Affiliate Marketing

Advertising was for a long time a trade where you paid before you saw the result. So much for a placement, so much per thousand impressions, and only much later could anyone check what that money brought in. A well-funded company can absorb that. An early-stage one cannot.

Pagedeck went through the same thing. Ravi ran a banner on a work-tools publication for a month at a cost of three million won, and two teams that came through it converted to a paid plan. At the next budget meeting Grace asked whether that meant one and a half million won per team, and the only answer Ravi had was that he'd try different copy next month. The money had already left, regardless of the outcome.

Affiliate marketing inverts that order. Nothing is paid up front to the person doing the recommending; payment happens when someone who arrives through their recommendation actually buys. The seller spends nothing when nothing sells, and the recommender earns more the better it sells. The risk hasn't disappeared, though — it has moved onto the recommender, so designing a commission worth that risk is half the work of running one of these programs.

Each person who recommends the product gets a link of their own, and when someone who follows it pays, a share of that payment is credited to them. In practice it runs in five steps.

1. Applications come in and get approved. Accepting anyone means the brand name ends up attached to strange writing, so most programs check the channel and the pitch first.
2. A unique link is issued. It looks like pagedeck.com/?ref=devnote, with the partner's identifier on the end of the address.
3. A visitor arriving through that link leaves a marker in their browser. Plenty of people look around today and pay five days later, so the sale counts for the partner as long as the marker is still valid — not only at the moment of the click.
4. The validity window gets set. The attribution window defines how many days after a touchpoint a result still counts, and affiliate programs usually run 30 or 60 days. A longer window favors the partner, a shorter one favors the seller.
5. Payouts get settled. Money rarely moves the moment a sale lands; the refund period has to pass first, and payouts usually go out once a month.

Commission rates come in two shapes: a percentage of the first payment only, or a percentage of every month the customer keeps paying. On a 30,000 won monthly plan, the first shape pays 20 percent once, or 6,000 won. The second pays 30 percent every month, or 9,000 won, which comes to 90,000 won over a ten-month customer. Subscription products lean toward the second shape because it pushes partners to bring in customers who stay. Under the first shape, a partner loses nothing when the signups they send disappear a month later.

Affiliate links work best where people are already reading to decide what to use. For someone comparing options, a recommendation link isn't an interruption — it's the next step.

Three places come up most. The first is comparison posts and review blogs. A post ranking for "collaborative document tools compared" through search engine optimization (SEO) reaches readers with a clear intent, and its conversion rate runs several times higher than an ad. The second is newsletters and courses, which have gathered an audience around one subject and carry enough standing with that audience for a recommendation to land. The third is agencies and consultants who already use the product with clients and take a commission for bringing it in.

The places it doesn't work are just as clear. When a contract requires a salesperson to meet the buyer several times, a single link almost never carries someone through to payment. Products like that handle introductions through partner agreements rather than per-sale commissions, because the contribution is months of work with a client rather than a click that can be traced to one address.

Pagedeck has opened its program to about forty blogs and newsletters that cover work tools. Three of them produce 80 percent of the revenue the program brings in. So Ravi's quarterly work isn't adding more partners — it's making better material for those three. The other thirty-seven took a link and never wrote anything, which is the usual shape of these programs: a long tail that costs nothing and produces nothing, and a short head worth treating like a real partnership.

"Thirty-two payments came through the program this month, and commissions came to 1.9 million won."

Cost only showed up in proportion to results.

"One of the partners is asking us to move the window from 30 days to 60."

A negotiation over how widely their sales get counted.

"This partner is bidding on our brand keyword. We need a clause about that in the policy."

They're intercepting customers who were coming anyway and collecting a commission for it.

Ravi is the one who actually runs this at Pagedeck. When an application arrives he opens the channel, approves it if it covers work tools and declines it otherwise. In the first week of each month he pulls results by partner and separates the ones producing payments from the ones producing only clicks. A channel with plenty of clicks and no payments is usually describing the product in a way that doesn't match it, so those partners get another copy of the material explaining which kinds of teams it suits and which it doesn't.

The most common mistake is thinking there is no risk because payment follows results. Nothing goes out up front, so it looks like failure costs nothing. But the person writing the pitch is the partner, not the company, and since commission attaches to payment, partners write toward payment. Features the product doesn't have get described, and words like "unlimited" appear.

Customers who arrive that way meet a different product than the one they were promised, ask for a refund, and land on the support team. Saving a 10,000 won commission ends up costing more than that to handle. Partners bidding on the brand name have the same root cause: commission attaches to customers who were already coming, so cost goes up while new customers stay flat. Working programs therefore write the restrictions into the policy in advance — claim limits, no bidding on brand keywords, and clawback of commission on refunds.