Lessons from running SaaS marketing tests

Most marketing teams I've worked with say they "run experiments." When I look closer, what they actually do is launch random campaigns, glance at the numbers a week later, and move on. There's no real test. There's no real learning. The next quarter looks the same as the last one.

Running real tests isn't hard, but it does need a small amount of structure. Here's the process I've used at LeadsChilly and GrowMeOrganic that actually moves the needle.

Start with a real question, not a vague idea

"Let's try LinkedIn ads" is not an experiment. "If we run LinkedIn ads targeting heads of sales at 50–200 person SaaS companies, can we get a demo at under ₹3,000?" is an experiment. The second version tells you what you're testing, who you're testing it on, and what counts as a win.

Write down the hypothesis before you spend a rupee

I keep a simple sheet with five columns: what we're testing, why we think it'll work, what success looks like, what we'll do if it works, and what we'll do if it doesn't. That last column is the one people skip. Knowing the kill condition in advance saves you from wasting another month on something that's clearly not working.

Run one variable at a time

If you change the audience, the creative and the landing page all at once, you'll never know what actually moved the result. Change one thing, hold the rest steady, and you'll learn something you can use again.

• Test the audience with the same creative
• Test the creative with the same audience
• Test the offer with the same creative and audience

Give the test enough data to be honest

I see people kill campaigns after 50 clicks and call it a failure. That's not a test, that's noise. As a rough rule, wait for at least 100 conversions or a clear pattern before deciding. Faster decisions on small data are how teams accidentally kill the winners.

Be ruthless about what you scale

Once a test wins, the temptation is to spread the budget across everything that looks promising. Don't. Pick the one or two clear winners and put real money behind them. The wins fund the next round of tests. The losers get killed without guilt.

Most growth isn't from finding a magic channel. It's from running boring tests honestly and doubling down on the few that work.

Keep a running log

Every test goes into one document — what we tried, what happened, what we learned. After a year, this document is more valuable than any course or playbook. You stop repeating the same mistakes, and new team members get up to speed in days instead of months.

Marketing experiments are not glamorous. They're spreadsheets, screenshots and small wins. But that's also how compounding growth actually happens.