Not with a disaster. With a campaign that ran, reported acceptably, and changed nothing about the business. We built Groundswell to be the opposite kind of company: small enough that the people who sold the work are the people who do it, and stubborn enough to argue about whether it is working.
We take a position
A recommendation with three options and no opinion is a way of avoiding responsibility. We tell you what we think and why, and we write it down so it can be held against us.
Measurement first
If we cannot agree how we will know whether this worked, we have not finished planning it. That conversation happens before the budget, not after.
Small and senior
The people in the pitch are the people on the work. We stay small on purpose, and we turn things down to keep it that way.
Boring where it counts
Accessibility, performance, data handling and contracts are not where we express ourselves. They are where we are dependable.
Kestrel had spent two years buying business current accounts on rate comparison sites. Acquisition worked, in the sense that accounts opened. It did not work in the sense that mattered: the accounts sat at low balances, switched again within a year, and the cost of holding them exceeded what they returned.
02 / 04
What we found
We interviewed thirty-one founders and finance leads, including eleven who had left. Almost nobody could remember what rate they had signed up for. What they remembered, in detail and with some heat, was the eight working days it took to get a second signatory added, and the fact that nobody at the bank could tell them why.
03 / 04
The move
We repositioned Kestrel around operational speed rather than price, and made the claim falsifiable: published turnaround times for the twelve most common business banking requests, updated monthly, including the ones that were still slow. The brand system was rebuilt around that transparency — a lot of plain type, a lot of published numbers, no photography of people shaking hands.
04 / 04
What changed in the media
Comparison-site spend dropped to a third of its previous level. The budget moved into search against operational queries, and into a long-running programme with accountants and bookkeepers, who turn out to be the people founders actually ask. Blended acquisition cost rose. Contribution per account rose considerably further, which was the point.
Median change across engagements running twelve months or longer, measured against a pre-agreed baseline rather than against the same period last year.
+184%
Non-brand organic sessions
Median, 12-month engagements
+72%
Qualified pipeline
B2B engagements only
3.4×
Contribution after media
Against holdout, not platform-reported
+48%
Enquiry to close
Where we own the full funnel
Figures illustrate the shape of the work. Replace them with audited client data before this site goes anywhere near a prospect.
How we work
Five stages. The last one is where the return actually lives.
01
Discover
Weeks 1–3
We go looking for the thing nobody wants to say out loud.
Interviews with your team, your customers and the people who chose somebody else. A full read of the analytics, the CRM and the finance export, reconciled against each other rather than taken on trust. We are looking for the gap between what the business believes about itself and what the data will support.
Research findings
Reconciled baseline
The uncomfortable question
02
Frame
Weeks 3–6
A small number of bets, sized and sequenced.
We turn the findings into a position and a plan: what you are for, who for, and the three or four things worth doing about it this year. Everything is costed. Everything has an owner. Anything that cannot survive that treatment gets cut here rather than in month seven.
Positioning
Growth model
Roadmap with owners
03
Make
Weeks 6–16
Build the smallest complete version, then put it in front of people.
Identity, site, campaign, measurement — whatever the plan calls for, built to a standard rather than to a deadline, and reviewed in the open every two weeks. We would rather ship one finished thing than four that need a caveat.
Design system
Production build
Measurement wiring
04
Launch
Two weeks either side
Launch is a fortnight, not an afternoon.
Soft launch, holdout in place, instrumentation verified against a known baseline before anyone celebrates. We watch the first fourteen days closely because that is when the assumptions break, and we would rather be in the room when they do.
Rollout plan
Holdout design
Day-14 readout
05
Compound
Quarterly, ongoing
The work that makes the earlier work worth more.
A quarterly cycle of experiments, editorial and media reallocation, with a written record of what we learned and what we were wrong about. Most of the return on a good position arrives in years two and three, and only if somebody keeps tending it.
Experiment programme
Learning log
Reallocation
In their words
“They talked us out of the campaign we came in asking for. It took two more months and it was the right call, which is not a sentence I expected to write about an agency.”
“The reporting is the part I would keep if I could only keep one thing. We stopped arguing about whose number was right and started arguing about what to do, which is a much better argument.”
“Six weeks in they told us one of our two flagship channels was not doing anything. Nobody enjoys that meeting. It saved us most of a year’s budget.”
“What I noticed was that the same three people were still on the account in month fourteen. In this industry that is close to remarkable.”
01 / 04
Insights
Latest thinking
What we are arguing about internally, written up before we are
entirely sure we are right.
The split between brand and performance is a budgeting convenience, not a description of how buying works. What happens when you stop pretending they are separate.
