Cut where the last unit of spend earns least
Every channel follows a curve of Diminishing ReturnsEach extra unit of spend or effort buys less extra result than the one before it.: the first money buys the most responsive customers, and each further amount buys fewer. The marginal returns and forecasting guide explains why the average return on a channel describes what past spend bought, while the marginal return, what the last increment bought, is almost always worse. A budget cut removes the last increments, so the marginal figures are the ones that matter.
That changes the shape of a good cut. A channel with a high average return can have a poor margin because it is near saturation, and a channel with a modest average return can still be earning well at the margin because it has room to grow. Cutting the channel with the lowest average, or cutting everything by the same share, misses this. The Google Ads guide makes the point inside one platform; the same logic applies across channels, with two complications: some costs cannot be cut quickly, and some spend does its work long after the quarter in which it is reported.
A budget that has to fall is also a chance to stop activity nobody chose deliberately. The guide to marketing on a small budget lists the usual suspects, from boosted posts with no job to listings that produce nothing. The strategy section covers how the budget was set and allocated before the cut.
What the cut has to protect
Before ranking anything, agree what the reduced budget still has to achieve. Three questions settle it.
- Which objectives survive? Some of the year's objectives will be kept, some deferred and some dropped. Write the list down, because the ranking that follows protects only what is on it.
- Where does the contribution come from? The products, customer groups and channels that produce most of the business's Contribution MarginRevenue left after every cost that varies with each sale. are the last places to cut, because losing a customer there costs most.
- What is the cut for? Cash this quarter, profit for the year or a lower cost base for the long term each point to a different cut. A cash cut cares about when money leaves the business, so commitments matter; a profit cut cares about what the spend would have returned; a structural cut cares about which activities the business wants to keep at all.
Agree the answers with finance and the leadership team before any channel owner proposes cuts. Otherwise each owner defends their own line, and the cut falls on whoever argues least.
Ranking spend by marginal return
Rank every line by what its last increment of spend has been earning, estimated from the business's own data.
- Recent changes in budget. When a channel's spend went up or down in the past, how did its results move? The change in results divided by the change in spend is a rough marginal figure.
- Saturation signs. Rising cost per click or per thousand impressions as spend rises, frequency climbing in the same audience and impression share already high on core terms all suggest the margin is poor, as the marginal returns guide describes.
- Incrementality evidence. Where a test has compared exposed and unexposed groups, it is the best evidence available. The incrementality testing guide covers the designs, and the budget increase incrementality test measures the marginal cost per incremental conversion directly, which is the figure a cut needs in reverse. The hierarchy of evidence explains why such a test outranks readings of past budget changes and saturation signs.
Express the margin in the measure the business manages to, a marginal cost per customer or a marginal return (Marginal ROASExtra revenue returned by the last increment of ad spend.), and compare it with what a customer is worth. The ROAS calculator does the arithmetic. The walk-through on entering a budget slice into the ROAS calculator shows how to enter each line's last increment of spend with the revenue it brought, and how to rank those increments against break-even.
| Kind of spend | How to read its margin | The usual trap |
|---|---|---|
| Paid search on category terms | Results at recent budget levels; impression share and cost per click trends | Treating a high average return as proof the last increment pays |
| Brand search and retargeting | IncrementalityConversions caused by marketing. tests or holdouts | Attribution credits them with sales that would have happened anyway |
| Paid social and video for reach | Reach, frequency and the lagged trend in branded search and new customers | Judging reach on last-click sales and cutting it first |
| Email and retention programmes | Results per send against the list's size and engagement | Cutting the platform or content that keeps the list alive |
| Tests and experiments | Whether the reduced budget can still produce a readable answer | Running every test thinner instead of fewer tests properly |
Attribution makes some of this harder than it looks. The brand versus performance guide explains how last-click and platform attribution flatter capture channels, which is why Branded SearchSearch queries that contain a company's or product's brand name. and retargeting often look like the safest spend on average and may be the least incremental at the margin. The Google Ads guide's advice holds across channels: neither keep nor cut such spend on an assumption; decide on evidence where it exists and record the assumption where it does not.
Fixed costs, variable media and commitments
A marketing budget is not one pool of money. Lines fall at different speeds, and a cut planned as if they were all the same either misses its target or breaks a contract.
| Category | Examples | How quickly it can fall | What cutting it costs |
|---|---|---|---|
| Variable media | Search, social, display and video bought by the click or impression | Days | Results at the margin, and a learning period on some platforms |
| Commitments | Signed insertion orders, agency retainers, event bookings, annual tool contracts | At the next break point or notice date | Penalties if broken early; relationships if handled badly |
| Production and content | Creative, video shoots, content writing | As current work finishes | A thinner pipeline of creative and content for later quarters |
| People | The in-house team and long-term contractors | Slowest, and hardest to reverse | Capacity, knowledge and morale that take longest to rebuild |
The guide to tracking spend against budget explains the difference between committed, spent and invoiced money, and why the headroom that matters is the budget minus what is committed. A cut measured against invoices alone looks larger than it is, because much of the remaining money may already be promised. Start the cut with a list of every commitment, its value and its next break point. A marketing budget tracker with a commitments log keeps that list current all year, so it is ready before a cut is announced.
Work that is slow to rebuild
Some activity can be stopped and restarted with little lasting harm. Other activity loses something on the way down that takes far longer to rebuild than it took to lose, and that work deserves protection out of proportion to how it looks in this quarter's reports.
- Brand reach. The brand measurement guides report that the IPA's published analyses of its effectiveness databank have argued that brand-building effects build over long periods and persist, and the marginal returns guide notes the same body of work cautions against judging brand-building spend on short-term response alone. The brand strategy guide reports Ehrenberg-Bass Institute research arguing that the cues that make a brand easy to notice and recall are built through consistent use over years. The diagnosis of results falling across every channel lists an earlier brand cut as the cause behind a slow slide months later, with branded search flat or falling.
- The email list and retention programmes. A list kept engaged by regular, useful sends is an asset; a list left silent loses engagement and becomes harder to mail well again. Retention programmes protect revenue from customers the business has already paid to win.
- Search content. The budget allocation guide notes that organic search is the clearest case of costs arriving long before returns. Content that ranks keeps earning; content that stops being maintained loses ground slowly and is slow to win back.
Protecting this work does not mean leaving it untouched. A shorter video flight at lower frequency, fewer but better emails, or maintenance of the best-performing pages instead of new content can all reduce cost while keeping the asset alive. What matters is that any cut to it is a decision, written down with its expected delayed cost, rather than the side effect of a ranking on short-term returns. Where that reach sits in a signed media plan of booked radio, posters, press or television, the tutorial on replanning a media plan after a budget cut takes the cut line by line, from each order's notice date to the reach the smaller plan still buys.
Sequencing the cuts
Order the cuts so that the money comes out of the places that cost least, and so that each change can be read.
- First, capture spend at the margin. Trim the top slice of each paid capture channel, beginning with the lines whose marginal cost per customer is furthest above what a customer is worth. The Google Ads guide covers the in-platform order, from exclusions to budget trims to target changes, and why it limits relearning.
- Then tests. The 70-20-10 budget rule reserves a share for experiments, and its own advice applies when money is short: a test needs enough spend to give a readable answer, so run fewer tests rather than thinner ones. Pause the tests that can no longer reach a verdict.
- Then low-yield tools and subscriptions. List every tool, what it is used for and what would happen without it. Cancel or downgrade at the next renewal where nothing would be lost. The tutorial on reviewing marketing tool licences and seats finds the unused seats, add-ons and duplicate subscriptions to start from.
- Then commitments, at their break points. Give notice where a contract allows it, rather than breaking it.
- Slow-to-rebuild work, reduced rather than stopped. And people last, because capacity is the slowest of all to rebuild.
Avoid making every cut on the same day. Changes made all at once leave nobody able to say which cut cost what, and on some platforms they trigger learning periods at the moment there is least money to absorb the noise. Stagger the changes, note each with its date and reason, and read each after its own delay. Before the first change is made, the marketing budget cut checklist gathers this guide's checks into one tickable gate.
Setting expectations and restore triggers
The people who asked for the cut need to know what it will cost before it happens. Forecast the lost volume from the marginal figures: for each line, the spend removed divided by the marginal cost per result of that spend. The total is smaller than the average cost would suggest, which is the case for a targeted cut and also the honest forecast. The marketing budget calculator works the same arithmetic backwards, from a target number of customers to the budget it implies, which shows what the reduced budget can still buy.
Say which effects will not show this quarter. A reduced brand flight will not move this quarter's orders much; it will show later, if at all, in branded search and in the cost of capture channels. Write that down now, so the later effect is not mistaken for bad luck.
Record the decision in the budget reallocation memo: what moves, the marginal evidence, the constraints, the expected loss, what would reverse it and who approved it. Then agree the restore triggers. The scenario planning framework supplies the structure: a signpost the business already measures, a threshold at which it acts, and an agreed response, such as returning a named amount to a named line, that has finance's agreement in advance so it can happen in days rather than weeks. Record the starting settings in each channel so that, when money returns, the cuts can be reversed in order. Each restore triggerThe condition agreed with finance when a budget is cut that says when money comes back and to which line: a measured signal, a threshold and an agreed response. belongs in the same memo as the cut, so whoever reads the decision later sees the route back beside it. When a trigger is met, the tutorial on restoring paid media spend after a cut brings the paid lines back from that record, re-reading the marginal cost at each step.
A garden supplier cuts a quarter's spend (illustrative)
Copperwhin Garden Supplies cuts £18,000 from a quarter
Copperwhin Garden Supplies, an invented online seller of compost, seeds, tools and planters, must cut its marketing budget for the coming quarter from £60,000 to £42,000. An average order brings in about £45 of contribution. Finance needs the saving in cash this quarter, so commitments that cannot end in time stay in the budget.
| Line | Quarter budget | Last-increment cost per order | Rebuild time | Decision | New budget |
|---|---|---|---|---|---|
| Category search | £18,000 | £38 (average £24) | Days | Trim at the margin | £14,000 |
| Brand search | £3,000 | Not tested | Days, if no rival takes the position | Keep and watch | £3,000 |
| Paid social prospecting | £9,000 | £55 (average £33) | Weeks | Trim | £5,000 |
| Retargeting | £4,000 | Last year's holdout: most buyers would have bought anyway | Days | Cut to a minimum | £2,000 |
| Online video for reach | £12,000 | Judged on reach and branded search | Months | Shorter flight, lower frequency | £8,000 |
| Email programme | £3,000 | £9 | Slow; lapsed subscribers are hard to win back | Keep | £3,000 |
| Tests | £4,000 | Not applicable | Not applicable | Pause two, keep one | £1,000 |
| Tools | £3,000 | Not applicable | Days | Cancel one | £2,000 |
| Agency retainer | £4,000 | Not applicable | Notice period ends next quarter | Unchanged; review at notice date | £4,000 |
| Total | £60,000 | £42,000 |
Paid social is trimmed first because its last increment costs £55 an order against £45 of contribution. Category search is trimmed next: still profitable at the margin, but less so than its average suggests. The cuts are staggered over three weeks, starting with retargeting and paid social, then category search, then the video flight.
The memo forecasts about 190 fewer orders this quarter: roughly 105 from category search at £38 each, 73 from paid social at £55 and around 10 from retargeting. It states that the shorter video flight will not move this quarter's orders and may show later in branded search, which the team will read monthly. The restore trigger, agreed with finance in advance: if online orders run above the same weeks last year for four weeks in a row, £4,000 returns to category search first, then £4,000 to the video flight.
Common mistakes
- Cutting every line by the same share. Good and poor spend go together; the margin should decide.
- Ranking on average return. Saturated channels look safe and growing ones look expendable.
- Trusting attribution for brand search and retargeting. They can look best on average and add least at the margin; test before deciding.
- Forgetting commitments. A cut measured against invoices alone is smaller than it looks.
- Stopping slow-to-rebuild work by default. Brand reach, the email list and search content can cost more to restart than to keep at a reduced level.
- Running every test thinner. Fewer tests with readable answers beat many without.
- Making every cut on the same day. Nobody can then tell which cut cost what.
- Forecasting losses from averages. Marginal figures give the honest forecast.
- No restore trigger. A cut without an agreed route back tends to become the new budget.
