The growth experiment log: documenting what you've tried.

A growth experiment log is a simple record of every marketing change you make, what you expected it to do, and what actually happened. It keeps you from running the same failed test twice. It keeps you from claiming credit for luck. And it turns scattered activity into a track record you can read.

Most owner-operated businesses do not keep one. They try a new headline, change the pricing page, run a round of paid ads, and then six months later nobody remembers what moved the number. The work happened. The learning did not. A log fixes that for the cost of a few minutes after each test.

Why bother logging experiments at all?

You log experiments so your business stops paying the same tuition twice. Every test teaches something. Without a record, the lesson evaporates, and next quarter someone runs the same idea with the same result and the same surprise.

I have sat in planning meetings where a team pitched an idea with real energy, and I had to say we tried that eighteen months ago. It cost $4,000 and did nothing. Nobody was wrong to suggest it again. They simply had no memory to check against. That is the gap a log closes.

There is a second reason that matters more as you grow. Memory flatters the person doing the remembering. We remember the wins and forget the duds. We credit our instincts and blame bad timing. A written log is honest in a way that memory is not. It shows the full ledger — the tests that worked, the ones that flopped, and the ones that taught you nothing because they were badly run.

That honesty compounds. After a year of logging, you can see patterns. Email tends to beat paid for this audience. Long-form pages convert better than short ones. Discount offers bring price-shoppers who churn. None of those insights come from a single test. They come from reading twenty tests side by side.

What goes in a single log entry?

A single log entry needs six things: the hypothesis, the change, the metric, the baseline, the result, and the decision. If any one is missing, the entry is harder to learn from later.

Start with the hypothesis. Write it as a sentence, not a vibe. "We think moving the phone number above the fold will increase call clicks because mobile visitors are not scrolling to find it." That sentence forces you to say what you expect and why. It also makes the test falsifiable. If call clicks do not move, the hypothesis was wrong, and that is useful.

Next, the change itself. Be specific enough that someone could reproduce it. "Added click-to-call button in the header on all service pages" beats "improved the header."

Then the metric. Pick one primary metric before you start. Call clicks. Form fills. Booked consultations. If you pick the metric after the results come in, you will pick whatever looks good, and you will fool yourself.

Record the baseline next — the number before the change. Without it, the result is a figure floating in space. "Call clicks went to 140" means nothing. "Call clicks went from 90 to 140 over the same four-week window" means something.

Then the result, with the window it covered. And finally the decision: keep, kill, or run again with a change. The decision is the point. An experiment that ends without a decision is just a thing that happened.

How do you keep it from becoming a chore?

You keep the log usable by making it smaller than feels responsible. A spreadsheet with six columns beats a document nobody updates. The best log is the one you actually fill in, not the one with the prettiest template.

I use a simple sheet. One row per experiment. Columns for date, hypothesis, change, metric, baseline, result, and decision. That is it. No dashboards, no automation, no tool to learn. The friction of a fancy system is the reason most logs die by week three.

Set a rule for when an entry gets written. Mine is simple: the entry opens the day the test launches and closes the day I read the result. Two touches. The opening entry takes two minutes. The closing entry takes three. If a test does not deserve five minutes of writing, it probably does not deserve to run.

One more discipline. Close entries on a schedule, not when you feel like it. Pick a day each week. Read every open experiment that has reached its window. Write the result. Make the decision. A standing thirty-minute block on Friday is enough to keep a log alive for a small business running three or four tests a month.

This is the same rhythm we bring when we run marketing for a client. A Fractional CMO lives or dies by whether the experiments get logged and read, because the whole point of fractional help is that the thinking stays in the business after the engagement ends. A log is how you keep the thinking.

What does a good log tell you after six months?

A good log tells you where your effort actually pays and where it only feels like progress. After six months, you can rank your channels by return, spot the tests that keep failing, and stop defending habits that the numbers never supported.

Here is the pattern I look for when I read a mature log. First, the hit rate. If nine of ten experiments move nothing, you are testing too small or testing the wrong things. If every experiment wins, you are not testing anything risky enough to learn from. A healthy log has a mix — some wins, some flops, some draws.

Second, the direction of the wins. Group them. Did the wins cluster around one channel, one message, one audience? That cluster is a signal. It is telling you where to spend the next quarter.

Third, the cost of the flops. Add up what the failed tests cost in money and time. If that number is small, good — cheap failure is the whole idea. If one flop cost $12,000, ask why you bet that much before you had evidence. Big bets belong after small tests have pointed the way, not before.

We worked with McShanes Solicitors on exactly this kind of discipline — small tests on the pages that mattered, logged and read, before any large change went live. The log kept the work honest. It also gave the partners a record they could read without a marketer in the room.

What experiments are worth logging, and which aren't?

Log any change you expect to move a number you care about. Skip the changes you are making for taste, compliance, or basic repair. The log is for learning about your market, not for tracking every edit.

Worth logging: a new headline on a service page, a different call-to-action, a change in how you ask for reviews, a new paid campaign, a price change, a different intake form, a fresh email sequence. Each of these has a hypothesis and a metric. Each can win or lose. Each teaches you something about who your clients are and what moves them.

Not worth logging: fixing a typo, updating your hours, swapping a stock photo nobody complained about, correcting a broken link. These are maintenance. They should happen. They do not need an entry because there is nothing to learn — you are not testing a belief about the market.

The gray area is the interesting part. A full homepage redesign is tempting to log as one experiment. Do not. A redesign changes thirty things at once, and when the number moves you will not know which of the thirty did it. Break it into separate tests where you can, or accept that the redesign is a judgment call, not an experiment, and log it as such.

Knowing which lever to pull is its own skill, and it is one of the reasons owners hesitate between hiring a part-time strategist and hiring an agency. We wrote about that choice in Fractional CMO vs agency: the difference that matters, and the short version is that the log belongs to whoever owns the strategy — not whoever runs the tactics.

What this log won't do for you

A log will not generate good ideas, and it will not tell you how much to bet. It records what you tried and what happened. The judgment about what to try next, and how much to risk, still sits with you. If your hypotheses are weak, a beautiful log will faithfully document a year of weak tests.

It also will not protect you from tracking the wrong metric. If you log form fills but the business runs on booked consultations, the log will cheerfully report success while revenue sits flat. Pick the metric that ties to money. If you cannot tie it to revenue, do not celebrate it.

And a log is not a substitute for knowing whether you even need this discipline yet. A business doing two tests a year does not need a system. It needs more tests. If you are not sure which side of that line you are on, we covered the honest version of that question in when you actually need a Fractional CMO — and when you don't.

Start the sheet this week. Six columns. One row per test. Open the entry the day you launch, close it the day you read the result, and make a decision every time. In three months you will stop guessing about what works. In six, you will have a record your whole business can read. That is enough.