Which post to fund: the signal nobody looks at

The question asked before a campaign on this platform is almost always "what do we produce?". It is the wrong question, and it costs two weeks and a production budget before a single dinar of delivery has been spent.

The right question is "what has already worked?". A profile that has been posting for six months has, without knowing it, run a series of free tests on its own audience. The results are there, they are readable, and they generally identify one or two images nobody would have chosen.

This article is about that selection and what follows: checking the profile before any spend, the signal that actually predicts, what funding changes about a post, formats and their invisible zones, testing without producing, how fast a small audience wears out, and the rights over the person in the photograph.

The article accompanying this one deals with the platform itself — the size and composition of the audience, the grid as landing page, the bio, stories, partnerships. It is worth reading first: it says whether you should be here, this one says what to fund once you are.

The figure shown first under a post is the likes, and it is the least predictive of all. It measures an approval that costs one gesture and commits to nothing, and it is heavily influenced by posting time and by the habits of existing followers.

The useful signal is the save, and it is so for a mechanical reason: saving a post is an act directed at the future. The person is not saying "this is good", they are saying "I will need this". It is the closest thing to purchase intent this platform can measure, and it is available free.

The second useful signal is the share by private message, for the same reason inverted: somebody sent your post to one specific person. In a market where the purchase decision is often taken by two people, that is a direct indicator, and it is almost always ignored in favour of followers gained.

The method therefore comes down to one operation: sort your posts from the last six months by saves rather than likes, and look at the top five. They will not be the ones you would have named from memory, and the gap between the two lists is the value of the exercise.

A word on what not to conclude from it. A post saved a hundred times is not necessarily the best advertising image: it is often a piece of useful information — a tariff, a size chart, opening times — that people keep in order to find it again. Funding that works, but it sells the information rather than the product, and you need to know that before measuring the result.

The story format has a reading rule of its own that changes the design: the person watches with their thumb ready to move to the next one. There is no moment of discovery, and nothing holds the screen by default.

The consequence is that the first second carries the whole message. A logo appearing first, an opening animation, a sentence that builds: those are three ways of using the one second you have to say something other than your company’s name.

The text has to work without sound, because a large share of these views happen in silence, and it has to be large. The right word here is not "legible" but "legible while walking": that is how this format is actually consumed, and text at a normal size is text nobody reads.

The gesture asked for must be single and explicit. A story offering to visit the profile, reply to the message and swipe up asks for nothing at all; one that asks for a single thing sometimes gets it. And in this market the thing most worth asking for is a message, because that is the gesture people already make.

Not all posts are funded at the same moment, and the timing matters more than people say. A recent post is delivered with its comments still forming, which helps it; a six-month-old post is stable but its context may have changed.

The sequence that works has two stages. You post normally, leave a week for the signals to form and for the comments to receive their answers, then fund whatever performed best. That week is what separates funding as a decision from boosting as a reflex.

Some things are never funded and deserve listing: a post announcing something dated that will be wrong in ten days, a post answering one specific comment, a post that thanks people. Those are messages to your existing audience and they make no sense in front of a stranger.

Finally, do not fund more than one post at a time when the audience is small. Two simultaneous campaigns compete for the same people, the frequency rises twice as fast, and you pay for competing with yourself — an effect that barely exists on a large audience and is immediate on this one.

What we do is bounded. We read the profile’s history and draw the selection of fundable posts from it, we check the profile as a landing page before any budget, we prepare the rotation, and we set up the agreed word that lets a partnership be measured.

We refuse to fund a post before reading the profile it points at. It is a refusal people dislike because it costs a week, and it holds: paying to bring people to nine images that do not say what you sell produces exactly the same cost and no enquiries, and then reads as a failure of the advertising.

We also refuse to buy a mention from an account whose saves and previous sponsored posts we have not been shown. The follower count is the one metric on this platform that can be bought, and it is the one presented to us first.

And we do not produce a series of images before exhausting what the profile already contains. That position reduces what we invoice, and it is justified by the fact that what exists was tested on your real audience. There is also work only you can do that decides the rest: answering messages while the campaign runs.