Every second, the global advertising industry places approximately $22,000 in digital ad spend. By any measure, this is the largest real-time allocation system in human commercial history. And the fundamental question underneath every dollar (”is this working?”) is answered with methodology borrowed from a world that no longer exists.

I built OpenLift to fix that.

This is the technical case for why multi-touch attribution is structurally broken, what replaces it, and why the replacement is now production-ready and open source.

The Wrong Question

The dominant measurement paradigm, multi-touch attribution (MTA), assigns fractional credit to touchpoints along a customer journey. Last-click, first-click, linear, time-decay, data-driven. These are all variants of the same question: given that a conversion happened, which ads deserve credit?

The question sounds reasonable until you realize it is the wrong question entirely.

The correct question is: would this conversion have happened without the ad?

That is a counterfactual question. Counterfactuals cannot be answered by observational data alone. They require causal inference.

Attribution tells you who was present at the crime scene. Incrementality tells you who committed the crime.

The consequences of getting this wrong are not abstract. Advertisers scaling campaigns on platform-reported ROAS of 4x may be generating true incremental ROAS of 0.8x. Destroying value while every dashboard signals success. Studies comparing platform MTA to randomized controlled holdout experiments consistently find the same result: platform attribution overstates incrementality by 2 to 5x.

That is not a rounding error. That is a structural lie baked into the measurement layer.

Why Attribution Fails: A Mathematical Autopsy

MTA has two failure modes that no amount of machine learning can fix, because they sit inside the problem formulation itself.

The Fundamental Problem of Causal Inference. For any individual, you observe either the outcome with the ad or the outcome without it. Never both. This was formalized by Holland in 1986 and it hasn’t changed. MTA computes a conditional expectation (what happened given someone saw an ad and converted) and calls it a causal effect. It is not. It conflates selection bias with lift.

The Retargeting Trap. Consider retargeting. Ads are shown to users who already visited your site. These users already have high purchase intent. When they convert, attribution credits the retargeting ad. But the counterfactual (would they have converted anyway?) is almost certainly yes for many of them. The “lift” from retargeting is largely a measurement artifact.

The data backs this up. Across industry holdout studies, the gap between platform-reported ROAS and true incremental ROAS looks roughly like this:

The pattern is consistent: the channels that look best in attribution are the ones with the largest inflation. Retargeting, where you are literally advertising to people who already want to buy, shows the biggest gap. Non-brand search, where the ad is actually causing discovery, shows the smallest.

And it gets worse. Safari’s ITP, Chrome’s CHIPS partitioning, iOS 14.5 ATT, and third-party cookie deprecation have destroyed the user journey graph that MTA relies on. By 2025, a typical e-commerce brand running deterministic MTA has observable journeys for roughly 20 to 35% of converters. The model is extrapolating from a systematically biased sample of a population it was already measuring wrong.

The Replacement: Bayesian Synthetic Control

OpenLift implements a Bayesian Synthetic Control. The core insight: find a weighted combination of control markets that closely tracks the test market before the campaign, then use that synthetic control as the post-campaign counterfactual.

In plain terms: if I run a campaign in Dubai, I find a blend of other markets (Abu Dhabi, Sharjah, Ajman) whose conversion patterns closely mirror Dubai’s behavior before the campaign launched. After the campaign, the gap between what Dubai actually did and what the synthetic Dubai would have done is the incremental lift.

The model uses a Student-t likelihood (heavier tails than Gaussian), which makes it robust to the promotional spikes, attribution storms, and data ingestion errors that would catastrophically bias a standard regression. Day-of-week effects are modeled explicitly. Priors are regularized and data-adaptive.

The output is not a point estimate. It is a full posterior distribution over lift. You get:

• Median lift (the central estimate)

• 90% Highest Density Interval (the shortest interval containing 90% of the posterior probability)

• P(lift > 0) (the probability the campaign had any positive effect at all)

Unlike a frequentist confidence interval, the HDI directly answers: “there is a 90% probability that the true lift lies in this range.” That is the statement marketers think confidence intervals make. It is not. Bayesian credible intervals actually make it.

Geo Selection: Why It Matters More Than the Model

The quality of the synthetic control depends entirely on how well the control geos track the test geo before the campaign. OpenLift uses Dynamic Time Warping (DTW) distance as the similarity metric.

Why DTW over simple correlation? Because correlation measures linear relationship at fixed lags. Two geos with identical conversion trajectories but one slightly phase-shifted (a market where weekends drive conversions a day later) will show low correlation but near-zero DTW distance. DTW finds the optimal alignment by warping the time axis within a constrained window.

Each candidate control gets a DTW score against the test market over a 90-day lookback. Geos above a threshold are excluded. This is non-negotiable. A well-calibrated posterior built on poorly matched control geos is worse than useless because it gives you false confidence.

Power Analysis: Design Before You Spend

The most common failure in marketing measurement is running an experiment that was never going to detect a real effect. Not because the effect was not there, but because the test was underpowered.

OpenLift’s power analysis module answers this before the campaign launches. Rather than assuming Gaussian errors and applying an analytic formula, it runs a Monte Carlo simulation: inject a synthetic effect into historical data, run the inference, check detection. Repeat. This captures the true correlation structure of the geo data.

The results are instructive:

• A 10% lift requires approximately 28 days to achieve 80% power with 4 well-matched controls

• A 5% lift requires 42+ days

• A 20% lift is detectable in under 21 days

If you are running a 14-day geo test with 2 control markets trying to detect a 5% effect, you are wasting everyone’s time. The math says so before you spend a dollar.

Multi-Cell Experiments and Channel Synergy

Real marketing systems run multiple channels simultaneously. OpenLift’s multi-cell framework tests whether Channel A alone, Channel B alone, or A+B together produce different incremental effects. And whether the combination is super-additive.

The synergy index is straightforward:

Synergy = Lift(A+B) / (Lift(A) + Lift(B))

Greater than 1.0 means the channels amplify each other. Equal to 1.0 means independent effects. Less than 1.0 means diminishing returns or cannibalization.

This is the question every CMO asks (”do my channels work better together?”) and no attribution model can answer, because attribution cannot decompose causation. Geo-lift can.

The 5-Tier Evidence Framework

Raw posterior statistics require statistical training to interpret. P(lift > 0) = 87% is precise but not immediately actionable for a VP of Marketing deciding whether to reallocate $2M in budget.

OpenLift maps the posterior to five decision tiers:

Critically, evidence strength gets downgraded if geo match quality is poor or if the credible interval is excessively wide relative to the mean. This prevents spurious “Strong” conclusions from well-calibrated posteriors built on poorly matched control geos.

The framework eliminates analysis paralysis. The posterior speaks directly to the business question.

Why Now

The statistical methods underneath OpenLift are not new. Synthetic Control was introduced by Abadie, Diamond, and Hainmueller in 2010. Bayesian computation has been theoretically tractable for decades. So why is 2025 to 2026 the inflection point?

Bayesian computation got fast. PyMC 5.x, built on PyTensor with JAX backend support, delivers MCMC performance that was unimaginable five years ago. A model that once took 45 minutes to fit on a mid-tier server now completes in 2 to 4 minutes on a MacBook Pro. The barrier dropped from “PhD in Bayesian statistics and a compute cluster” to “pip install pymc.”

Privacy killed the old way. iOS 14.5, GDPR enforcement, cookie deprecation, and ATT have not merely inconvenienced MTA. They have structurally broken it. Brands that relied on deterministic cross-device identity graphs have lost 60 to 80% of their observable user journeys. The industry is abandoning correlational attribution not by conviction but by data availability.

Emerging markets never had MTA in the first place. OpenLift’s geo-lift methodology works especially well in markets where user-level data was always sparse by design. In Sub-Saharan Africa, Southeast Asia, and Latin America, mobile-first consumers operate primarily through apps where deterministic attribution was always limited. Geo-level measurement provides the same quality signal in Lagos as in London.

The next trillion dollars of e-commerce growth will come from markets where MTA never worked in the first place.

LLMs remove the last adoption barrier. Bayesian credible intervals and posterior distributions are the right statistical objects for marketing measurement decisions. They have also historically required statistical training to interpret. OpenLift’s LLM integration converts posterior statistics into plain-language business recommendations. The statistics PhD is no longer a prerequisite for using the right methodology.

Empirical Validation

To validate calibration, OpenLift was tested against synthetic data where ground truth lift is known by construction. 200 experiments were simulated across varying effect sizes and geo counts. Key findings:

• The 90% HDI achieves approximately 90% empirical coverage, confirming calibration

• Estimation bias is consistently below 0.5 percentage points

• Power at 30 days with 4 well-matched controls exceeds 80% for effects of 10% or greater

• Dropping to 2 controls cuts power to 61%. Going to 8 controls pushes it to 94%

• Poorly matched controls (high DTW scores) degrade both coverage and power

The model works. The question is whether the industry will adopt it at the scale the measurement crisis demands.

The Bottom Line

Marketing measurement is in crisis, and the crisis is not temporary. The user-level tracking infrastructure that underpinned a decade of performance marketing is being systematically dismantled. The causal validity of what remains was always questionable.

OpenLift is:

• Statistically rigorous. Bayesian Synthetic Control with full posterior uncertainty.

• Principled in design. DTW geo matching and simulation-based power analysis.

• Decision-ready. 5-tier evidence framework mapping posteriors to business actions.

• Globally applicable. Geo-level methodology that works regardless of cookie availability.

• AI-augmented. LLM translation layer that removes the statistics expertise barrier.

• Open source. MIT License. Production-ready.

In a world where platform-reported ROAS routinely exceeds true incrementality by 3 to 5x, causal truth is not a luxury. It is the difference between brands that scale what works and brands that scale what appears to work until the budget runs out.

OpenLift v3.0.0 is open source under the MIT License. The methodology is based on Abadie et al. (2010), Hoffman & Gelman (2014), and Brodersen et al. (2015).

There is a line from Harvey, the old James Stewart film, that has stayed with me longer than I expected: “In this world, you must be oh so smart, or oh so pleasant.” Elwood P. Dowd says that for years he was smart, but he recommends pleasant.

I understand that line more now than I probably would have a few years ago, because there was a time when I thought the highest form of intelligence was being able to see through everything, cut through every argument, detect every weakness, and say the thing exactly as it was, whether or not the room had asked for that level of precision.

There is a particular pride that comes with being right, especially when you have spent years proving to yourself that you are not stupid, not naive, not easy to fool, and not the kind of person life can simply push around without consequence.

But after a while, you start to realise that intelligence without warmth can become a very expensive way to live, because you may win arguments and still lose rooms, you may see things clearly and still make people feel small, you may be technically correct and still become difficult to be around, and you may protect yourself so aggressively from being misunderstood that you slowly become someone people no longer want to understand.

I am not saying intelligence is useless. God forbid. The world is already full of enough loud foolishness pretending to be wisdom.

But I am starting to think pleasantness is deeply underrated, not the fake kind that smiles while resenting everyone, and not the weak kind that avoids truth because it is scared of tension, but the kind that has enough internal strength to not turn every interaction into a courtroom.

I had surgery a few days ago, and recovery has a way of making life feel smaller, slower, and more immediate, because you notice who makes things easier, who makes things heavier, who brings peace into a room, and who arrives with the emotional atmosphere of unpaid debt.

Maybe that is why the line feels more useful now.

For years, I wanted to be smart enough to never be caught off guard by life. Now I think I would rather be pleasant enough that life, even when it is hard, does not make me hard in return.

So yes, be smart, because stupidity is not a virtue and softness without sense is just poor decision-making with good manners.

But where possible, I recommend pleasant.

🤍 db

There is something unsettling about Easter if you actually take it seriously. Not the sanitized version, colored eggs, soft music, polite sermons, but the raw claim at its center: that someone saw humanity clearly, in full resolution, without illusion, and chose sacrifice anyway.

That’s the part most people skip.

Because if you look at human beings honestly, without sentimentality, the conclusion isn’t flattering. We are inconsistent, self-serving, fragile under pressure, and often blind to our own contradictions. We betray what we claim to value. We avoid responsibility until it becomes unavoidable. Even at our best, we are unreliable.

And yet, the claim of Easter is not that humanity improved first. It’s not that people became worthy and then were saved. It is the opposite. The sacrifice precedes the redemption. The decision is made in full awareness of human failure, not ignorance of it.

That changes everything.

Because then sacrifice is no longer transactional. It is not “I give because you deserve.” It becomes something far more dangerous: “I give because I choose to bear what you cannot.”

That is not soft. That is not sentimental. That is terrifyingly strong.

Easter is not just about suffering, it is about voluntary suffering. And that distinction matters. Anyone can endure pain when they have no choice. But to step into it willingly, knowing the cost, knowing the people you are doing it for will misunderstand you, abandon you, even deny you, that is a different category entirely.

That is not weakness. That is control.

It also destroys the excuse most people hide behind: “I would do the right thing if it were easier.” No, you would not. The entire point is that it is not easy. It is not convenient. It is not fair. And it certainly is not optimized for comfort.

Easter confronts you with a question you cannot dodge if you are honest:

What are you willing to carry, even when you do not have to?

Because that is where meaning lives, not in comfort, not in avoidance, but in the voluntary acceptance of burden. Not because it benefits you immediately, but because it is right.

Most people want the reward without the weight. Easter rejects that completely.

It says the weight is the point.

And maybe that is why it hits deeper than any other holiday. It does not celebrate what is easy or natural. It celebrates what is almost impossible, choosing sacrifice in a world that gives you every reason not to.

That is not a story about the past.

There is a quiet shift happening that few are willing to name plainly: men are becoming increasingly tolerant of what should never be tolerated. Not out of strength, but out of confusion. Not because they have mastered themselves, but because they no longer know where their boundaries are.

A man without boundaries does not become peaceful; he becomes negotiable. And a negotiable man is eventually shaped by whatever pressures surround him, rather than by any internal standard he has chosen and committed to uphold.

This is not framed as a problem. In fact, it is often presented as progress. Men are encouraged to be more understanding, more accommodating, more emotionally aware. These are not inherently flawed ideals, but when they are not paired with firmness, they produce something incomplete—a man who can explain everything, justify anything, and ultimately stand for nothing.

The ability to empathize has quietly replaced the ability to draw a line.

And without a line, there is no structure. Without structure, there is no identity. A man who cannot say “no” with conviction does not have principles he has preferences. And preferences collapse under pressure.

What is particularly dangerous about this shift is that it does not feel like decline. It feels like maturity. It feels like growth. Conflict is avoided, tension is minimized, and everything becomes easier to navigate in the short term. But over time, something essential erodes. Self-respect is not lost in a single moment; it is traded away gradually, in small decisions where discomfort is avoided and standards are quietly lowered.

Men are not being forced into this. They are adapting to it. It is easier to tolerate than to confront, easier to rationalize than to reject, easier to adjust oneself than to risk losing something by standing firm. And so the line keeps moving. What was once unacceptable becomes understandable. What was once unthinkable becomes negotiable.

This is not strength. Strength requires the capacity to endure discomfort in service of something higher than immediate peace. It requires the willingness to lose, if what is being preserved is not worth keeping.

A man who avoids conflict at all costs does not eliminate conflict—he internalizes it. And over time, that internal compromise manifests as resentment, confusion, and a quiet sense that something is wrong, even if it cannot be clearly articulated.

Respect, both from others and from oneself, is built on the enforcement of standards. Not arbitrary standards, not rigid or performative ones, but real lines that define what is acceptable and what is not. Without them, a man becomes fluid in the worst sense—shaped by circumstance, rather than shaping it.

The tragedy is not that men are becoming more understanding. It is that they are becoming less decisive. Less willing to act. Less certain of what they stand for. And in that uncertainty, they begin to accept situations, dynamics, and outcomes that slowly diminish them.

There are things a man should walk away from without hesitation. There are lines that, once crossed, should not be renegotiated. Not out of ego, but out of clarity.

Because a man who cannot protect his own standards cannot protect anything else.

And if he loses that, he does not collapse dramatically. He continues. He functions. But he does so without alignment, without conviction, and without the quiet certainty that his life is truly his.

That is what should concern you. Not visible failure, but invisible compromise, the slow erosion of standards that once defined what it meant to stand as a man in the world.

For most of the last decade, marketing measurement has been built on a convenient illusion.

If a user clicked an ad and later converted, the ad “worked”.
If spend went up and revenue followed, the channel “scaled”.
If the dashboard looked green, decisions felt justified.

That illusion is collapsing.

Privacy changes broke deterministic attribution. Platforms increasingly grade their own homework. GA4 drops data and fills the gaps with modeling few teams fully understand. Meanwhile, the questions executives actually care about have not changed:

Did this spend cause growth?
Or did it simply coincide with it?

Most marketing tools still can’t answer that honestly.

The question attribution avoids

Attribution systems try to answer:
“Who gets credit?”

But the harder and more important question is different:

“What would have happened if we did nothing?”

That is a counterfactual question. And attribution, by construction, cannot answer counterfactuals. It assigns credit within observed activity. It never asks whether the activity mattered at all.

This is why teams argue endlessly about last-click vs data-driven models while budgets quietly drift away from reality.

Incrementality is not new. It’s just inconvenient.

Economists have been asking counterfactual questions for decades. Policy evaluation, clinical trials, and causal inference all start from the same principle:

You don’t measure impact by observing outcomes.
You measure impact by comparing outcomes to a credible alternative world.

In marketing, that alternative world is usually created through experimentation. Geo tests. Holdouts. Pauses. Controlled rollouts.

Enterprise tools like Haus or Recast do this well. They are powerful, opinionated, and expensive. They also abstract away most of the underlying assumptions and tradeoffs.

That abstraction is useful for delegation.
It is less useful for understanding.

The gap we kept running into

In practice, many teams fall into an awkward middle ground.

They are not large enough to justify a full-service incrementality vendor.They are too sophisticated to trust platform-reported ROAS.
They have data teams, but no shared, inspectable measurement primitive.

So they either:

• avoid incrementality entirely, or

• stitch together fragile notebooks and one-off analyses that don’t survive scrutiny.

What was missing was not another dashboard. It was measurement infrastructure.

Something local, transparent, and rigorous enough to be trusted, but simple enough to run quickly.

What OpenLift is

OpenLift is an open-source system for measuring incremental impact using Bayesian synthetic control methods.

At a high level, it does something simple:

1. You change something in one market.

2. You compare it to similar markets where nothing changed.

3. You estimate what would have happened without the change.

4. You quantify the difference, with uncertainty.

Under the hood, OpenLift builds a weighted synthetic baseline. Instead of saying “Saudi went up”, it learns a relationship like:

Saudi ≈ 0.6 × UAE + 0.4 × Bahrain

That relationship is learned from historical data, not assumed. It is probabilistic, not deterministic. And it produces a full distribution of possible outcomes, not a single point estimate.

The result is not a number you are told to trust. It is a range you can reason about.

Why Bayesian methods matter here

Real-world data is noisy. Seasonality is strong. External shocks are constant. Pretending otherwise leads to false certainty.

Bayesian modeling forces uncertainty into the open. Instead of asking “what is the lift?”, OpenLift asks:

• What is the most likely lift?

• How wide is the plausible range?

• How confident are we that the effect is even positive?

This matters because decisions are made under risk, not certainty. A channel with a +5% mean lift and a wide confidence interval is a very different decision from one with the same mean and tight bounds.

Most marketing tools hide this distinction. OpenLift makes it unavoidable.

What OpenLift is not

OpenLift is not an attribution model.
It does not optimize creatives.
It does not replace experimentation discipline.
It does not magically fix poor data.

It answers one question, well:

Did this intervention cause impact, and how sure are we?

That constraint is intentional.

Who this is for

OpenLift is not designed for CMOs to download and run experiments themselves.

It is designed for:

• data and analytics teams who want inspectable methods

• growth teams running real experiments

• agencies and specialists who need a reliable core

• founders who want defensible answers, not comforting ones

Some teams will still choose full-service vendors. That makes sense at scale. But delegation and understanding are not the same thing. Infrastructure exists to raise the floor, not replace services.

Why open source

Measurement only works if it is trusted. Trust requires transparency.

OpenLift is open source because:

• assumptions should be visible

• weights should be inspectable

• uncertainty should be explicit

• results should be reproducible

You should be able to defend these numbers to finance without appealing to vendor authority.

The direction this points to

As privacy constraints tighten and platforms become more opaque, marketing will look less like optimization and more like applied economics.

Teams that can run clean experiments, reason about uncertainty, and distinguish signal from noise will allocate capital better. Teams that cannot will continue to argue over dashboards.

OpenLift is a small piece of that future. Not a product pitch. Not a silver bullet. Just a primitive for telling the truth more often.

If you want to explore it, the repository is here:
👉 https://github.com/daramolaworks-create/openlift

Feedback from people who care about causal inference, measurement rigor, and decision-making under uncertainty is welcome.

Performance marketing has reached a point where additional tooling no longer produces better decisions. The limiting factor is no longer access to media, bidding sophistication, or analytics dashboards. The limiting factor is where intelligence resides and who controls it. The prevailing operating model places advertisers inside platform-defined abstractions, where optimisation logic, data visibility, and even the interpretation of success are mediated by systems whose incentives are not aligned with the advertiser’s objectives. At scale, this is not merely inefficient. It is structurally fragile.

Advertising platforms such as Meta, Google, and TikTok optimise through probabilistic auction mechanisms that estimate expected value under uncertainty. Bids are effectively derived from predicted conversion probabilities conditioned on user context, creative features, and historical priors, multiplied by advertiser-supplied value signals. These models are trained on vast, platform-internal datasets, incorporate delayed attribution, and adapt continuously across advertisers and campaigns. The optimisation is coherent from the platform’s perspective, but it is not designed to expose its internal reasoning to any single advertiser.

What advertisers receive in return are aggregated reports. Conversion counts, blended CPA, and directional ROAS are surfaced without posterior probabilities, uncertainty bounds, or saturation curves. The system reports outcomes, not mechanisms. As spend increases and markets become non-stationary, this lack of explanatory access becomes a binding constraint. Decision-making shifts from reasoning to reaction, and optimisation becomes a matter of trusting interfaces rather than understanding systems.

Most performance tooling attempts to resolve this gap by reprocessing platform outputs. Data is ingested through APIs, visualised in dashboards, and augmented with recommendation layers that operate on the same abstractions they consume. This does not restore agency. It reinforces dependence. The logic remains downstream of the platforms, and proprietary performance data is routinely externalised into shared infrastructure that aggregates learnings across many advertisers. Over time, strategic behaviour converges, and differentiation erodes.

An Independent Performance Intelligence Layer (IPIL) reverses this dynamic. Instead of treating platform reporting as ground truth, IPIL treats it as an input signal. Performance data is consolidated into an owned analytical environment and modelled explicitly around spend, exposure, and outcomes across channels, geographies, and creatives. Core performance metrics are computed deterministically as part of the data model rather than inferred dynamically by dashboards. This creates a stable analytical surface on which system-level reasoning can occur.

Within IPIL, machine learning models are not positioned as autonomous optimisers. They function as analytical interfaces that translate intent into interrogation. Natural-language questions are compiled into explicit queries against a known schema. Patterns such as pacing instability, creative fatigue, or geographic decay are identified through direct inspection of the data rather than inferred from platform narratives. Recommendations emerge as structured hypotheses with stated assumptions and evidentiary support, not as black-box prescriptions.

The objective of IPIL is not to compete with platform bidding systems at the auction level. Those systems are already near-optimal within their informational constraints. The objective is to operate above them by identifying misallocation regimes, cross-channel opportunity cost, and shifts in performance that platform-level models are not incentivised to surface. This is where strategic leverage exists. IPIL does not replace platform optimisation; it contextualises it.

From an engineering perspective, IPIL requires abandoning marketing-specific tooling in favour of general systems principles. Columnar data stores are selected for their ability to handle high-cardinality facts efficiently. Ingestion is immutable, with transformation and type enforcement occurring explicitly rather than implicitly. Analytical views are treated as contracts. Reasoning models are stateless, auditable, and decoupled from execution. Every recommendation can be traced back to a query, and every query back to a defined data structure.

This shift is becoming necessary. As advertising platforms reduce signal granularity, extend attribution windows, and push broader optimisation defaults, the gap between reported performance and underlying system behaviour continues to widen. Maintaining control now requires an intelligence layer that is independent of platform incentives and external tooling ecosystems.

Performance marketing is transitioning from dashboard-driven optimisation to system-level inference. The Independent Performance Intelligence Layer is not a new category of software. It is the architectural correction required to keep reasoning inside the organisation.

Every system — human or machine — eventually drowns in what it knows.

We are living in the age of infinite input. Every scroll, click, ping, and post adds another layer to a growing ocean of unfiltered data. Our dashboards have never been fuller, our feeds never louder — yet clarity has never felt more scarce.

The problem isn’t information. It’s weightlessness.

When everything carries equal importance, intelligence breaks down. The mind freezes. The model overfits. The algorithm loops endlessly, predicting the predictable. In that moment, both human and machine lose the one thing that makes them useful — the ability to decide what truly matters.

The Physics of Relevance

Intelligence, at its core, is an act of weighting.

The human brain does it constantly — prioritizing signals, ignoring noise, assigning relevance to a few bits out of billions. The same happens inside a neural network: every “weight” determines how much importance a data point holds. Learning, for both, is simply the process of adjusting those weights over time.

Attention, then, isn’t just focus. It’s architecture.

Modern AI systems, for instance, rely on something called an attention mechanism — a way for the model to decide which tokens to amplify and which to mute. That simple concept is what gave rise to GPTs, translation engines, recommendation systems, and the illusion of understanding.

But this isn’t just about machines. Every organization, dashboard, and marketing framework you’ve ever seen — from attribution models to MMMs — is really a giant act of weighting: deciding which signals drive meaning, which outcomes define success, and which truths get ignored because they’re inconvenient to measure.

Every intelligent system is a weighting system disguised as something else.

The Bias in Every Weight

Weighting, however, is never neutral.

The moment we decide what matters, we encode bias.
Humans overweight short-term feedback. Algorithms overweight what scales.
Marketers overweight metrics that look good in a deck.
Investors overweight stories that confirm their worldview.

Machines mirror us — not because they are conscious, but because they inherit our distributions of belief. The weights we set, explicitly or implicitly, determine what the model sees as signal and what it dismisses as noise.

In practice, this means we are constantly building systems that learn our blind spots faster than our strengths.

Every model is a mirror. What we choose to weight is what we choose to believe.

The Age of Abundance, The Poverty of Meaning

We often mistake more data for more intelligence.
More dashboards. More metrics. More pipelines.
But the more we collect, the less we discriminate — and discrimination is the soul of intelligence.

The real scarcity today isn’t information. It’s discernment.

I’ve seen this everywhere: in AI models chasing loss functions, in growth systems optimizing for vanity KPIs, in teams buried under metrics they don’t understand. The abundance of information creates the illusion of progress — while clarity silently erodes.

At some point, the signal-to-noise ratio becomes a moral issue.

Towards Weighted Thinking

What we need isn’t more data. We need better hierarchies of meaning.

Weighted thinking begins with restraint — the willingness to assign importance deliberately, to treat relevance as a scarce resource.

Weight what endures. The things that stay true under pressure — trust, quality, craft — deserve disproportionate attention.
Weight the feedback loop. Systems become intelligent only when consequence feeds back into decision.
Weight the invisible. Infrastructure, culture, data hygiene — the quiet things that make the loud things possible.

This isn’t theory. It’s how living systems survive. From the brain’s synaptic pruning to a company’s focus discipline, progress depends on forgetting the irrelevant as much as learning the relevant.

What Matters, Matters Because We Decide It Does

In the end, intelligence is less about memory and more about judgment.
The machine learns by adjusting its weights.
The mind grows by doing the same.

When we stop weighting, we stop thinking.

Information doesn’t want to be free — it wants to be weighted.
Because only through weight does it become meaning.
And only through meaning does it become wisdom.

We have misunderstood miracles. We imagine them as interruptions in the order of the world, sudden flashes of the impossible. But the deeper truth is quieter, more demanding: the miracle is not in what arrives uninvited, but in what we permit ourselves to create.

A miracle is the moment a person decides they are allowed. Allowed to step out of the script handed to them. Allowed to imagine a life different from what was expected. Allowed to break the weight of fear with one act of courage. Miracles do not descend; they emerge from the hidden permission we give ourselves.

And this permission is costly. To walk toward what matters means entering the fog of uncertainty, facing rejection, failure, exile from the familiar. Yet to choose otherwise — to wait passively for something outside to rescue us — is to shrink from the dignity of existence. The miracle belongs to those who act, even trembling, even unsure.

But pursuit is dangerous. It corrodes many who take it up. There is always the risk of becoming hard, bitter, or ruthless in the chase for a goal. The world rewards speed, not grace. Power, not tenderness. The true pathway to nobility is not in victory alone, but in the manner of the striving.

Nobility is the refusal to betray yourself, even when ambition whispers otherwise. It is the discipline to want fiercely without letting that want consume your integrity. It is walking forward while holding still the capacity for wonder, kindness, and restraint. The noble person achieves without disfiguring themselves in the process.

This is the miracle that matters most: not simply reaching a destination, but remaining intact along the way. To fight and not grow small. To dream and not become cruel. To hunger and still carry gentleness.

We speak of miracles as lightning, but perhaps they are more like seeds. They begin invisibly, as convictions in the soil of one’s inner life, watered by persistence, cracked open by struggle, and finally breaking into the light. By the time others notice them, they no longer look like miracles at all. They look like inevitabilities.

But you will know. You will remember the silence before it began, the countless moments when you could have surrendered, the days when no one believed. And in that memory, you will see clearly: the miracle was never given. It was made.

— 🤍 db

This week had its own rhythm. Long hours, unexpected meetings, conversations that felt heavier than they should, and small wins that carried more weight than they looked like at first glance.

I noticed how much of the outcome depended on framing. The same event could look like friction or learning. A mistake could feel like failure, or it could prove the value of taking responsibility quickly. A conversation could spiral into wasted energy, or it could be shut down before it eroded peace.

None of these were extraordinary in themselves, but together they built a pattern: progress through reframing. Choosing how to interpret what’s in front of you. Extracting signal from noise.

That’s what this week was for me, not about closure or completion, but about seeing the same things differently, and finding clarity in the shift.

— 🤍 db

The CMO’s job has always been simple to state and hard to do: convert attention into durable, predictable cash flows. Everything else—brand, creative, channels—is a means to that end. In the broadcast era, when distribution was scarce and measurement was crude, we tolerated very wide error bars. The CMO was a storyteller with budgetary discretion because the market rewarded reach and consistency more than precision. That world is over. The constraint today isn’t distribution; it’s proof. Capital is more expensive, competition is faster, and executives want a line of sight from dollars out to dollars back. The role hasn’t changed in purpose, only in the standard of evidence. And that single shift explains why performance marketers will increasingly run marketing.

Think about what you actually hire a CMO to do in a modern company. First, you want someone who treats marketing as capital allocation, not content production. Marketing is a portfolio of bets that either accelerate payback or build a compounding advantage—ideally both. That requires a leader who can decide: what to fund now, what to sunset, what to test, and what to scale, all under a clear hurdle rate. Performance leaders are the ones trained to make those calls weekly, not annually. Their instinct is to treat cash like oxygen: budget moves to where returns clear the bar. That is how boards expect a CMO to behave when money is not free.

Second, you want a CMO who turns uncertainty into knowledge. Most marketing debates are really epistemology: what’s true, how do we know, and how quickly can we update our beliefs? In a high-noise environment, the winning operating model is test → learn → redeploy. Performance leaders don’t worship channels; they run them. They collapse opinions into experiments and route budget toward whatever the results justify. That habit matters because the premium today is on speed of learning. The faster you cycle evidence into allocation, the more efficient your growth becomes. A modern CMO must be the architect of those feedback loops. Performance marketers already live inside them.

Third, you want the CMO to be the company’s demand truth-teller. Sales feels pipeline. Product hears feature requests. Finance sees lagging revenue. Only marketing has the cross-section of top-of-funnel signal and cost to acquire it. The CMO should be the executive who steps into the room and says, with credibility, “Here is where demand truly is, what it costs, what it takes to move, and what we should stop doing.” That statement is valuable only if it’s grounded in reality. Performance leaders earn the right to make it because their craft forces contact with the market every day. They see elasticity in real time. They know where the acquisition engine stalls and why. That proximity to demand is not a “tactic”; it’s the vantage point a modern CMO needs to steer the company.

Fourth, the CMO has to reconcile a false dichotomy: brand versus performance. The job has never been either/or. “Brand” is simply the compounding effect of consistent promises kept; “performance” is the discipline that ensures we’re not paying tomorrow’s profits for today’s vanity. In healthy companies, brand is defended not with poetry but with evidence that long-term investment shifts the slope of the curve—lowering future acquisition costs, raising conversion, increasing retention. That is a quantifiable claim over time. The kind of leader who can make it convincingly is the one who knows how short-term signals ladder into long-term economics and is disciplined enough to fund both with intent. Performance marketers who grow up—who move from button-pushing to portfolio thinking—are the ones who can finally integrate this. They don’t kill brand; they underwrite it.

Fifth, the CMO must connect marketing to the rest of the operating system. Growth fails when functions pursue local maxima: sales optimizes for near-term quota, product for velocity, finance for cost control, creative for awards. Someone has to force a global optimum. The modern CMO is that integrator. They translate market signal into product bets, into sales enablement, into financial planning. Performance leaders are credible integrators because their work already sits at the intersection of those trade-offs. They know what kinds of prospects convert and stay. They know which messages create lift and which features close the gap. And they can state the cash impact in a way finance respects. That is how you align a company around growth as a system, not a slogan.

If that’s the job—a capital allocator who manufactures learning, tells the truth about demand, funds both the compounding and the immediate, and integrates the firm—then ask who is closest to doing it today. It’s not the pure brand marketer. They are essential for voice and taste, but the bar has moved from taste to proof. It’s not the channel specialist who knows one platform deeply but cannot price risk across a portfolio. The person closest is the performance leader who has outgrown the console, who can step back from daily spend and answer the only questions that matter: Where is our next efficient dollar? What stops us from scaling it? What is the cash consequence of choosing A over B?

This is not an argument for narrowness. The performance leaders who will become CMOs are the ones who developed range: they can steward narrative because they understand compounding; they can partner with product because they understand why customers say yes; they can earn the CFO’s trust because they defend budgets with outcomes, not aspirations. They will hire taste around them and give it the oxygen to create advantage, but they will keep the company honest about what the market is actually telling them. In other words, they will look less like a campaign executive and more like a general manager of demand.

You could call this a pendulum swing—from brand to performance—but that misreads the moment. It’s not a swing; it’s a consolidation. The CMO role is collapsing into its essence: the stewardship of growth as an asset. In times of cheap capital, we tolerated soft claims about that asset. In times of harder capital and faster competition, we insist on verification. The leaders trained to verify, to learn faster than they spend, and to allocate as if the money were their own are performance marketers. When they mature into portfolio owners instead of channel operators, they match the job to be done better than anyone else.

That is why the outcome is predictable. The board will increasingly pick CMOs who behave like investors in the company’s demand curve—who can compound it, price it, and defend it. The supply of that mindset comes from performance. And the companies that understand this early will scale more efficiently, waste less budget, and build brands that earn their way onto the balance sheet the hard way: by being too effective to ignore.

This week carried a different weight. Not the heaviness of chaos, but the gravity of clarity.

There are moments when everything feels scattered—effort without direction, motion without meaning. And then, almost suddenly, the fog lifts. The same work, the same hours, but this time they begin to cohere into something larger. That’s what this week felt like: alignment.

It wasn’t just about results, though there were results. It was about the quiet satisfaction of knowing the compass is pointing true. That what once felt like strain is now flow. That what seemed uncertain is now anchored.

There is a profound relief in that shift. A sense of being not just busy, but purposeful. Not just moving, but moving in the right direction. The kind of progress that fuels belief—not just in the outcome, but in yourself.

This week reminded me that achievement isn’t loud. It’s not always visible. Sometimes it’s the subtle but unmistakable feeling that the pieces are finally in their rightful place.

And that feeling is everything

🤍 db

Lately, the days have felt compressed, like more is being asked of each hour than the hour can hold.

There’s been a lot of context switching. Multiple arenas. Multiple conversations. Some urgent, some just loud. It’s easy to confuse activity for progress, but I’m learning to separate the two.

Not every task needs to be closed immediately. Not every thread deserves to be pulled. Sometimes the smartest move is to let certain pieces breathe while you reallocate energy to where it will compound.

This week wasn’t about wins or losses. It was about load management. About protecting the capacity to play the long game.

And that, for now, is enough.

🤍 db

Lately, life has been less about scaling and more about pruning.

I’ve been deliberately quiet. Some of it by design. Some of it because when things start shifting internally or externally it’s better to listen than to speak. I’m learning that noise isn’t momentum. And movement without clarity is just distraction dressed up as productivity.

In the background, I’ve been doing the less glamorous work: simplifying. Reducing surface area. Saying no more often. Watching closely where I was previously quick to act. These are the things that compound over time, even if they don’t generate dopamine in the short term.

There are still wins, still forward motion but they’re less performative now. More private. More principled.

Sometimes the best strategy is to step back, recalibrate, and wait for the next inevitable asymmetry.

And that’s where I am.

🤍 db

There’s a moment in every high-performing organization when the cracks begin to show. What worked yesterday — scrappy experimentation, isolated wins, and brute force scaling — suddenly becomes inadequate. Growth stalls. Metrics plateau. The playbook feels empty.

It’s not for lack of effort or talent. Often, the problem lies in the system or rather, the lack of one.

Growth isn’t magic. It’s math. It’s engineering. And most importantly, it’s process-driven. To truly scale, businesses need to shift their mindset from chasing tactics to designing systems that optimize the entire growth stack, from acquisition to retention.

This is the first principle of Levered thinking.

The Misalignment of Effort and Outcomes

Too often, growth efforts fail because they’re disconnected from the realities of a business’s lifecycle. Startups operate like speedboats, pivoting quickly but often without direction. Scaleups are like cargo ships, powerful but slow, burdened by the complexity of their own success.

What both miss is a framework that aligns effort with outcomes:

Startups need repeatability. Early traction is often fueled by founder hustle or viral luck. But to reach Product-Market Fit (PMF), you need scalable, data-driven systems that remove the guesswork.

Scaleups need efficiency. Growth at scale isn’t about doing more. It’s about doing what works, better. CAC (Customer Acquisition Cost) must be optimized, and the growth engine tuned for cost-efficiency without compromising velocity.

The solution isn’t more campaigns or bigger budgets. It’s a system, a growth framework as disciplined as it is ambitious.

The Case for High-Fidelity Growth

In a world flooded with low-signal marketing noise, the winners are those who operate with high fidelity, clear signals, precise execution, and ruthless prioritization. Chamath Palihapitiya often talks about the “predictable, repeatable, scalable” nature of growth. These aren’t buzzwords; they’re survival principles.

What does high-fidelity growth look like?

1. Inbound Precision: A strategy where every touchpoint works cohesively to attract high-intent users. No scattergun approaches, just deep alignment between audience, messaging, and channels.

2. Pipeline Predictability: Building predictable acquisition funnels that rely on robust experimentation frameworks, backed by strong attribution models to understand what truly moves the needle.

3. Infrastructure Overhaul: Growth requires tech sophistication. Marketing automation, analytics pipelines, and scalable ops are non-negotiable. Without these, your efforts are little more than a house of cards.

The Missing Growth Equation

At the core of every stalled growth story is a missing piece of the equation. Perhaps it’s a lack of data granularity, an unclear ICP (Ideal Customer Profile), or weak experimentation loops. Growth isn’t a one-size-fits-all playbook; it’s a custom equation that evolves with your business.

What’s missing in many organizations today is a levered approach, a systematic method that identifies the right levers to pull at the right time. Consider these questions:

• Are your acquisition channels optimized for long-term ROI, or are they feeding short-term wins at unsustainable costs?

• Is your retention strategy reactive or proactive, with clear pathways to maximize LTV (Lifetime Value)?

• Are your growth metrics vanity-driven or rooted in actionable insights?

The answers lie in rethinking growth from first principles. What does your business fundamentally need to move from stagnation to scale?

Growth as a Discipline, Not a Department

One of the biggest misconceptions in growth is that it’s a function siloed within marketing. Real growth transcends departments. It’s a company-wide discipline that touches everything from product development to sales alignment.

To scale effectively, growth must operate at the intersection of creativity, data, and technology. It’s about building feedback loops that connect acquisition, activation, and retention into a seamless, iterative process.

Feedback loops Growth isn’t linear; it’s cyclical. A well-designed loop amplifies what works while quickly discarding what doesn’t.

Network effects Scaling requires more than incremental gains. True scale happens when growth becomes self-sustaining through referrals, community, and virality.

Compounding growth The best growth systems don’t just add value. They compound it, leveraging past learnings to drive exponential impact.

Why Levered Thinking is the Future

In today’s hyper-competitive landscape, businesses can’t afford to rely on intuition alone. Growth demands precision, technical rigor, and a willingness to experiment fearlessly.

A levered approach combines the creativity of startups with the operational efficiency of scaleups. It’s a mindset that prioritizes systems over tactics, impact over activity, and long-term growth over short-term wins.

Whether you’re a startup seeking PMF or a scaleup optimizing for efficiency, the principles of levered growth apply universally.

Because growth isn’t just about moving fast. It’s about moving with purpose.

https://levered.us/

WTF is the Attribution Reporting API in Google’s Privacy Sandbox?

Google has introduced the Attribution Reporting API within its Privacy Sandbox initiative, a landmark in the evolution of digital advertising. This tool enables advertisers and ad techs to track ad-driven conversions while prioritizing user privacy. It signifies a departure from traditional tracking, phasing out the use of cross-party user identifiers. This move is in response to the growing demand for data protection and the necessity to preserve the efficacy of digital advertising. The Attribution Reporting API is pivotal in marrying privacy-conscious practices with the need for precise advertising metrics.

The trends defining the $1.8 trillion global wellness market in 2024

McKinsey reports on the trends defining the $1.8 trillion global wellness market in 2024 highlights several key developments. McKinsey's insights into the 2024 global wellness market, valued at $1.8 trillion, spotlight the shift in consumer focus towards health and wellness. Emphasizing home-based health solutions and sophisticated biomonitoring, the wellness industry is poised for continued expansion. Consumers are increasingly willing to invest in products and services that improve their health, fitness, nutrition, appearance, sleep, and mental well-being. This growth is fueled by heightened consumer interest and a growing array of wellness offerings.

We’re doing great work with health and wellness businesses at HOP, let’s have a chat.

The anatomy of a Facebook account heist

A thought-provoking piece reveals the growing problem of Facebook and Instagram account takeovers, often for monetary gains. Many culprits are in Vietnam, exploiting social media account vulnerabilities. The article reveals their systematic methods, which present a considerable risk to online security and privacy.

On Reddit

Gen Z girls are becoming more liberal while boys are becoming conservative.

No reactions here 😐

On X

MrBeast officially broke the retweet record on X. Big Man shattered the retweet record on X, with over 3.8 million retweets. The organic reach is noteworthy, with one tweet amassing 160 million views in just 72 hours, underscoring the formidable influence of X's platform. With just over a year since its launch, X's potential seems limitless.

Other reads

Measuring Saas Engagement

Is it time to revive the Arab Peace Initiative?

Afcon’s decision to allow 24 countries to play is paying off

At the heart of this exploration is the model-centric approach, which stands as a cornerstone in contemporary marketing strategies. This approach utilizes statistical and machine learning models to analyze, predict, and optimize various growth functions. Unlike traditional methods, a model-centric approach offers a more nuanced understanding of market dynamics, customer behavior, and the effectiveness of marketing efforts.

Bayesian Optimization

Bayesian Optimization (BO) is a strategy for the optimization of objective functions that are expensive to evaluate. It works excellently in marketing contexts where evaluating the impact of strategic decisions is resource-intensive. BO builds a probabilistic model of the objective function and uses it to select the most promising inputs to evaluate in the true function, thus balancing exploration and exploitation.

Application in Budgeting and Media Planning

In budgeting and media planning, BO can be employed to allocate resources efficiently. By iteratively updating the probabilistic model with results from previous campaigns, marketing teams can predict the most effective allocation of budgets across various channels.

Attribution and Performance Incrementality

For attribution and performance incrementality, Bayesian models offer a more sophisticated approach than traditional last-click models. They can account for the complex, non-linear interactions between various marketing channels and customer touchpoints, providing a more holistic view of what drives conversions.

Open Source BO Models

Several open-source models stand out for their utility in marketing optimization. LightweightMMM, a Bayesian Marketing Mix Modeling library, allows for easy attribution across channels. Another notable tool is Facebook's Prophet, a forecasting tool adept at handling the seasonality inherent in marketing data.

The Correct Approach

Achieving optimality in marketing requires more than just sophisticated models; it involves a strategic approach encompassing various aspects of marketing operations.

Attribution is not Incrementality, Seek Incrementality

Understanding the difference between attribution and incrementality is crucial. While attribution models assign credit to different marketing touchpoints, incrementality focuses on understanding the additional value generated by a marketing action. Incrementality testing, therefore, becomes vital in measuring the true effectiveness of marketing strategies.

Don't Invest Time on Tools You Could Not Scale

Scalability is a key factor when selecting tools and models for marketing optimization. Choose tools that not only address current needs but also have the capacity to scale as the business grows. Opt for solutions that integrate easily with existing systems and can handle increasing volumes of data.

Experiment as Much as You Can…… Reliably

Reliable experimentation is the backbone of a data-driven marketing strategy. Test & gather data on what works and what doesn't. However, ensure that these tests are statistically sound.

Models Still Need Business Priors

Priors are your domain knowledge about customer behavior, market trends, and business constraints. Even the most advanced statistical models cannot replace the nuanced understanding that comes from experience in the field. For instance, while a model might suggest a particular marketing channel as optimal based on historical data, business priors might indicate a shift in market trends rendering this channel less effective. Balancing data-driven insights with business acumen is key to achieving true optimality.

Calibrate Your Model

Models reflect reality and provide reliable predictions. This involves tuning model parameters, validating model predictions against known outcomes, and continuously updating the model with new data. Calibration ensures that your model adapts to changing market conditions and consumer behaviors, maintaining its relevance and accuracy over time.

Take Action

The ultimate goal of all these analyses and models is to inform actionable strategies. The insights garnered should lead to concrete steps that can enhance your processes. This could mean reallocating budgets to more effective channels, adjusting campaign messages, targeting different customer segments, or exploring new marketing platforms. The key is to translate data-driven insights into real-world actions that drive tangible improvements in marketing performance.

Thoughts on Growth is a thematic platform dedicated to the in-depth analysis and discourse on the multi-faceted nature of growth across different domains. Here's a technical breakdown of what such a title might encompass:

1. Economic Growth Metrics: A technical discussion on economic indicators such as GDP, GNP, and other metrics that quantify the economic health of a nation or region. It may delve into compound annual growth rates (CAGR), market size expansion, and per capita income changes.

2. Business Expansion Models: Exploration of frameworks like Ansoff's Matrix for market penetration, product development, market development, and diversification. It could include case studies on scaling strategies, network effects, and sustainable growth trajectories in various business models.

3. Psychological Underpinnings of a Growth Mindset: Analysis of the cognitive-behavioral patterns that define a growth mindset, drawing on Dweck's research. This may include discussions on neuroplasticity, motivation theories, and the psychology of resilience.

4. Algorithmic Growth Hacking: Technical strategies for digital market growth, such as A/B testing, funnel analysis, user segmentation, and viral coefficient optimization. It could cover the utilization of machine learning algorithms to predict and enhance user acquisition and retention.

5. Growth Equity Investment Strategies: Deep dives into investment approaches focused on mid-stage startups that show potential for exponential growth. It may include discussions on valuation methods, due diligence processes, and post-investment growth support.

6. Sustainable Growth in Environmental Context: Application of sustainability models to ensure long-term growth without depleting resources. This could involve discussions on the circular economy, renewable energy adoption, and the role of green technology.

7. Technological Growth Accelerators: Examination of the catalysts for technological growth, such as Moore's Law in the context of computational advancement, network effects in platform growth, and the diffusion of innovations theory for technology adoption.

"Thoughts on Growth" serves as a knowledge hub, blending research with practical, real-world case studies and expert opinions. It would likely target professionals, academics, and students who are deeply interested in the mechanics, implications, and strategies of growth within their industries and beyond.

Subscribe now