How to Start an MLM Company

Most failed direct selling launches did not skip steps. They did them in the wrong order — designed the plan before knowing the margin, chose software before finalising the plan, recruited the field before piloting the commission run — and paid for each inversion twice.

What follows is the sequence, with the reason each step precedes the next.

1. Product economics, to a number

Before anything else: gross margin per unit, after cost of goods, fulfilment, payment processing, returns and chargebacks.

Not list price minus cost. The real number.

Everything downstream is a claim on it. A 78%-margin digital subscription supports a plan that would liquidate a 55%-margin physical product. You cannot design compensation without this figure, and companies that try end up designing the plan they wish they could afford.

Also settle reorder rate expectations. A company whose revenue depends on reorders and a company whose revenue depends on new enrollment are different businesses with different plans and different regulatory exposure.

2. Legal review — before the plan is written, not after

The instinct is to design the plan and then have it reviewed. Reverse it: talk to counsel about the shape first.

Some structures carry inherent exposure that no configuration resolves. A board plan paying on board completion, or a monoline paying as positions join beneath you, compensates entry rather than product movement — and that is structural. Learning this after the plan document has gone to the field costs field trust, which is not recoverable with an apology.

What to get reviewed at this stage:

• The plan family and its payout trigger.
• Which market(s) you will operate in — the specifics differ. See is MLM legal in the USA and is MLM legal in South Africa.
• Whether your programme falls within any state business opportunity registration requirement.
• Your product category’s own regime — supplements, cosmetics and health claims bring additional regulators.

3. Design the plan, and model the total

Now pick a family and set percentages. The decision framework works through margin, sales motion and market in order.

Then do the arithmetic that most first-time founders skip. A unilevel at 5% per level over five levels does not cost 5% of volume — it costs roughly 22.6% at full build, because every unit of volume pays every upline inside the paid depth. Add a matching bonus and a rank tier and a plan sketched as “5% per level” is a 35% plan.

Model total payout across every component at three organisation sizes — early, mid, full build. Plans routinely pass the early test and fail at full build, because depth multiplication only becomes visible once the organisation is deep.

Run it in the plan calculator, then subtract the result from step 1’s margin along with support and platform cost. Positive at all three sizes, or the plan changes now rather than in year two.

4. Write the documents

With the plan modelled and reviewed:

• Compensation plan document — the field-facing version, matching the configuration exactly. A discrepancy between the document and the engine is a dispute waiting to happen.
• Distributor agreement — plain language, as the CPA requires in South Africa and as good practice everywhere.
• Policies and procedures — including a claims policy your field acknowledges.
• Income disclosure statement — typical results, basis stated.
• Privacy policy — the document your consent checkbox references, and the basis for POPIA and GDPR processing.
• Buyback policy — unsold resalable inventory, stated window, stated percentage.

5. Choose the platform

Only now, because the plan is what you are configuring.

Five questions to insist on demonstrations for, not descriptions of:

Is the plan configuration or code? If switching a 3x3 matrix to a 5x5, or adding a generation bonus, needs development work, then every future plan revision does too — and you will revise repeatedly in the first two years.

Can a closed period be reproduced after the rules change? Change a rank threshold, re-run last month, get the same numbers you paid. Ask to see it happen. This is the single property most likely to be described convincingly and implemented poorly.

Are sponsor and placement stored separately, with history? In a binary or matrix they diverge constantly, and conflating them makes every future sponsorship-bonus question unanswerable. See upline and downline.

Is every order classified retail-customer versus participant at order time? You cannot report a retail ratio you never captured, and it is the first number a regulator, a payment processor or an acquirer asks for.

Do reversals land in the right period? Returns, chargebacks and cooling-off cancellations have to reverse the commission they generated, in the period it belongs to.

What each looks like implemented is on the MLM software page and the compensation plan software page.

6. Pilot the commission run — with real people

Not a test dataset. A small real field: ten to thirty distributors, real orders, real returns, a real period close, real payouts.

What a pilot catches that a demo cannot:

• Qualification rules that read clearly and behave differently.
• A rank threshold that nobody can actually reach, or that everybody reaches immediately.
• Reversals landing in the wrong period.
• Commission statements the field cannot understand — which becomes your support load.
• The gap between the plan document and the configuration.

Run at least two full periods. One period does not exercise carry-forward, compression, or rank maintenance, and those are where the problems are.

7. Launch to the field

With the plan modelled, reviewed, documented, configured and piloted.

And with these instrumented from day one, because retrofitting them onto a live field is the expensive version:

• Retail versus participant volume reporting, per period.
• Qualification caps on self-purchase, enforced by the engine.
• Buyback tracking with commission reversal.
• An auditable payout trail: who was paid, on what volume, under which rule version.
• Income claim monitoring across your field’s public posts.

The inversions to avoid

Stated plainly, since each is a specific and common failure:

Plan before margin. You design compensation you cannot fund and discover it in year two.

Software before plan. You pay to configure a plan that then changes, or worse, you let the platform’s limitations dictate the plan.

Field before pilot. Your first commission run is a live incident with an audience.

Launch before legal. The most expensive of the four, because the remedy — changing the plan — takes income away from people who built under the old rules.