The Elastic Band Model: why small budget accounts snap

2026 was the year I finally took the thing I do every day and put it in front of rooms full of people who do it too.

Three stages. SearchDay in Aarhus in April, run by s360, in front of around 200 people in an old industrial workshop. Hero Conf at the Brighton Centre two weeks later, in front of a room of PPC specialists. Then the Performance Marketing Conference in Manchester in June, ten minutes from my desk.

One idea across all three, because I only wanted to stand up and talk about the thing I know better than most: what happens to a Google Ads account when the budget is small and there is no room to get it wrong. Three formats, three very different rooms, and each one made the idea sharper.

01Why small budgets

Almost every client I run is small budget by definition. Plumber in Didsbury. Dentist in Altrincham. Ecommerce brand in the Northern Quarter. Solicitor in Chorlton. Different industries, same problem.

Most of the advice in our industry is written for accounts with real data, real volume and real margin for error. The tactics are not wrong in isolation. They are wrong for the context, and nobody talks about the context.

On a large account a bad decision costs you some wasted spend. You spot it in the data a few weeks later, you fix it, you move on. On a small account a bad decision does not give you time to notice. A bad week is a quarter of the monthly budget gone, the client asking questions, and trust starting to go.

Small budgets do not bend. They snap.

Big accounts have flex. They absorb bad months, bad tests and bad decisions. Small ones do not have that flex, so when you stretch them they do not slowly degrade. They break. The campaign goes dark, the algorithm loses its way, the client loses faith.

So the question I ask on every small account is not how to optimise it. It is how much tension the account is under right now.

02The Elastic Band Model

That question needed a shape, so I gave it one. Picture a rubber band with four forces acting on it. Work with them and you survive. Ignore them and it snaps.

It is a diagnostic lens rather than a list of tactics. Every recommendation I made on those three stages came back to one job: take tension out of the account.

03What high tension actually looks like

The Aarhus version had room for detail, so I spent time on the three places tension usually comes from.

Structure. A thousand pounds a month spread across five campaigns that compete with each other. None of them get enough data. Smart Bidding on all five, all of them guessing, none of them learning. A full funnel built just in case, before the bottom of the funnel is even profitable.

Targeting. Broad match with no efficiency target, so the budget goes on exploration that will never pay back. Audience layers stacked on top of two conversions a month, which is noise rather than data. Geo targeting spread so wide that you are present everywhere and competitive nowhere.

Optimisation. This is where good intentions do the most damage. A/B tests that will never reach significance. Smart Bidding running on a fraction of the conversions it needs. Constant changes that reset learning every single time, so the algorithm never gets to finish a thought.

04The account behind the talk

I used one real account in all three versions. A UK service business with genuine demand, decent margins and a Google Ads account that was not working. Here is what I inherited.

Thirteen campaigns on a tiny daily budget. Under 10% impression share almost everywhere, so effectively invisible. Hundreds of conversions in the interface that did not match reality, because page views and other micro actions had been set as primary goals. The algorithm thought it was smashing it. The client had barely any new customers.

Then it got worse. Ninety ad groups. 726 keywords. Sixty eight locations. All broad or phrase with almost no negatives, which on that budget is 726 ways to leak money. Search partners on, Display on. Target CPAs set at 92p when the real cost per lead in that market is £30 to £60, so the algorithm was chasing a number it could never hit. Smart Bidding sitting on top of all of it with nothing trustworthy to learn from.

What I did next was not clever, and that is the point.

• Fixed conversion tracking first. Form fills and calls only. No click counting dressed up as leads.
• Consolidated thirteen campaigns into one at £70 a day, so the whole budget worked together instead of fighting itself.
• 726 keywords down to 34, 90 ad groups down to 3, 68 locations down to 4. Commercial intent only.
• Mostly exact and phrase match, with 2,540 negatives and counting, so every click became intentional.
• Search Network only. Partners and Display off.
• Rebuilt the CPA target from the unit economics rather than from what anyone wished it was.
• Smart Bidding back on, this time with inputs worth learning from.

Same budget, same market, same three month window a year apart. The difference was tension, not tactics.

Fundamentals are 90% of the result on a small budget.

That line is uncomfortable in a conference room, because people come to hear something new. We want tactics. We want the thing nobody else knows. But the gap between good and bad performance on a small account is almost never a missing tactic. It is the fundamentals being skipped in favour of something that looks more impressive in a deck.

05The four principles

Each principle takes tension out of one dimension of the band.

Keep it simple, and consolidate by default. The question is not whether to split something. It is what your reason for splitting is. Different budgets, different targets or genuinely different intent are all valid reasons. Everything else is one campaign. Segmentation feels like control. On a small budget it is usually just complexity.

Capture high intent demand before you try to create it. Impression share is the first metric I look at. If you are losing it to budget on terms that convert, you are spending money somewhere less valuable. Own the bottom of the funnel completely, then earn the right to do anything more creative.

Track what pays the bills. Leads, calls, sales, revenue. Micro conversions make the dashboard look healthy and teach the algorithm the wrong lesson. Bad tracking is not only a reporting problem. Nothing downstream of it can be trusted, including your bid strategy, your keyword decisions and your targets.

Optimise gradually. The hardest one, because when performance is poor the instinct is to act. Pick the single biggest constraint, fix that, let the data accumulate, then evaluate. Constant changes do not improve a small account. They keep it in learning.

On Smart Bidding specifically: 15 conversions a month is where Target CPA technically starts, not where it works. Thirty is where it becomes functional. If a campaign cannot get there, that is a signal to consolidate rather than to add another strategy on top.

06What forty five minutes taught me that fourteen could not

The interesting part of taking one idea to three stages was watching what fell away each time.

Aarhus had room for everything: the model, the three sources of tension, the case study, the four principles and a closing checklist. Brighton cut it to eighteen minutes with a hard stop and no Q&A, which works out at roughly half a minute per slide. The checklist went, because there is no point putting something on screen that nobody has time to read. A fifth principle went with it.

Brighton also taught me something I did not expect. The organisers sent my deck back. Too much content on most slides, and the font sizes were too small for the room. They were right. Splitting one crowded slide into three, with a single idea on each, made the talk sharper without adding a second to the runtime.

By Manchester I was down to fourteen minutes. The four labelled principles came out entirely and the talk became the band, the account, and then the checklist, which earned its place back. What survived every cut was the model itself and the before and after. Everything else turned out to be supporting cast.

If you ever get the chance to build one idea out at three lengths, take it. Compression is the best editor I have ever worked with.

07The checklist that made it back in

Five questions. You can run these against any account you inherit in about ten minutes.

• Can each campaign get enough conversions to learn? If not, consolidate.
• Can you justify why every moving part exists? Cut what you cannot.
• Are you losing impression share on profitable campaigns? Stop spreading thin.
• Is your conversion tracking measuring real business outcomes? Not page views. Leads, calls, sales.
• Does your bid strategy have enough data to learn? If not, simplify the structure until it does.

---

← Back to all articles