Foundations & core concepts of budget allocation & forecasting

Marketing is usually the largest discretionary line a company controls. Netflix reports marketing expense as its own item in the income statement, north of $2 billion a year. Kraft Heinz moves hundreds of millions between brands, media and retailer promotions every planning round. Yet the words used to describe that money get used loosely in most planning meetings: budget, forecast, working spend, incremental, baseline. The looseness is expensive. Two people agreeing on a number while meaning different things by it is how a plan falls apart in month four. This lesson defines the objects. The rest of the module (response curves, zero-based rebuilds, brand versus performance, the boardroom fight) assumes you have them straight.

What a marketing budget and a marketing forecast actually are

A marketing budget is a commitment. It is an amount of money, assigned to owners and time periods, that the company has agreed to spend. It has three properties worth naming: a scope (what counts as marketing at all), a shape (how it splits across brands, markets, channels and quarters), and a liquidity profile (how much of it can still be moved once the year has started).

A marketing forecast is a prediction. It says what that spend will produce, expressed as a range with the assumptions written down: gross adds, signups, pipelinepipelineAll active sales opportunities across the stages of the sales process, together with their combined potential value and probability of closing.View full definition →, sell-out volume, whatever the business counts.

One is a decision, the other is an estimate. The budget is approved once and amended rarely. The forecast should change every time reality contradicts an assumption behind it. Teams that treat their forecast as a second copy of the budget, a number to defend rather than update, end up unable to tell finance anything useful in June.

Key sub-concept 1: working and non-working spend

Working spend buys attention: media, placement, the distribution of the message itself. Non-working spend is everything that makes the message exist and tells you whether it landed: production, agency retainers, research, martechmartechThe connected set of software tools a marketing team uses to plan, run, measure and automate campaigns across channels.View full definition → licences, and in most companies a share of marketing salaries.

The distinction matters because the two behave differently. Working spend can be bought in small increments and turned off. Non-working spend is largely fixed inside a year and often locked into contracts signed before the year began.

Kraft Heinz made this split the operating unit of its marketing plan. Under 3G Capital's zero-based budgetingzero-based budgetingA budgeting method where every expense is justified from zero each cycle, rather than carrying last year's spend forward.View full definition →, every cost line was rebuilt from scratch each year, and non-working marketing costs (agency structures, production, overhead layers) were the first target, on the argument that money saved there could move into media that reaches shoppers. As cost control it worked. As brand building it did not: in February 2019 Kraft Heinz took a $15.4 billion writedown, most of it on the Kraft and Oscar Mayer brands, and Miguel Patricio, who became CEO that year, said publicly that the company had underinvested in its brands and would put money back into marketing. A high working ratio is not a goal in itself. A company with thin research and cheap production can buy an enormous amount of attention for a message that does not work.

Key sub-concept 2: what counts as marketing spend at all

Before anything gets allocated, agree what is inside the line. The usual candidates: paid mediapaid mediaVisitors arriving via paid ads or sponsored placements, where you pay a platform to display your message rather than earning visits organically.View full definition →, agency fees, production, martech and data tooling, events and sponsorship, marketing headcount. Then the awkward ones. Trade promotion (money paid to retailers for price reductions, displaysdisplaysThe total number of times an ad or piece of content is displayed, regardless of clicks. Each display counts as one impression, even to the same person.View full definition → and features) exceeds measured advertising at many consumer goods companies, Kraft Heinz included. Discounts, free trials and referral bonuses are acquisition spend in economic terms even when finance books them as contra-revenue or cost of sales.

Two budgets built with different scopes cannot be compared, which is why most published "marketing spend as a percentage of revenue" benchmarks are close to useless. Write your scope down and keep it stable across years, or your own trend line lies to you.

Key sub-concept 3: baseline and incremental

Baseline is what the business would have sold this period with no marketing activity at all: repeat purchase, existing distribution, brand memory, search demand that already exists, seasonality. Incremental is what the spend caused on top of that.

Netflix illustrates the gap. A large share of any week's viewing and subscription activity comes from members already there and from conversation around a title that would have happened anyway. So the question worth forecasting for a launch campaign is not how many signups arrived during the window. It is how many would not have arrived without it. Segment (customer data infrastructure, acquired by Twilio in 2020; it sells the tooling other companies use to track marketing, so treat its measurement advice as coming from an interested party) grew on a free tier, open source libraries and developer word of mouth. InboundInboundA strategy that attracts prospects organically via valuable content (blog, SEO, social) rather than interrupting them.View full definition → arrived whether or not a campaign ran that month.

IncrementalityIncrementalityThe share of results (sales, conversions, revenue) that only happened because of a marketing action, not what would have occurred anyway.View full definition → is measured with holdouts, geogeoThe practice of making your brand and content visible and citable inside AI-generated answers from tools like ChatGPT, Gemini and Perplexity.View full definition → tests and modelling. The mechanics belong to the frameworks lesson. What matters here: a forecast that predicts total outcomes rather than incremental ones will overstate marketing's contribution every single time, and the overstatement grows with the size of the brand.

Key sub-concept 4: the planning cycle and how much of the budget can move

The standard cycle: finance sets an envelope in the third or fourth quarter for the following fiscal year, marketing splits it by brand, market and channel, commitments get signed, the year starts, the forecast gets revised monthly or quarterly, and the year-end reconciliation feeds the next plan. Most companies run some version of this even when they claim to be agile.

Inside that cycle, money has very different liquidity. Biddable search and social can be changed within hours. An agency retainer runs the full year. A sponsorship or an upfront television commitment signed in March is spent whether or not the plan changes in September. When someone says they reallocated 20% of the budget, they almost always mean 20% of the movable part, which might be half the total.

Netflix runs its cycle around the title slate rather than the calendar. Marketing weight follows release dates, so the working planning unit is the campaign around a title, aggregated up into a quarterly number that investors then read as one line.

Real-world cases

Kraft Heinz, 2015 onward: zero-based budgeting gave the company precise control over non-working cost and almost no view of incrementality. Every planning round could answer what was spent and on what. It could not answer what the spend caused. The 2019 writedown and the subsequent decision to put money back into brands are what happens when a budget is well defined and a forecast is not.

Netflix, 2020: marketing expense fell year over year (roughly $2.65 billion in 2019 to about $2.23 billion in 2020) while paid net additions hit a record of around 36 million. Production shutdowns and lockdown demand did the work. It is the clearest available demonstration of baseline: a period where spend went down, results went up, and nobody serious credited the campaigns.

Segment: much of what acquired customers sat outside the marketing budget entirely. Documentation, client libraries and the hosted free tier are engineering costs on the P&L and acquisition costs in reality. Draw the scope boundary at "paid media plus marketing headcount" and you conclude Segment acquired customers almost for nothing, which is wrong. The cost was real, it just lived in another department.

CMO action items

• Write one page defining your marketing budget scope: what is in, what is out, and where trade promotion, discounts, free tiers and martech licences sit. Circulate it to finance and get their sign-off before the next planning round starts.
• Split last year's actuals into working and non-working, then look at the ratio across three years. A trend in either direction that nobody decided on is the finding.
• Put a baseline estimate on the table before you argue about allocation: what does the business do next quarter with zero net-new campaign spend, and how much would anyone bet on that number?
• Map the liquidity of the current year's budget into three buckets: contracted, committed but cancellable, and movable this week. State the movable percentage in the plan document so nobody promises a reallocation the contracts do not allow.

Common mistakes that kill results

Mistake 1: presenting a forecast as a target. A forecast is your best estimate of what will happen. A target is a commitment you have agreed to be measured against. Once the two are the same document, the estimate stops being honest, because it gets negotiated to whatever number is acceptable. Keep both, label them differently, and expect the gap between them to be the actual subject of the planning conversation.

Mistake 2: reporting total outcomes as marketing results. Every signup in the campaign window gets counted, baseline included. This flatters the team in good quarters and destroys credibility in bad ones, because the moment demand falls for reasons unrelated to marketing, the same arithmetic makes it your fault.

Mistake 3: assuming the budget is liquid. Plans get rewritten in October by people who have not checked what is already contracted. Upfront commitments, sponsorship deals and annual retainers are spent money. If you have never quantified how much of your budget you can actually redirect mid-year, you do not yet know what you are allocating.