floor, meaning you are acquiring customers at a cost that does not make LTV sense

• Inventory is approaching a level where a spend increase would generate demand you cannot fulfil within a reasonable timeframe

The critical distinction here is between a structural efficiency decline and a temporary variance. A single bad day is not a spend flex signal. Three days of consistent underperformance across multiple metrics is. The framework is designed to prevent the day-trading mentality from expressing itself as spend volatility.

Holding budget reserves

One of the most practical applications of spend flexibility is the budget reserve - holding back 10-15% of the monthly or moment budget and deploying it only when the efficiency signals support it. This is not being conservative. It is maintaining optionality.

A brand that deploys 100% of its budget on day one of a promotional moment has no ability to capitalize on a stronger-than-expected day three. A brand that enters the moment with 15% in reserve and clear deployment criteria can push harder when the signals support it and arrive at the end of the period with a better efficiency read than the plan projected.

This is the most nuanced part of performance marketing planning, and the part that most growth teams get wrong in one of two directions: either they hold their standard efficiency targets through a peak moment and leave new customer volume on the table, or they abandon efficiency targets entirely during a peak and acquire customers at a cost that never pays back.

The right approach is neither. It is the concept of peak permission - a deliberate, data-backed decision to lower efficiency floors during a peak moment for a defined period, because the LTV math of a customer acquired during a moment justifies a higher acquisition cost than the same customer acquired in a standard period.

Why the LTV math changes during a peak

A customer acquired during a promotional moment often behaves differently from one acquired during a standard period. They came in through a higher-intent environment, they often have a larger first order (driven by promotional mechanics or bundling), and they are more likely to engage with post-purchase retention flows because the moment created a stronger brand impression than a standard ad conversion.

This does not mean all peak-acquired customers are higher LTV - cohort data is essential here. But for brands with strong retention data, the L3M repurchase rate and average order value on peak-acquired cohorts typically justifies a CAC that is 20-40% higher than the standard-period target, while still achieving a positive LTV-to-CAC ratio.

Setting floor and ceiling targets for peak periods

The framework for peak efficiency targets should be explicit and agreed before the moment opens - not renegotiated in real time based on daily results. The structure we use looks like this:

The MER floor naturally compresses during a peak because total revenue is elevated - the blended efficiency of all spend looks different when the denominator is much larger. This is expected and should not trigger a spend pull-back. The metric to watch during a peak is nMER, not blended MER - because nMER isolates the efficiency of new customer acquisition spend and tells you whether you are buying growth at a justifiable rate.

ACoS as a lever, not a guardrail

During a peak moment, ACoS should be treated as a flex lever rather than a fixed constraint. The ACoS ceiling expands to allow for the higher CPM environment and the higher intent audience that a peak generates. But it should expand within a defined range, not indefinitely.

The practical read: if ACoS on conversion campaigns is running at 28% during a peak where your standard target is 25%, and nMER is holding above its floor, that is a spend-forward signal. If ACoS is running at 38% and nMER is approaching the floor, that is a hold signal regardless of the moment context.

Post-peak efficiency recovery

One of the most reliable patterns in peak performance marketing is the post-peak efficiency recovery. In the 2-3 weeks following a significant peak moment, MER and nMER typically improve as the cohort acquired during the peak begins their second and third purchase cycle. ACoS often improves as the algorithm carries forward the conversion signal quality from the peak window.

This recovery is not guaranteed - it depends on retention mechanics being in place to activate the new cohort in the post-peak window. An immediate post-peak welcome sequence, a second-purchase incentive email at day 14, and a retention-focused paid re-engagement campaign in the 3-4 weeks after the moment are the levers that convert the peak acquisition into the post-peak efficiency recovery.

Here is a real example of what happens when this structure is in place.

One of our brands ran a promotional moment. Before the period opened, we had a full day-by-day plan: projected sales, spend targets by channel, MER and nMER floors, peak-permission ACoS ceiling, allocation split by segment, and an inventory check confirming stock depth was sufficient for the planned demand volume.

Day 3 - we missed. Revenue came in below plan.

But we were already ahead of the cumulative target because days one and two had over-delivered. The lever structure meant we did not react to day three with panic - we held position and let the plan run.

Day 7 - we missed again.

Day 8 - we over-delivered and recovered the gap.

On target for nMER, MER, and ACoS. Within planned budget. New customer CAC within the peak-permission ceiling.

A team without that structure would have reacted on day three - cut spend, changed creative, disrupted the algorithm - and probably landed somewhere worse despite more effort and more noise. The plan held because the architecture held. The architecture held because it was built from the data up, anchored to prior year shape, with explicit peak permission targets and a lever framework agreed before the moment opened.

• Forecasts built backwards from a revenue goal are not forecasts - they are targets dressed up as plans. Build from the data up: historical trajectory, inventory depth, moment shape, and desired future state.

• Prior year marketing moment data is the most valuable forecasting asset most brands are not using. Track the shape of every moment - the day-by-day curve, the efficiency read, the NC/RC split - and index forward from it for the equivalent event this year.

• Inventory position sets the spend ceiling. You cannot productively deploy more paid than the available stock can fulfil. Align the media plan with the inventory plan before the moment opens.

• The marketing calendar is a strategic document, not a scheduling tool. Design the whitespace as deliberately as the peaks. The periods between moments are where retention compounds and the next peak becomes cheaper.

• Manufactured moments - built on scarcity, social proof, and a warm pre-built audience - create urgency without requiring margin sacrifice. The pre-launch list-building phase is often more valuable than the launch day spend itself.

• Spend flexes should follow signals, not emotions. Hold budget reserves. Flex up only when conversion rate, CTR, inventory, and nMER all support it. Pull back on structural signals, not daily variance.

• During peaks, lower the efficiency floor deliberately. nMER is the anchor metric, not blended MER. Set an explicit ACoS ceiling for the moment, agree peak-permission CAC targets in advance, and architect the post-peak retention window to convert the acquisition into long-term revenue.