Real-world application: budget allocation & forecasting in practice

In the last week of March 2020, Airbnb's forward bookings were falling faster than the finance team could rebuild the model. Cancellations ran ahead of new reservations, and Brian Chesky later described the company as having lost about 80% of its business in eight weeks. Sitting on the 2019 books was roughly $1.6 billion of sales and marketing spend, and a 2020 plan built on that base. Airbnb pulled about $800 million of planned marketing, switched off most paid acquisitionpaid acquisitionVisitors arriving via paid ads or sponsored placements, where you pay a platform to display your message rather than earning visits organically.View full definition →, stopped brand campaigns, raised expensive emergency debt in April, and cut around 1,900 jobs in May. This lesson stays inside that one decision: what held when the spend stopped, what the numbers did not actually prove, and how the plan got rebuilt around brand.

The shock: what got switched off

The cut was not a trim across the plan. Airbnb took the working spend line (the working versus non-working split the foundations lesson sets out) close to zero in paid search, paid social and most partner channels, while keeping the parts of marketing that talked to hosts and press. Two things forced the shape of it. First, cash: with travel frozen and a $1 billion debt raise priced on distressed terms, every dollar of variable spend was competing with runway. Second, speed: paid channels can be switched off inside a day, whereas people, content commitments and sponsorships cannot. That asymmetry is why performance spend absorbs the first hit in almost every shock, regardless of what the response curves say about marginal return. The order of cuts follows liquidity, not efficiency.

Key sub-concept 1: what held when the spend stopped

The result is the reason this case gets quoted in every budget defence since. With performance marketing off, Chesky has said traffic came back to roughly 95% of prior levels. Airbnb has also disclosed that the large majority of its traffic arrives direct or through unpaid channels, around 90%. Revenue for 2020 landed at about $3.4 billion against $4.8 billion in 2019, a drop of roughly 30% in a year when airline and hotel revenue fell much further. The demand that returned was domestic, drive-to, longer stays: people searching "airbnb" by name, opening the app, not clicking an ad. What Airbnb had was a base of unpaid demand built over a decade, plus a brand name that had become a verb. The paid spend had been buying a slice of traffic that would largely have arrived anyway.

Key sub-concept 2: why this was not a clean test

Read the pause carefully before you copy it. Airbnb switched off paid acquisition at the exact moment global travel intent collapsed, which means the counterfactual is not "same market without ads" but "almost no market at all". Search auctions were empty, so the spend that got cut was spend that had little left to buy. The 95% figure compares traffic against a shrunken baseline, and it says nothing about the incremental bookings paid search would have delivered in a normal quarter. There was no 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 → holdout, no staggered restart, no control group: the pause was a cash decision that produced a natural experiment, not an experiment anyone designed. Two quarters also tests almost nothing about brand decay, which shows up over years, not months. If you want the answer for your own business, you run the pause as a split by geography or platform, with the same seasonality on both sides, and you hold it long enough for repeat purchase cycles to turn over.

Key sub-concept 3: the costs a pause hides

Switching off paid does not leave the market untouched. When you stop bidding on your own brand terms, intermediaries buy the click: Booking Holdings and Expedia between them spend billions a year on performance marketing, and Booking's marketing expense alone was around $5 billion in 2019. Airbnb's app share and near generic brand cushioned that leakage. A mid-sized brand in the same position hands the transaction to a reseller and pays a commission on it forever.

Three other costs rarely appear in the savings memo. A two-sided marketplace that pauses demand marketing puts host supply at risk, which is why Airbnb kept host communication and PR running while paid was dark. Paid channels carry restart friction: bidding history, algorithm learning phases and agency teams all have to be rebuilt. And the internal story hardens fast, because once a CFO has seen revenue hold without performance spend, getting the money back requires evidence, not a plan.

Key sub-concept 4: rebuilding the plan around brand

Airbnb did not restore the 2019 plan. It rebuilt around brand and earned attention under Hiroki Asai, who came in as global head of marketing in 2020, with no return to a classic performance-led structure. "Made possible by Hosts" launched in February 2021 as the first large brand campaign in years. Product launches became marketing moments in their own right, the May 2022 categories redesign being the clearest example. Chesky has framed marketing as a fixed investment set against the brand rather than a variable cost that flexes with bookings, which changes the planning conversation: the question becomes how much brand investment the business can sustain, not how many bookings this month's spend bought.

Real-world case 1: the rebuild in the numbers

2021 revenue reached about $6.0 billion, ahead of 2019's $4.8 billion, on sales and marketing of roughly $1.2 billion against $1.6 billion two years earlier. Sales and marketing fell from around a third of revenue in 2019 to roughly a fifth, and by 2023 sat near 18% on revenue of about $9.9 billion. That is the durable result: not a one-off saving, but a permanently lower marketing intensity carried by unpaid demand, with performance spend kept in a narrower role.

Real-world case 2: where Airbnb's own logic runs out

The same company shows the limit. Absolute marketing spend has grown again since 2021, and Airbnb has been explicit about investing behind markets where it is under-penetrated relative to the United States. In those markets the 95% argument does not apply, because there is no decade of word of mouth to fall back on and no unpaid base to hold the floor. The rule that comes out of the case is conditional: a brand with 90% unpaid traffic can pause paid and lose little, while a brand with 60% paid traffic that pauses is not testing brand strength, it is cutting revenue.

CMO action items

• Calculate your unpaid demand share this week: direct plus organic plus app sessions as a percentage of total qualified traffic. That single number decides whether an Airbnb-style pause is a test or a self-inflicted revenue cut.
• Design the pause before you need it. Pick two comparable regions or two platforms, define the holdout, the duration (at least one full repurchase cycle) and the metric you will read, so a cash shock finds a ready experiment rather than an improvised one.
• Price the leakage. Estimate what share of your brand-term clicks a reseller, marketplace or competitor would capture if you stopped bidding, and what commission that traffic would then cost you.
• Write the restoration criteria into the cut memo. Name the specific evidence and threshold that returns the money, before anyone discovers that revenue held without it.

Common mistakes that kill results

• Reading a crisis pause as proof of brand strengthbrand strengthThe commercial value your brand adds beyond functional product attributes: the price premium, preference and loyalty it generates.View full definition →. When category demand collapses, cut spend looks free because the auctions are empty. Separate the demand effect from the marketing effect before you make the saving permanent.
• Assuming the brand base is yours rather than the category's. Airbnb's unpaid traffic came from years of word of mouth and a name people type directly. Traffic that arrives through an aggregator, a retailer or a platform's recommendation feed is rented, and it disappears when the intermediary reprices it.
• Pausing one side of a marketplace. Cutting demand marketing while supply sits idle drives host, seller or partner churn, and rebuilding supply takes far longer than rebuilding an ad account.
• Freezing the new intensity everywhere. A lower marketing-to-revenue ratio that works in a mature home market becomes underinvestment in a market where nobody knows the brand yet. Set the ratio by market maturity, not as one company-wide target.