Why this matters
Boards rarely refuse brand spend because they dislike brands. They refuse it because the request arrives as a belief, while the performance budget arrives with a weekly report. The brand versus performance guide explains why that report flatters capture channels; this exercise is about turning that explanation into a case a sceptical board can test, without promising results nobody can yet show, and honest about what the market did.
The broker and its rising costs
Keyhollow Mortgages is an invented online mortgage broker. It earns a fee when a mortgage it arranges completes, takes applications through its website and app, and over six quarters moved its budget out of regional radio, podcast sponsorship and online video into paid search and paid social.
Keyhollow Mortgages: six quarters of spend and outcomes
| Quarter | Paid search | Paid social | Brand activity | Applications | Branded search index | Market applications index |
|---|---|---|---|---|---|---|
| Y1 Q1 | £140,000 | £60,000 | £90,000 | 2,900 | 100 | 100 |
| Y1 Q2 | £160,000 | £70,000 | £60,000 | 2,850 | 98 | 99 |
| Y1 Q3 | £180,000 | £85,000 | £30,000 | 2,700 | 93 | 97 |
| Y1 Q4 | £200,000 | £95,000 | £10,000 | 2,600 | 88 | 95 |
| Y2 Q1 | £215,000 | £100,000 | £0 | 2,500 | 84 | 93 |
| Y2 Q2 | £225,000 | £105,000 | £0 | 2,450 | 80 | 92 |
- Impression share on generic mortgage terms rose from 71% to 89%; average cost per click on those terms rose from £4.10 to £5.60.
- Share of applications from people with no earlier visit to the site fell from 46% to 33%.
- Estimated paid share of voice among online brokers fell from 7% to 3%; Keyhollow's share of broker applications is about 5%.
- Weekly spend and application records go back three years. Brand spend was almost constant for the first eighteen months, then fell as shown.
Why the balance tipped and how to reset it
Last-click AttributionAssigning conversion credit to touchpoints. credited search and social with applications that earlier brand activity had prepared, so each review moved money towards them. The guide's signs of a tipped balance are the ones to look for: rising capture costs, branded search falling faster than the market, a shrinking share of genuinely new customers and capture channels near saturation. A reset needs a split, a funding source judged on marginal rather than average return, and a proof designed before the money moves, because brand effects arrive over quarters.
Task
- Calculate the change in total spend, applications and cost per application across the six quarters, and compare the fall in applications and branded search with the market index.
- Write the diagnosis for the board in plain words: what the market explains, what it does not, and which signals point at the budget mix.
- Propose a brand and performance split within a total no larger than the latest quarter, and say where the brand money comes from.
- Plan the evidence: what will be measured, how often, against what baseline, and what result would make you change course.
- Say when a mix model becomes worth building and what it needs that the current data lacks.
Worked answer
The arithmetic behind the diagnosis
- Total spend: £290,000 in Y1 Q1, £330,000 in Y2 Q2, up 14%.
- Applications: 2,900 to 2,450, down 15.5%, against a market index down 8%.
- Cost per application: £100 to about £135, up 35%.
- Branded search index: down 20%, two and a half times the market's fall.
- Proposed Y3 split within £330,000 a quarter: brand £90,000 (27%), paid search £170,000, paid social £70,000. The £55,000 cut from search comes from generic terms bought above 85% impression share; the £35,000 cut from social comes from broad prospecting with the weakest new-visitor share.
Diagnosis. The market explains part of the fall in applications, not the rise in cost. Keyhollow lost volume nearly twice as fast as the market while spending more, and branded search fell far faster than category demand. Impression share near its ceiling and rising cost per click show the capture channels buying the same searches at higher prices. Fewer applicants are new to the brand, so the pool that performance harvests is shrinking. Its Share of VoiceYour share of market visibility. has fallen below its share of applications, which the share of voice guide associates with losing ground over time.
The split. Ask for a protected brand line of about a quarter of the budget, funded from the least efficient capture spend rather than new money, with the total held flat. Brand activity goes back into the channels that previously ran, aimed at people likely to buy or remortgage within the next year, with financial promotion rules applied to every script before it airs.
The evidence. Run a tracking survey baseline before launch and quarterly after it, asking unaided awareness and consideration among people planning a mortgage, as the brand awareness and lift guide describes. Track branded search as a share of category searches weekly, annotated with rate changes and campaigns. Watch cost per application in the capture channels, where brand's return is cashed. Where radio and video can be bought by region, stagger the launch so some regions start a quarter later and compare branded search and applications by postcode. The board's review points are at two and four quarters: if branded search share and consideration have not moved beyond their noise by the second review, the creative is examined; if capture costs are still rising at the fourth, the split is reduced.
The mix model. A model (MMMStatistical model of channel impact.) cannot yet help, because brand spend barely varied and then fell in step with a falling market, so the two cannot be separated. The mix modelling basics guide explains why variation and controls matter. The staggered launch creates that variation; with roughly two years of weekly data including it, and the market index and lender rates as controls, a model becomes worth building. The comparison of mix modelling and multi-touch attribution is the argument for keeping attribution for keyword and campaign decisions inside each channel.
Common mistakes
- Blaming everything on brand cuts. The market fell too; a case that ignores it loses the board's trust.
- Asking for new money. Funding brand from the least efficient capture spend is easier to approve and test.
- Building a mix model on flat history. Create the variation first.
Check your judgement in the brand investment checkpoint, or revisit awareness planning at the intermediate brand marketing level.
