How to prioritize growth experiments in a small business

If you run a small business and have started testing things to grow, the problem shows up fast. There are plenty of ideas: change the homepage headline, show indicative prices, call back the quotes that never replied, open a new Google listing, try an ad. There is little time to test them. Deciding which one goes first is the part almost nobody does with a method.

This article is about that: ordering. What growth hacking is and how an experiment cycle works is covered in growth hacking for SMEs. Here we assume you already want to experiment and go straight to choosing what to test, on what basis, and how often to review it.

Before you score anything: where your business gets stuck

The best idea in the world is useless if it targets a part of the business that isn't holding anything back. So the first filter isn't a formula. It's a question: where are people dropping off today?

Walk through a customer's path and look for the stretch that narrows the most. Few website visits. Visits, but hardly anyone asks for a quote. Quotes, but few close. Customers who buy once and never come back. Each one needs different experiments, and mixing them spreads the effort thin.

Pick one bottleneck for the next few weeks and park the rest. If you aren't sure in what order to build the pieces (brand, website, visibility, content, automation), that is an earlier decision, and we cover it in how to prioritize a growth system in 5 stages. Prioritizing stages and prioritizing experiments are different jobs: the first decides what you build, the second what you test on top of it.

A backlog of hypotheses, not hunches

Put every idea in one place. A spreadsheet is enough. Each row should have:

• The hypothesis, in one sentence: "if we do X, Y will happen, because Z".
• The bottleneck it targets.
• The metric you will check to know whether it worked.
• What it takes to test: hours, money, who.

The "because" is what separates a hypothesis from a whim. "Try a video on the homepage" isn't a hypothesis. "If the homepage explains how we work, more visitors will reach the form, because right now the page doesn't say what happens after you get in touch" is one. You can check it, and if it fails you learn something.

ICE, RICE and PIE in plain terms

They are three ways to give each idea a score so you can rank them. None is better in the abstract. All of them depend on honest scores.

ICE scores three things, usually from 1 to 10: impact (how much it would move the metric if it works), confidence (how sure you are it will work) and ease (how little it costs to test). You multiply or average them. It is the quickest, and the best fit for a small team.

RICE adds reach: how many people the change affects in a given period. The formula is reach times impact times confidence, divided by effort. It helps when ideas touch audiences of very different sizes, for example a page everyone sees versus an email that goes to part of your customer list.

PIE looks at potential for improvement, importance of the page or step, and ease. It is built mainly around web pages, and makes sense if most of your experiments are page changes.

All three share the same flaw: they look objective and they aren't. An impact score of 8 is an opinion with a number on it. If the person scoring is the person who proposed the idea, the score drifts up on its own. Use them to structure a conversation, not to end it.

How to estimate with little data without fooling yourself

A small business rarely has enough history to calculate impact. You have to estimate, and you can estimate well or badly.

For impact, think about the ceiling. If the idea improves a page few people see, its maximum impact is already capped, however good it is. Look at how many people pass through that step today before imagining how much it could improve.

For confidence, ask what evidence you have. A complaint customers keep repeating, a call where someone got lost in the form, a figure from Search Console: those raise confidence. "I read it on a blog" or "it works for a competitor" keeps it low.

For effort or ease, count everything: the hours to make the change, the hours to measure it, the time of whoever has to sign it off, and the work of undoing it if it goes wrong.

One simple trick: have two people score separately, then compare. Wherever the scores are far apart, there is an assumption nobody has said out loud.

With low traffic, not everything is an A/B test

An A/B test needs enough conversions in each version to tell a real improvement from chance. A local business site that gets few enquiries a month can take a very long time to get there, and meanwhile the season, the ads and the competition all change. Before you launch one, run your numbers through a sample size calculator. If the answer is months, find another way to test.

Alternatives that do work at low volume:

• Before and after, with the date written down and a comparable period. It's less clean than an A/B test, which is why you shouldn't touch anything else while it runs.
• Qualitative tests: ask five customers or acquaintances to try requesting a quote on your site while you watch. Whatever repeats is a strong clue.
• Ask directly: a "how did you hear about us?" field on the form, or a question when you close a sale.
• Test away from the website: a new script for follow-up calls, a different response time on quotes. The volume is small, but the effect usually shows in every conversation.

Set the duration and the success criterion before you start. If you decide them after seeing the data, you will find what you were looking for.

To measure anything you need contact tracking that works. Which tools you need and which you can skip is in how to build a growth marketing stack.

An example with made-up scores

To show the method, a hypothetical case. The scores are examples, not from any client.

A home renovation company gets website visits but few quote requests. That's its bottleneck. It has three ideas:

The photos win, and not because they're the most ambitious idea. They target the step where people drop off, there is evidence (several customers asked about past work before calling) and they can be tested in an afternoon. Rebuilding the site may make sense, but it's a project, not an experiment. The ads would bring more visits to a page that already doesn't convert, so they hit the wrong stretch.

The ritual: short, and written down

Without a fixed review, the backlog turns into a drawer. Half an hour every week or every two weeks is enough, always with the same people:

1. Review open experiments: what the data says, whether it's time to decide.
2. Decide on the ones that are finished.
3. Rescore the backlog with what you've learned.
4. Pick the next one, only if there is room.

One or two experiments open at a time is enough for a small team. More than that and nobody really follows them.

Keep a decision log. One line per experiment: date, hypothesis, what changed, result, decision and why. A few months from now someone will propose the same idea under another name, and the log saves you from running it twice. It also shows whether your confidence scores are right or whether you always overshoot.

When to kill, iterate or scale

At the end of each test there are only three outcomes.

Kill it if the result is clearly worse or flat and there is no obvious variant that tests something different. Write it down and move on. A discarded experiment with a log entry is not wasted time.

Iterate if there is a signal but it's weak, or if you saw something unexpected in the qualitative tests. Change one thing from the previous version.

Scale it if it works: make the change permanent, extend it to other pages or services where it fits, and look at the bottleneck again. Sometimes fixing one stretch moves the jam to the next one. That's a good sign, and it tells you what to score next.

Who runs this work is a separate decision. If you're weighing your own team against outside help, growth marketing agency vs in-house team compares both. And if your experiments touch sales more than the website, it pays to have a clear sales strategy first.

Frequently asked questions

Which prioritization model is best for a small business?

ICE, nearly always. It's fast and forces a conversation about impact, confidence and effort. RICE is worth it when your ideas reach audiences of very different sizes. What matters is sticking to one and revising the scores as you learn.

How many experiments should I run at once?

One or two, if you're a small team. That way each one has someone following it and their effects don't mix in the same metric.

Can I experiment if my website gets very little traffic?

Yes, with different tools. Usability tests with a few people, questions to customers, well-documented before-and-after changes and experiments on the sales side. Keep A/B tests for when the volume allows it.

How often should I review the experiment backlog?

Every week or every two weeks. What matters is that the review doesn't get skipped and every decision is written down.

Where to start

If you don't know where your bottleneck is, or your tracking doesn't let you see where people drop off, start there. In the free audit we review your website, your visibility and how enquiries reach you, and tell you what to fix before you test anything. Different Growth is a digital studio based in Spain, and we build that foundation as a growth system.

Request a free audit