What to Recommend: Picking Products That Actually Sell

Most advice here is a version of "recommend what you actually use". True, and not enough: plenty of things you genuinely use will earn nothing, for reasons that have nothing to do with sincerity.

Here is what actually decides it.

#The rate is the least useful number

The equation is not the commission rate. It is the rate times the proportion of people who buy, times the proportion who keep it.

12% on a ₹6,000 item that two people buy and one returns pays less than 4% on a ₹700 item forty people buy and three return. The second also asks less of your audience's trust, which is the asset you are really spending.

So the first question is not "what pays most" but "what would these people have bought anyway".

#Decision time versus the tracking window

This is the criterion almost no guide mentions, and it disqualifies whole categories.

Swipe

Programs pay on a purchase made inside a tracking window that starts at the click, and some are 24 hours. An impulse buy - a lipstick, a phone case, a book - is decided in the same session, comfortably inside it.

An expensive, considered purchase is not. Someone tapping your headphone link reads three reviews, watches a comparison, waits for a sale, and buys eleven days later - possibly through someone else's link. You did the work that sold it and the window closed on day two.

That does not mean avoid expensive things. It means a short window and a long decision are a bad pairing, and the same recommendation performs completely differently depending on which you are in.

#Returns quietly take the rest

Almost every program reverses commission on a returned order, and returns are not marginal: the National Retail Federation put 19.3% of online sales as returned in 2025.

Apparel and footwear run well above that, because people deliberately order two sizes. If you recommend clothing, your effective rate sits well below your headline rate - and no dashboard shows that until the reversals land.

This is not an argument against fashion. It is an argument for knowing your real number rather than your advertised one - which you can work out from your own dashboard.

#Price band has to match the audience

An expensive pick to an audience that cannot afford it converts at roughly zero, and it also costs you something subtler: people stop reading your recommendations as being for them.

The useful check is not "is this too expensive" but "have these people bought at this price on my say-so before". If not, the first time is a test, not a strategy.

#The boring one that costs the most: dead links

A recommendation is worth nothing the moment it points at a product that is out of stock, discontinued, or region-locked.

This is where most affiliate income silently leaks, because nothing tells you it happened. The post still looks fine. The link still opens. It just lands on "currently unavailable", and the reader leaves.

Two habits fix most of it:

• Link to the product, not a category or a homepage. Every extra tap is a chance to leave.
• Check your back catalogue occasionally. The content you posted six months ago is still being found; if its links are dead, it is being found and wasted.

#A short filter, in order

1. Have people already asked you about it? If yes, most of the rest is a formality.
2. Would you recommend it at zero commission? If no, stop here.
3. Is it bought quickly, or researched for a fortnight? Match that against the program's window.
4. What is the return rate in this category? Adjust your expectations, not your honesty.
5. Is it in stock, in your audience's country, at a price they buy at?
6. Only now, look at the rate.

The thing you are actually optimising is the number of people who buy because you said so. Everything above is a way of not spending that.

---

Returns figures: the National Retail Federation and Happy Returns, 2025 Retail Returns Landscape, October 2025.

---