Run a marketing experiment in 2 weeks (free template)
By William Walczak · Hiilite Creative Group · June 2026
TL;DR
Most small business owners skip testing because they picture a lab, a data scientist, and a sample size in the thousands. You don’t need any of that. A useful marketing experiment is: one clear question, the smallest test that answers it, two weeks, and a decision at the end. This guide walks you through it step by step. The copyable template is at the bottom.
Why experiments matter more than best practices
Best practices are averages. They describe what worked for someone else, in a different market, with a different audience, at a different moment. Running an experiment tells you what works for your business, right now.
Stefan Thomke, Harvard Business School professor and author of Experimentation Works, makes the point plainly: companies that build a culture of testing — where small, fast experiments are the default response to uncertainty — outperform those that rely on expert opinion or industry norms. Not because the experiments are complex. Because they’re honest. They produce evidence instead of assumptions.
Ron Kohavi and colleagues make a similar case in Trustworthy Online Controlled Experiments — even online, even with small samples, even with imperfect tools, a deliberate test beats a confident guess.
For a small business owner with limited time and budget, that framing is liberating. You don’t need perfection. You need a process.
The 2-week experiment loop
This is the loop the Hiilite Growth Mapping framework uses under the Retention domain. Run it once and you’ll run it again.
Step 1: Write one hypothesis (10 minutes)
A hypothesis is not a goal. “I want more email opens” is a goal. A hypothesis is testable: if I change X, then Y will happen, because Z.
Example: “If I add a deadline to my email subject line, open rate will increase by at least 5 percentage points, because urgency signals relevance to subscribers who are on the fence about opening.”
Write it in that structure. If you can’t complete the “because Z” part, your hypothesis is not specific enough. Keep narrowing until you can.
One hypothesis per experiment. Changing two things at once means you won’t know which one moved the needle.
Step 2: Identify the one metric that decides the winner
Pick the metric before you run the test — not after. Choosing it afterward is how results get twisted to confirm what you hoped.
The metric should be: – Directly tied to the thing you changed (not a downstream proxy) – Measurable with the tools you already have – Something you can read in 14 days without needing a statistician
Examples: email open rate, landing page click-through rate, form completion rate, booking rate, reply rate on an outreach sequence.
One metric. Write it down.
Step 3: Define the smallest test
This is where most owners over-engineer it. You do not need two identical audiences split at random. You need a condition and a baseline.
If you’re testing an email subject line, send the new version to a small segment and compare the open rate to your average. If you’re testing a landing page headline, swap it and watch the click-through rate for two weeks.
The goal is a test small enough that you can run it now, with the resources you have, without disrupting your normal operations. You can add rigor later. Start with signal.
Write down: – What you will change – What you will keep the same – How many people (or impressions, or sessions) you expect to get in two weeks
If the expected sample is under 50, you probably won’t get a conclusive result — but you’ll still learn something directional. Flag it as exploratory.
Step 4: Set a duration and a decision rule
Two weeks works for most small business contexts because it captures a full work cycle without drifting into seasonal noise.
Before you launch, write down the decision rule:
“If the metric improves by [X]% or more, we adopt the change. If it doesn’t, we revert or explore a different hypothesis.”
The threshold matters. A 1% improvement in email open rate may not be worth the operational change. A 15% improvement probably is. Set the bar at a level that would actually change your behavior.
Step 5: Run it. Don’t touch it.
Launch the test on day one. Do not adjust anything mid-experiment. Do not extend it because it’s not looking good on day ten. The discipline of letting it run is where most tests break down.
Block a 30-minute slot in your calendar for day 14 to read the result.
Step 6: Read the result and make a decision
On day 14, compare the metric to your baseline and your decision rule.
Three outcomes are possible: 1. It worked. The metric cleared the threshold. Adopt the change. 2. It didn’t move. The metric stayed flat or declined. Revert and rethink the hypothesis. 3. Inconclusive. The sample was too small to read. Treat it as directional signal, not a verdict.
Write down what you learned. One sentence is enough: “Adding a deadline to subject lines improved opens by 8% — adopt.” Or: “Deadline subject lines made no difference for our audience — our subscribers may not be deadline-sensitive.”
That sentence is the asset. It saves you from running the same test again in six months.
Step 7: Feed the result into the next experiment
The output of each test is the input to the next hypothesis. What did you learn about your audience? What question does that raise?
Kalaignanam et al. (2021) describe marketing agility as the capacity to shorten feedback cycles — to sense a signal, act on it, and recalibrate faster than the competition. This loop is that capacity, running at the pace of a small business.
The template (copy this)
[LEAD MAGNET: experiment template]Below is the copyable version. Paste it into any doc, spreadsheet, or project tool. Fill in one row per experiment.
## Marketing Experiment Log
| Field | Details |
|------------------|----------------------------------------------|
| Experiment # | |
| Date started | |
| Date ended | |
| Hypothesis | If I [change], then [metric] will [direction] by [amount], because [reason] |
| What changes | |
| What stays same | |
| Primary metric | |
| Baseline value | |
| Decision rule | Adopt if metric improves by [X]% or more |
| Expected sample | |
| Result (day 14) | |
| Decision | Adopt / Revert / Inconclusive |
| Learning (1 line)| |
| Next hypothesis | |
Tips for using the log: – Keep every experiment in the same document. The pattern across ten rows is more valuable than any single result. – Mark inconclusive experiments clearly. They’re not failures — they’re sample-size flags. – Share the log with anyone who runs marketing for your business. It’s a live record of what your audience responds to.
Frequently asked questions
Do I need a control group?
Not always. If you have enough historical data to establish a reliable baseline (at least 30 days of clean readings on the metric), comparing against the baseline is a reasonable substitute for running a simultaneous control. If your data is patchy, try to run a split: send version A to one segment and version B to another in the same send.
What if my audience is too small?
Sample size under 50 means results will be noisy. That doesn’t mean the experiment is pointless — it means you should label the result as directional, not conclusive. Run the same experiment again in a larger window, or wait until your audience grows. Some signal beats no signal.
How many experiments should I run at once?
One per marketing channel, maximum. Running two experiments on the same channel simultaneously corrupts both — you can’t isolate what moved the needle. Parallel tests are fine if they touch different channels (email and landing page, for example) and the metrics don’t bleed into each other.
What counts as a “marketing experiment” for a small business?
Anything testable: email subject lines, call-to-action copy, pricing presentation, offer framing, page headlines, ad creative, follow-up timing, onboarding sequence length. If you can change one thing and measure the downstream effect, it qualifies.
Do I need special software?
No. Your email platform’s open-rate report, your website’s analytics, and a simple spreadsheet are enough to run the experiments in this guide. Use the tools you already have. Upgrade your toolset only after you’ve built the habit.
How this fits into the Growth Mapping framework
This experiment loop is one module inside the broader Retention pillar of the Growth Mapping framework. Retention sits alongside Recruitment and Revenue as one of the three operational domains (the 3Rs) where growth compounds.
A single experiment is a point-in-time test. A running log of experiments is a learning system — what Thomke calls a “culture of experimentation” and what Kalaignanam et al. describe as the feedback cycle at the core of marketing agility. The loop doesn’t just tell you what worked last month. It tells you how your audience is changing, and it gets more accurate the longer you run it.
For a deeper look at how this connects to the Sense → Seize → Transform model and the platform that automates the measurement layer, see The Agentic Agency paper and the customer retention guide.
Get the template as a formatted download
[LEAD MAGNET: experiment template]The template above is free to copy. If you want a formatted, printable version with a guided worksheet for the hypothesis-writing step, download it here.
[Download the 2-week experiment template] — no cost, one click.
Ready to close the measurement loop automatically?
Running experiments manually works. What works better is a system that reads your live data, flags where the gap between your goal and your current performance is widest, and tells you which experiment to run next.
That’s what the Hiilite Agentic Advisor does. It connects your revenue, marketing, and client data into a single loop — so the measurement layer runs continuously, not just when you have a free afternoon.
[Book a discovery call] to see how it works for a business like yours.
About the author
William Walczak, MBA is the founder and CEO of Hiilite Creative Group (est. 2014), a Kelowna, BC marketing agency. He is a PhD candidate in Interdisciplinary Graduate Studies at UBC-Okanagan (dissertation: “Growth Mapping: A Mixed-Method Study of Growth Hacking”), and holds an MBA and Engineering degree from UBC and Simon Fraser University. His research interests include consumer behavior, machine learning, predictive analytics, and consumer experience.
He was named Marketing Strategy CEO of the Year 2023 (BC) by CEO Monthly and was recognized in the Daily Courier’s Top 40.
Peer-reviewed publication: Walczak, W., Li, E. P. H., & Nelson, S. (2024). “Logarithm: A Cinematic Exploration of Time.” Journal of Customer Behaviour.
Connect: hiilite.com/team/william-walczak · LinkedIn · Google Scholar