How to Know Which Marketing Experiments to Kill (And Which to Scale)

You're running three marketing tests simultaneously. Budget's tight. One's showing promise, one's bleeding cash, and one's… unclear. Which do you feed? Which do you kill? Most founders wait too long to make this call, hoping underperforming tests will magically turn around. This article gives you a practical decision framework based on two core metrics that remove emotion from the equation.

Why Most Founders Wait Too Long to Kill Tests (And Scale Too Early)

Picture this: you're running Google Ads, LinkedIn outreach, and content marketing. You've spent $4,000 across all three. LinkedIn's costing you $350 per lead. But you've already invested two weeks optimising the targeting. Surely one more tweak will fix it?

This is the sunk cost fallacy in action. You keep feeding a losing test because you've already committed time and money. The logic feels sound: "We're so close to making this work." But every dollar you pour into a failing channel is a dollar you're not putting into something that actually works.

The opposite mistake is just as common. You get one good week from Facebook ads—three customers at $60 each—and immediately triple your budget. Then CAC shoots up to $180 because you scaled before you had statistical significance.

Dave Girouard, founder of Upstart, argues that timing matters more than the decision itself. Waiting costs more than being wrong. When you're burning through runway, indecision is expensive. A mediocre decision made today beats a perfect decision made in three weeks when you've run out of cash.

The Two Numbers That Actually Matter

Forget impressions. Forget click-through rates. Forget engagement metrics. When you're deciding whether to kill or scale a marketing test, only two numbers matter: CAC trajectory and time to payback.

CAC trajectory tells you if your acquisition cost is improving as you optimise. Time to payback tells you how fast you recover that cost. Together, they give you a clear kill/scale signal.

Vanity metrics feel good but don't inform resource allocation. A channel with 10,000 impressions and 2% engagement means nothing if it's costing you $400 per customer with a 12-month payback. The following sections break down exactly how to calculate and use these two metrics.

Customer Acquisition Cost (CAC) trajectory

CAC is total marketing spend divided by new customers acquired. Simple enough. But the single number doesn't tell you much. What matters is the trend.

Is your CAC improving week-over-week as you optimise? Or is it flat? Worsening?

Here's what a promising trajectory looks like: Week 1 CAC of $150. Week 2 drops to $120. Week 3 hits $95. That's a signal. The channel's responding to your optimisation. It might be worth scaling.

Don't calculate CAC until you have at least 30-50 conversions. Ten customers tells you almost nothing. One enterprise deal can make a channel look amazing, but it's not repeatable data. You need volume before the trend means anything.

If you're tracking CAC in a spreadsheet and manually calculating conversions, you're wasting time. Lead Recorder automatically tracks where your leads come from and calculates CAC by channel in real time, so you can spot trajectory shifts immediately.

Time to payback

Time to payback is how many months it takes for a customer's revenue to cover their acquisition cost. This matters more than CAC alone.

A $200 CAC with a 1-month payback beats a $50 CAC with a 12-month payback every time—especially when you're cash-strapped. The $200 customer pays for themselves in 30 days. The $50 customer ties up your cash for a year.

Here's the calculation: if your CAC is $200 and your average customer pays you $100 per month, your payback period is 2 months. B2B SaaS companies typically aim for 12-month payback, but early-stage founders need faster. You can't wait a year to recover acquisition costs when you've got six months of runway.

When the data lies: sample size and false signals

Small sample sizes produce false signals. Elliot Shmukler's A/B testing framework emphasizes that even data-driven decisions need sufficient volume to be valid.

Rule of thumb: wait for at least 100 visitors per variation and 30+ conversions before drawing conclusions. One enterprise deal can make a channel look incredible, but if you can't repeat it, it's noise.

The founder's dilemma: you can't always wait for perfect sample size when cash is tight. Fair enough. But if you're making decisions on 10 conversions, acknowledge you're taking a risk. Don't pretend the data's conclusive.

The Kill/Scale Decision Matrix

Every marketing test falls into one of four quadrants based on CAC trajectory (improving vs. flat/worsening) and payback period (fast vs. slow). This framework removes emotion. The data tells you what to do.

Plot your tests on this matrix after you've collected sufficient data. You'll immediately see which tests to kill, which to scale, and which need one more iteration. The following sections cover each quadrant with specific action steps.

Quadrant 1: Kill immediately (high CAC, no improvement trend)

CAC is above your target AND showing no improvement after 3-4 optimisation cycles. This is a kill signal.

Example: LinkedIn ads with a $400 CAC that's been flat for six weeks despite creative and targeting changes. You've tested different headlines, audiences, and offers. Nothing's moved the needle. Stop spending today. Reallocate the budget within 48 hours.

Killing a test isn't failure. It's smart resource allocation. If you've already iterated multiple times without improvement, one more tweak won't save it. Move on.

Quadrant 2: Scale aggressively (low CAC, fast payback)

CAC is below target AND payback is under 3 months with improving or stable trajectory. Pour fuel on this fire.

Increase budget by 50-100% immediately. If performance holds, double again. Example: Facebook ads delivering $80 CAC with a 6-week payback. This is a winner. Scale it hard.

Watch for the scaling ceiling. Most channels have a saturation point where CAC rises as you increase spend. But when you find a winner, hesitation costs you growth.

Quadrant 3: Iterate one more cycle (promising trend, needs optimisation)

CAC is improving week-over-week but hasn't hit target yet, or payback is slow but trending faster. Give it one more shot.

Run one more 2-week optimisation cycle with a specific hypothesis. Example: Google Ads with CAC dropping from $200 to $150 to $120. One more iteration might get you to your $80 target.

Set a deadline. If CAC doesn't hit target after this cycle, move it to Quadrant 1. Eric Reis's principle applies here: startups learn through experimentation, but you need clear success criteria. Don't let tests drift indefinitely.

Quadrant 4: The dangerous middle (when to trust your gut vs. the data)

Mixed signals. Maybe CAC is good but payback is slow. Or CAC is high but improving rapidly. This is where founder judgment matters most.

Ask yourself: does the trajectory suggest this will reach Quadrant 2, or is it stuck? If you can't articulate a clear hypothesis for why it will improve, kill it.

Data informs, but founders must decide. Don't let tests languish here indefinitely. Set a 2-week deadline to move it to another quadrant.

What to Do When Your Co-Founder Disagrees With Your Call

Kill/scale decisions create co-founder tension. One of you is emotionally attached to a test. The other wants to pull the plug. This isn't adversarial—it's usually different risk tolerances.

Ego battles disrupt progress. Personal attachment clouds judgment. You need a protocol to resolve disagreements quickly without letting indecision drain your budget.

The 72-hour rule for contested decisions

When co-founders disagree, the person advocating for the test gets 72 hours to present new data or a revised hypothesis. After 72 hours, you make the call using the matrix. No more debate.

Example: your co-founder wants to keep running Instagram ads despite Quadrant 1 data. They have 3 days to show why trajectory will improve. If they can't, you kill it.

This rule prevents analysis paralysis while respecting both perspectives. Three days of discussion beats three weeks of indecision.

How to separate ego from evidence

The ego trap: "I chose this channel, so killing it means I was wrong." This is human nature. The workaround is simple: pre-define your kill criteria before you spend a dollar.

Flatiron Health's decision matrix approach removes emotion from the call. Document the decision criteria in writing. "We'll kill this test if CAC doesn't drop below $150 after 4 weeks." Agreed upfront. No room for ego later.

This creates accountability. When the data says kill, you kill. No renegotiation.

Speed Beats Perfection (But Only If You're Watching)

Rapid kill/scale decisions compound over time. Founders who decide fast test more channels and find winners sooner. Velocity of decision-making is a competitive advantage.

But speed only works if you're monitoring CAC and payback weekly. Don't set and forget. If you're not tracking where leads come from and how much they cost, you're flying blind. Lead Recorder gives you real-time visibility into which channels are working so you can make fast, confident decisions.

Final action step: review all active tests this week using the matrix. Kill at least one underperformer. Double budget on your best performer.

Killing tests isn't failure. It's the fastest path to finding your scalable growth channel.