By Louis Newman, founder of Magnetite (magnetite.ai)
TL;DR: Standard ad creative decays in two to three weeks and leaves nothing behind when the spend stops. A proof inventory (graded, rights-cleared praise from real users) compounds instead: every sweep adds units, every cleared advocate is reusable, and every ad result teaches you which proof converts. That asset is what the Proof Engine builds.
Every B2B ad budget answers one question, whether or not anyone asks it out loud: when the spending stops, what do you keep?
For most LinkedIn programs the honest answer is nothing. The impressions are gone, the creative is fatigued, the agency retainer bought you a folder of assets nobody will ever run again. Next quarter starts from zero, exactly like this one did.
I think that is a choice, not a law of physics. This post is the case for the alternative: treating advertising spend as the acquisition of a compounding asset rather than the rental of a decaying one. It is the argument underneath everything we build at Magnetite, so I want to lay it out properly, including the mechanics of how our ranking model encodes it and the places where the compounding story has real limits.
The treadmill everyone has normalized
The standard motion in B2B paid social looks like this. You brief a piece of creative. You produce it. You launch it. Performance peaks early, while the audience is still seeing it fresh, and then it sags as the same feed shows the same people the same asset again and again. The working assumption across performance teams is two to three weeks before fatigue sets in. Then the cycle restarts: new brief, new production, new launch.
Produce. Fatigue. Produce again.
Notice what the spend is doing in this model. Every dollar buys impressions, and impressions evaporate the moment they are served. The creative itself, the thing you paid an agency or an in-house team real money to make, has a useful life measured in weeks. The budget is rent. You are renting attention by the month, and the landlord never offers you equity.
The treadmill is also why ad accounts quietly stagnate. In our audit work we found a live ad that had been running unchanged for ten months. Not because the team was lazy, but because the refresh cycle is genuinely expensive: every new asset needs a concept, a production pass, an approval chain. When the machine that feeds the treadmill slows down, the treadmill does not stop politely. It just runs the stale asset into the ground.
The industry's standard prescriptions are all supply-side: hire more creators, iterate faster, generate variants with AI. Those help at the margin, but they accept the underlying physics. I have written separately about why creative fatigue is a supply problem, not a copywriting problem. The deeper move is to stop fighting decay with volume and start buying something that does not decay the same way.
The alternative: buy an asset, not a campaign
Here is what a Magnetite engagement actually produces: a permanent, searchable, rights-cleared, performance-tracked library of the best things real people have said about your product. We call the units of that library proof: practitioner workflow posts, expert recommendations, reviews, mentions, found across LinkedIn, X, Reddit, podcasts and review sites.
That library has three properties that no campaign has.
Inventory persists and grows. Every discovery sweep adds to the same library. Nothing expires. A recent client sweep surfaced 55 candidate units for five cents of API cost; 27 graded into the inventory and 28 were excluded as off-target or not actually about the client. The exclusions matter as much as the inclusions, but the point here is the economics: the marginal cost of adding to the asset is tiny, and the additions are permanent.
Cleared advocates are reusable. The expensive part of proof-led advertising is not finding praise, it is the human yes: the author granting permission for their post to run as a Thought Leader Ad. That yes is not a one-shot transaction. A cleared advocate is a relationship, and one yes can cover future posts too. The roster of people who have already said yes is itself a compounding asset, and almost nobody tracks it.
Performance feeds back into the library. Every ad result is recorded against the specific proof unit it ran from. The library does not just grow, it learns. Over time you know not just what praise exists, but which kinds of praise convert for your specific audience.
And the part I always lead with when someone asks what they are actually buying: cancel the ads any time, you keep the asset. The library, the grades, the cleared roster, the performance history. They are yours. That is the difference between rent and equity, and it is the practical expression of the thesis that your customers already wrote your best ads. The ads were always out there. The asset is the system that finds them, clears them, and remembers what they did.
The mechanics: how we rank proof, in plain language
Compounding is a nice word, and marketing is full of nice words. So let me show you the actual arithmetic, because our ranking model is where the thesis stops being rhetoric and becomes code.
Every graded proof unit in the library gets a compellingness score, recomputed at rank time. The score is three things multiplied together.
Tier weight. A detailed, unprompted workflow post from a real practitioner (we grade that gold) carries a base weight of 100. An expert recommendation carries 70. A review carries 45, and a passing mention carries 30. Credibility and specificity, not reach, set the base. A one-reaction gold post outweighs a high-engagement listicle mention by design.
Recency, with a floor. Newer proof should generally rank above older proof, so the score decays with a half-life of 60 days: a unit posted two months ago counts at half strength, four months ago at a quarter. But, and this is the line of code I care most about, the decay stops at 15%. Old proof never decays to zero.
That 15% floor is the entire compounding thesis expressed in arithmetic. A practitioner post from eight months ago should not outrank a fresh equivalent, and the model makes sure it does not. But the thing the old post documents, a real person getting real value from your product, did not stop being true when the post aged. It is still evidence. It is still searchable, still in inventory, still clearable, still usable. Compare that with ad creative, which decays to effectively zero in weeks, not because the asset changed but because the audience wore it out. Proof ages. Creative expires. The floor is how we encode the difference.
Evidence bonuses. On top of tier and recency, the score multiplies up for the signals that make proof land harder: specific numbers in the text (a 1.2x bonus), strong persona match with your buyers (1.15x), video (1.25x) or image (1.1x) over plain text, third-party verification (1.1x), and raw, unproduced delivery over polished content (1.1x).
Because the score is recomputed daily, the library continuously reorders itself: new finds surface to the top, older units settle toward their floor without ever leaving. Nothing is deleted. Every sweep makes the ranked inventory deeper, and every ad result sharpens what we know about which units to run next. That is what a compounding loop looks like at the implementation level.
What this means for budget allocation
Now run the budget math under each model.
On the treadmill, $20k a month at a $14 CPC buys roughly 1,400 clicks (illustrative at industry averages; an audit runs this on your own account's trailing baseline, not averages). Every one of those dollars is media rent, and the creative feeding the spend needs replacing every few weeks at additional cost.
The published Thought Leader Ads benchmarks, from the Fractional Demand and ZenABM reports (2026) set against LinkedIn Ads industry benchmarks (2026), put TLA performance at around $0.51 CPC against a $2.42 standard, and 4.65% CTR against 0.68%, which is where the headline claim of 4 to 6x the outcomes from the same budget comes from. Those numbers deserve scrutiny, including the vendor-bias caveats nobody quotes, and I go through them line by line in Thought Leader Ads: the 2026 economics. Treat them as direction, not gospel, and demand the math on your own baseline.
But even a skeptical read of the benchmarks misses the bigger reallocation. The benchmark delta is a snapshot; compounding is a trajectory. On the treadmill, efficiency resets to baseline every refresh cycle, so next quarter's budget buys roughly what this quarter's did. With a proof library, next quarter inherits everything: a deeper ranked inventory, a larger roster of cleared advocates who do not need re-clearing, and a performance history that tells you what to run first. The marginal cost of the next good ad falls over time, because most of the work that makes it good has already been done and banked.
Practically, that means part of your budget stops being pure media rent and becomes asset acquisition: discovery sweeps, grading, rights clearance. That part of the spend does not evaporate with the impressions. It accrues.
Where compounding stops being magic
I want to be precise about the limits, because the compounding case is strong enough that it does not need exaggerating.
Compounding needs raw material. The engine amplifies existing praise; it does not manufacture it, and we never pay for it. If your product does not yet have a real user base generating genuine praise, there is nothing to compound. Zero compounds to zero. This is exactly what the free audit establishes first: if your proof inventory comes back thin, we tell you, show you what we looked at, and you have lost nothing.
Week one is inventory building, not magic. The first sweeps populate the library. Grading takes a pass. Rights clearance involves real humans saying yes on their own schedule, usually days. The early weeks of an engagement are asset construction, and anyone who promises compounding returns in week one is selling you the treadmill with better vocabulary.
The asset does not spend itself. A rights-cleared library with no distribution is a well-organized filing cabinet. The compounding shows up when the library feeds a live ad program and the results feed back. Asset and distribution need each other.
The floor is a floor, not a fountain. Old proof stays usable; it does not stay equally potent. A library that stops getting fresh sweeps will still beat starting from zero, but the model deliberately favors fresh equivalents. Compounding here means the base never resets, not that effort becomes optional.
The question to ask your budget
So, back to the opening question, the one I think every B2B advertiser should put to their own ad account: if you stopped spending tomorrow, what would you keep?
If the answer is nothing, you are on the treadmill, and every quarter will start the way this one did. If the answer is a permanent, searchable, rights-cleared, performance-tracked library of the most credible voices in your market, plus a roster of advocates who have already said yes, you own something that gets more valuable with every sweep.
The first step costs you nothing: we run a free Proof Audit on your market, build the initial inventory, and show you exactly what praise is already sitting out there unpromoted. Request your free Proof Audit.