A nutra ad spy tool: how to find what is actually running on Facebook

Why the Ad Library is not enough

The Facebook Ad Library is official, public and free. That is where the good news ends. It answers the question "what is this page advertising". Working a vertical requires a different one: what is running in my niche, in my geo, and for how long.

• No sorting by how long an ad has been live — the single best signal, from the outside, that a campaign pays for itself.
• No search inside the creative. You can search advertisers and ad text, not what is shown in the image or said in the video.
• No grouping of variants. One creative in thirty versions reads as thirty ads and buries everything else.
• Media does not last. Video URLs expire, and a month later you are left with a card missing the only thing that mattered.
• Geography is where the ad was seen, not where it was targeted, and a single ad rarely tells you where the money is going.

A spy tool exists to solve those five things: it walks the library continuously, keeps the result in a database, stores the media itself, and lets you search what is inside a creative rather than only the name of the page that ran it.

What a campaign actually is

It is not a creative. It is four things at once: creative, offer, geo and landing page. Copying one of the four is copying none of them. Most people copy the creative, keep their own offer and their own prelander, see nothing, and conclude that spy tools do not work.

So all four have to be visible together, on one screen: the video itself, the ad copy, the country, and where the ad leads.

Signal one: how long it has been running

Longevity is the most honest signal available from outside. A nutra ad in its third week is almost certainly profitable — nobody in this vertical pays for impressions out of sentiment, and a losing ad is switched off within days.

Which gives a practical rule: filter by active status and start date first, and look at creatives second. Done the other way round you will pick whatever looks best, and in nutra "looks best" and "works" overlap less often than you would like.

Signal two: stacks

When one creative is running in twenty variants — frames reordered, a different opening shot, slightly different copy — that is not clutter. That is scaling. Somebody found an approach that works and is now hunting for its best version.

A single ad can be a hundred-dollar test. A stack of twenty variants cannot: behind it is a decision to keep spending.

Reading the creative itself

A nutra creative is almost always built from three layers, and it pays to take it apart along them.

• The pain. How the video opens — what is on screen in the first two seconds, and which condition is being described.
• The mechanism. How the result is explained: an ingredient, a device, a procedure, a doctor talking. This is the layer most often swapped during testing while the pain stays put.
• The offer. What is being sold and on what terms. It is the only layer visible solely on the landing page, which is why the landing page has to be kept whole.

Take ten live ads apart this way and you end up with something better than an idea for a creative: a map of which pains your geo is currently paying to solve, and which mechanisms are being used to close them.

Geo: why campaigns do not transfer

The same pain sells through different mechanisms in different countries, and that is not a translation problem. What changes is what people believe: a white coat works in one market, a neighbour of the same age in another, a reference to a television programme in a third. Move a campaign across geos by its pain layer, and pick the mechanism again from what is already running there.

In practice that means a database has to cover dozens of markets rather than two or three — otherwise there is nowhere to check the neighbouring geo at all.

Twenty minutes to a first map

• Pick one sub-niche and one geo. Not "nutra", but joints in Mexico.
• Filter to active ads launched more than two weeks ago.
• Sort by stack size: the creatives with the most variants first.
• Take the top ten and split each into pain, mechanism, offer.
• Open all ten landing pages. That is where the half you cannot see from the feed lives.

What comes out is a list of three or four recurring campaigns. That is what works in your geo now, as opposed to what worked when the last case study was written.

Common mistakes

• Judging a tool by the size of its database. Depth in your niche and your geo decides the outcome, not the total.
• Copying the creative together with its copy. Ad text burns out fastest and is the first thing moderation catches.
• Ignoring dates. A year-old ad describes a market that has moved on.
• Stopping at the card. Without the landing page you are looking at half a campaign and repeating it blind.