Zero-based budgeting revisited: what it actually demands from a CFO

Zero-based budgeting is one of those finance concepts that sounds self-evident until you try to implement it. Build every budget from scratch each cycle. Justify every pound or dollar before you spend it. No automatic rollovers. In theory, rigorous. In practice, the experience at companies from Kraft Heinz to Unilever has shown that ZBB can simultaneously reduce costs and destroy organisational capacity, sometimes in the same quarter. The concept itself is not the problem. The confusion about what it actually demands is.

Why it matters for CFOs specifically

The CFO's relationship with ZBB is different from any other executive's. A COO thinks about ZBB as a cost discipline. A CEO thinks about it as a narrative for investors. The CFO has to think about what it does to the quality of financial information flowing through the organisation.

When ZBB is applied carelessly, the FP&A function drowns. Every budget holder is suddenly required to document and defend every line item, which generates enormous volumes of low-value justification work. Finance teams spend their bandwidth processing paperwork rather than analysing decisions. The 2026 PEX report on AI adoption in finance noted that finance teams remain cautious about delegating analytical control to AI systems, and that finding points to a structural problem: if your team is buried in line-item reviews, they cannot exercise the judgment that ZBB is supposed to produce.

There is also an asymmetry of effort. ZBB demands the most work in the areas where the numbers are smallest. A business unit manager spends four hours justifying a training budget of £8,000 while a capital allocation decision worth £4 million gets thirty minutes of board time. That inversion is not hypothetical. It is the most common complaint from divisional finance directors who have lived through a ZBB implementation.

The current macro environment makes this relevant again. With producer price indices rising on energy price surges and the Federal Reserve's rate trajectory still uncertain, CFOs are under pressure to demonstrate cost discipline without impeding the investments that drive growth. ZBB gets proposed in exactly this kind of environment, and CFOs need to know precisely what they are taking on.

How it actually works: the mechanics

In a conventional incremental budget, last year's approved spend is the starting point. Each line adjusts up or down from that base. The assumption embedded in that process is that whatever was funded last year was worth funding. ZBB removes that assumption entirely.

The practical mechanics start with decision packages. Each spending unit identifies what it does and breaks its activities into discrete packages: a baseline package (what happens if you fund the minimum), and incremental packages (what you get for each additional level of funding). Budget holders rank their packages, and the organisation then allocates resources by working down the ranked list until money runs out.

Consider a concrete example. A consumer goods company with £500 million in annual operating costs applies ZBB to its marketing division. The marketing team does not submit "last year's budget plus 3%." Instead, they submit a baseline package covering only media commitments already contractually locked in, and then three incremental packages covering brand awarenessbrand awarenessThe degree to which your target audience recognises or recalls your brand, either prompted or unprompted. It measures how present your brand is in people's minds.View full definition → campaigns, performance marketing, and sponsorships. Finance reviews each package against expected revenue contribution. The sponsorship package, which would have rolled forward automatically under incremental budgeting, gets cut because no one can document a return. The performance marketing package gets funded above last year's level because the data supports it.

That outcome, a simultaneous cut in one area and increase in another, is what ZBB is supposed to produce. Understandinghow the decision-package structure connects to driver-based modelling is what separates a meaningful ZBB exercise from a paperwork exercise dressed up in ZBB language.

The hidden time cost is real. A full ZBB cycle at a company of 5,000 employees typically takes three to four months of preparation. At Kraft Heinz, 3G Capital ran ZBB aggressively from 2015 onward and achieved significant margin expansion in the short term, but by 2019 the company had written down £13.8 billion in asset value, partly because the cost discipline had also hollowed out brand investment. The ZBB tool had worked. The ZBB judgment had not.

When to use it and when not to: the honest tradeoffs

ZBB earns its cost in specific circumstances. It works well for organisations that have grown through acquisition and have never rationalised overlapping cost structures. It works well for companies entering a genuinely new strategic phase where historical spending patterns have no connection to future requirements. And it works for discrete functions, shared services, IT infrastructure, real estate, where the activities can actually be defined and ranked.

It does not work well as an annual discipline applied across an entire organisation. The cognitive and administrative burden is too high to repeat every twelve months without degrading the quality of the outputs. Most organisations that claim to run ZBB annually are actually running a more rigorous incremental budget with better documentation requirements. That is not a bad outcome, but it is worth calling accurately. Knowingthe genuine differences between rolling forecasts and zero-based approaches lets you deploy each where it actually belongs rather than defaulting to one methodology for every planning cycle.

The honest tradeoff is this: ZBB produces better resource allocation decisions in the first cycle you run it properly. It produces diminishing returns if you repeat it without changing the strategic questions it is meant to answer. The organisations that get the most from ZBB treat it as a periodic reset, not a permanent operating model.

One further consideration for 2026 specifically: the same AI capabilities that Anthropic disclosed blocking actors from misusing for biological weapons research are the capabilities that finance teams now have access to for scenario modelling and decision-package analysis. AI can process decision packages at scale in ways human reviewers cannot. But, as the PEX data shows, finance functions are not yet comfortable delegating analytical judgment to those systems. That gap matters for ZBB implementation: the tool that could make ZBB genuinely efficient is available, but the organisational trust to use it is not yet there.

ZBB is a good idea applied at the wrong cadence, in the wrong scope, with insufficient attention to where the real decisions live. Run it on the right perimeter, with a clear question it is meant to answer, and it produces real results. Run it as an annual ritual and it consumes the analytical capacity it was supposed to free up.