The Scientific Method (Business Edition)

It put a man on the moon and photographed a black hole. The same method will build your business — if you run it properly, one variable at a time.

June 21, 20269 min read

The scientific method is the most powerful idea humans have ever had. It put a man on the moon, photographed a black hole, and built modern medicine. If it can do that, it can get you to a point where you are successful in business. I know, because it is exactly how I did it.

Most people treat business as a matter of taste, instinct, or luck. It is not. It is a search problem. You are looking for a combination of variables that produces a result, and the scientific method is the only reliable way anyone has ever found to search a space like that without fooling themselves.

The cycle

There are four steps, and they run in a loop:

Formulate hypothesis. Define your variables and assemble them into a version — a v1 — you can test.

Test hypothesis. Expose that version to the market to see how well it actually works.

Observe results. Gather the data and read the effect of that combination of variables.

Iterate hypothesis. Use the data to change one thing, and make the next version slightly better.

                     ╔══════════════════════╗
                     ║    ● 1. FORMULATE    ║
        ┌───────────▶║      HYPOTHESIS      ║────────────┐
        │            ╚══════════════════════╝            │
        │                                                ▼
┌────────────────────┐                      ┌────────────────────┐
│      4. ITERATE    │                      │        2. TEST     │
│     HYPOTHESIS     │                      │     HYPOTHESIS     │
└────────────────────┘                      └────────────────────┘
        ▲                                                │
        │            ┌──────────────────────┐            │
        │            │       3. OBSERVE     │            │
        ┘────────────│        RESULTS       │◀───────────┘
                     └──────────────────────┘

Throughout this piece I will use cold calling as the worked example, because it is concrete and most people understand it. But the method is identical whether you are testing an outbound process, an offer, a landing page, or a hiring funnel.

One thing to be clear about up front: the scientific method operates at the micro level. Your business is a tree of variables. At the top sit the big things — the avatar, the offer, the outbound process. Each of those breaks down into smaller variables, and those into smaller ones still. You do not experiment on the whole tree. You experiment at the bottom, shuffling the small Level-3 variables to find the Level-2 combinations that hit your target.

                                     ┌───────────────┐
                                     │  BUSINESS v1  │
                                     └───────────────┘
           [ NICHE ]                         │                        [ PROBLEM ]
         ┌───────────┬───────────┬───────────┬───────────┬───────────┬───────────┐
         │           │           │           │           │           │           │
    ┌─────────┐ ┌─────────┐ ╔═════════╗ ┌─────────┐ ┌─────────┐ ┌─────────┐ ┌─────────┐
Lv.1│  Avatar │ │  Offer  │ ║ Outbound║ │ Content │ │  Sales  │ │ Solution│ │ Identity│
    └─────────┘ └─────────┘ ╚═════════╝ └─────────┘ └─────────┘ └─────────┘ └─────────┘
                                 │
                 ┌───────────────┼───────────────┐
                 │               │               │
          ┌────────────┐  ╔════════════╗  ┌────────────┐
Lv.2      │ Cold Email │  ║  Cold Call ║  │  Conv. Msg │
          └────────────┘  ╚════════════╝  └────────────┘
                                 │
Lv.3                 gatekeeper · hook · pitch

    DNA (information) · critical information variants
    ░░▒▒▓▓░░▒▒▓▓░░▒▒▓▓░░▒▒▓▓░░▒▒▓▓░░▒▒▓▓░░▒▒▓▓░░▒▒▓▓░░▒▒▓▓░░▒▒▓▓░░▒▒▓▓░░▒▒▓▓░░▒▒▓▓░░▒▒▓
DNA ▒▒▓▓░░▒▒▓▓░░▒▒▓▓░░▒▒▓▓░░▒▒▓▓░░▒▒▓▓░░▒▒▓▓░░▒▒▓▓░░▒▒▓▓░░▒▒▓▓░░▒▒▓▓░░▒▒▓▓░░▒▒▓▓░░▒▒▓▓░
    ▓▓░░▒▒▓▓░░▒▒▓▓░░▒▒▓▓░░▒▒▓▓░░▒▒▓▓░░▒▒▓▓░░▒▒▓▓░░▒▒▓▓░░▒▒▓▓░░▒▒▓▓░░▒▒▓▓░░▒▒▓▓░░▒▒▓▓░░▒

Step 1 · Formulate the hypothesis chain

1.1 Define the key variables

There are usually two to four of them at the level you are working on. For cold calling, the variables are:

gatekeeper process · hook · pitch

1.2 Define the metrics and KPI

Decide how you will measure success before you start. For cold calling, the headline metric is appointment booking rate, supported by the gatekeeper bypass rate, the pitch rate, and the conversation rate. If you cannot measure it, you cannot iterate on it.

1.3 Establish the sample size

How many times will you run the test? For cold calling, you might commit to 1,000 calls. A larger sample gives more validity and protects you from reading noise as signal.

1.4 Establish latency and time frame

Over what period will you run the test and reach that sample size? Fix it now, so you are not tempted to call the result early.

1.5 Make the test airtight

Ask what could interfere with the test and skew the results through no fault of the variables — cold calling over the Christmas week, a bad phone connection, a broken dialer. Remove those threats before you begin.

1.6 Create your version variables

Build the specific versions that make up this experiment:

gatekeeper v1 + hook v1 + pitch v1 = Cold Calling Process v1

1.7 Establish the constants

Define everything that affects the outcome but is not one of the variables you are testing — phone number, tone of voice, speech pacing, lead source, time of day. These must be held constant for the entire run. If you change your tonality and your time of day while you change the hook, and the booking rate moves, you have NO idea which change caused it.

This is Ceteris Paribus — Latin for “all other things held the same.” It is the single rule that makes the whole method work. You may change ONE variable at a time. EVERYTHING else stays constant. Test two things at once and you learn nothing, because a result with two causes has no lesson in it. So you test ONLY one variable at a time.

Step 2 · Test the hypothesis

2.1 Run the test

Get to work and let it run. The hard part here is not the work — it is the patience.

2.2 Do nothing

Do NOTtouch the test while it runs. No changes, no impulsive tweaks, no “quick improvements” halfway through. The moment you change a constant mid-run, you have corrupted the experiment and wasted the sample.

2.3 Log the data

Keep a daily log of the metrics in a spreadsheet, attributed to its source. Boring, daily, accurate. The log is the only thing that will tell you the truth at the end.

Step 3 · Observe the results

3.1 Verify the data

Before you decide anything, re-check the data and make sure it is solid. Decisions made on dirty data are worse than no decisions at all, because they feel justified.

3.2 Determine success

Read the result honestly. Did the experiment land within KPI? Did you move from baseline toward the goal, or away from it? The data does not care how hard you worked. Let it speak.

Step 4 · Iterate the hypothesis

4.1 Re-record the key variables and result

If you are not yet at KPI, note exactly what this version was and what it produced:

gatekeeper v1 + hook v1 + pitch v1 = ABR 2%

4.2 Pick the needle-mover

Decide which single variable, changed well, will have the biggest impact. Maybe your gatekeeper process already gets you to the decision-maker every time and your hook lands cleanly — then the pitch is the needle-mover. Use judgment here, but only ever pick one.

4.3 Recognise the new constants

The two variables you did notpick now become constants. You cannot touch them, or you lose the ability to read the needle-mover's true effect. Ceteris Paribus again: one variable moves, the rest hold.

4.4 Pick the modification level

Decide how much to change the needle-mover. There are three settings, chosen by how far you are from KPI:Slight — out of KPI but close (target 3.5%, got 2.3%). The next version keeps most of this one.Moderate — out of KPI but alive (target 3.5%, got 1.3%). The next version keeps some of this one.Complete — dead in the water (target 3.5%, got 0.3%). The next version starts fresh.

4.5 Modify and rebuild

Apply the change and assemble the next version:

gatekeeper v1 + hook v1 + pitch v2 = Cold Calling Process v2

4.6 Repeat Step 1

Run the new version through the exact same test. Before you draw any conclusion, confirm the five rules held:— The full sample was collected and attributed.— The test stayed airtight.— Every constant stayed constant.— You changed nothing mid-run.— You logged the data daily, accurately.

This is how you build everything in a business. Not by taste, not by guru advice, not by a single A/B test someone sold you as a shortcut. By running the loop, one variable at a time, until the numbers land where you need them.

Follow the eighteen steps properly and you cannot fail — and I mean that literally. The only remaining variable is time. Give the method enough of it to shuffle the variables, and it works. It always works.