Platform
Use Cases
Many Possibilities. One Platform.
AI and Automation
The Always-on Incrementality Platform
Solutions
Teams
Built for your whole team.
Industries
Trusted by all verticals.
Mediums
Measure any type of ad spend
Use Cases
Many Possibilities. One Platform.
AI and Automation
The Always-on Incrementality Platform
Teams
Built for your whole team.
Industries
Trusted by all verticals.
Mediums
Measure any type of ad spend

Try this: write down everything you remember about the last restaurant you went to. Then rate how confident you are, from 1 to 10.
We ran this exercise with a room full of senior marketers at Smartly's Customer Advisory Board. The first details came fast. Dim lights. Long table. Who sat across from you. Confidence: a solid 8, maybe a 9.
Then come the specific questions. How many windows were in the room? What color were the chairs? What music was playing? Confidence dipped a little. Accuracy dropped a lot.
The brain didn't store the room. It stored a summary - "nice place, good food, a bit loud, sat across from Andy" - and every time people recall it, it rebuilds the details from that summary. Confidently. Often wrong. Psychologists have known this for close to a century: memory is reconstruction, not playback.
Marketing ran on the same trick for two decades. Reports stored the summary. Analysts rebuilt the details from context that lived in their heads. That worked while humans made the decisions. It breaks the minute an AI does.
Back in 2018, Laura Desmond, now Smartly’s CEO, wrote that marketing's greatest looming crisis was the loss of intelligence. Too much data, too little understanding, and an industry on a fast track to becoming dumber. AI was supposed to be the cure. Fed the wrong signals, it just gets us there faster.
The fix starts at the smallest level there is. The atomic level.
There's a historical parallel I really like. In the late 1800s, when factories first got electricity, most of them simply swapped the giant steam engine for a giant electric motor. Same factory floor, new power source. Productivity barely moved. It took roughly four decades to redesign factories so that every machine had its own small motor. That's when productivity took off.
That's where most marketing teams are with AI today: copilots and generated copy bolted onto the old workflow. The capability is there. The factory just isn't built for it.
AI is our electricity. The real shift is rebuilding the factory.
Rebuilding means taking a workflow like "build the Q4 media plan" and breaking it into its smallest units: set channel budgets, forecast sales, QA the setup, monitor pacing, spot creative saturation, reallocate. Each one is a task an AI marketing agent can own, through the skills and MCPs the platforms are now shipping. The work is becoming atomic.
But there's a part almost nobody is talking about: an atomic task is only as good as the signal it runs on. Miss one, swap one, mislabel one, and the agent won't fail loudly. It will fail confidently.
When I wrote about the Marketing Data Engineer, I shared what Advertiser data often looks like. Spend reported only as "Google, global." 50% of conversions from an "unknown" country. TV spend dumped on the 1st of the month because that's how the agency reports it.
In chemistry terms, these are molecules: several atoms bonded into one thing that can't be pulled apart again. You can always aggregate up. You can never disaggregate down.
A human analyst could live with that, because the missing context lived in their head. They knew the TV "spike" on the 1st wasn't real, that "Unknown" really meant the US, and which campaigns were quietly retargeting existing customers while claiming new ones.
An agent has none of that context unless you give it to it. Hand it a molecule and it will do exactly what you did with that restaurant: rebuild the missing details from the summary, with total confidence.
In creative, we call that an AI hallucination, and we can see it - a wrong logo, a sixth finger.
In data, nobody sees it. Well, not until it’s way too late.
Atom (Atomos in Greek) means "uncuttable." (Ironic, since we did eventually split the atom.) But the idea holds: an atom is the smallest unit of something that still behaves like that thing. One atom of gold is still gold. Split it, and none of the pieces are.
Borrow that for marketing, and you get this: an Atomic Signal is the smallest unit of marketing data that still carries meaning on its own - enough meaning that an AI can act on it without a human there to explain it.
Three properties make a signal atomic:
Here's what the gap looks like in practice:
| The data we often see today | The signals agents actually need | |
|---|---|---|
| Digital spend | "Google, global, October" | Spend per campaign, country, platform and day |
| TV | Monthly budget booked on the 1st | Daily impressions or GRPs per market, bonus delivery included |
| Influencer | One $1M line item | Spend allocated by post date, country and reach |
| Conversions | "New user," as your 3rd party attribution defines it by default | Your own definition, written down and attached to the data |
You can get every input right - spend, delivery, context, definitions - and still get it wrong, because of one atom that doesn't exist in any platform report.
Every agent optimizes toward a reward. A score that says "that worked, do more" or "that didn't, stop." In most marketing stacks today, that reward is an attributed conversion, , the output of marketing attribution
Attributed conversions are fool's gold. Shiny, convincing, and often worthless. In our attribution vs. incrementality benchmark, covering 65 enterprise Advertisers and over $1B in spend, Social received less than a third of the credit it actually deserved. Search, meanwhile, got over-credited right up to the cliff where the next conversion costs more than the customer is worth.
Feed that reward to an agent and it will do what agents do best: mine fool's gold faster, and at a bigger scale, than any human team ever could.
The real reward atom - did this cause an outcome that wouldn't have happened otherwise? - isn't collected. It's measured. Continuously, and at the same granularity as every other atom. That's the loop: the agent acts, incrementality measures what it actually caused, and the next decision gets sharper. Marketing measurement stops being a report someone reads on a Monday and becomes the reward function the machines learn from.
Six months ago, INCRMNTAL became part of Smartly. This is why.
A reward signal is only worth something if a system is listening to it. Smartly is where the atomic work happens - creative, media and execution across channels.
Put the reward atom right next to the work it's meant to reward, and the loop finally closes.
Today we’re announcing INCRMNTAL Connect as part of Smartly’s intelligence suite of products.
Today we're announcing INCRMNTAL Connect as part of Intelligence Pro, Smartly's new intelligence offering.
Remember the factories? Productivity didn't take off when electricity arrived. It took off when every machine got its own small motor, right where the work happened. Intelligence works the same way. A signal that shows up in a report three weeks later is the giant engine in the basement. A signal that shows up at the moment of decision is the motor on the machine.
That's the thinking behind Intelligence Pro. It brings four capabilities together, each one wired to a question marketers ask before they move budget.
Where should I invest? Reach Planner compares media scenarios and forecasts cross-channel reach and frequency, before anything goes live. Where can I scale? Headroom Analysis shows where more budget still has room to drive growth, and where returns start to flatten.
How does my performance compare? Benchmarks put your results in context, using aggregated, anonymized performance data. And where should I investigate or optimize next? That's the job of INCRMNTAL Connect. It brings the reward atom to the table: an incrementality-based view, sitting right next to what the platforms report.
Connect wires INCRMNTAL's signal into Smartly's factory floor - Reporting, Predictive Budget Allocation and Triggers - so budget and automation decisions follow what actually moved each KPI, not whichever campaign claimed the most conversions.
It starts with the media you run through Smartly. For everything else - every other channel, organic included - there's the full INCRMNTAL platform.
None of these signals is the whole picture, and none of them pretends to be. One number that claims to answer every marketing question is just another summary, and you already know what happens when you trust a summary with total confidence. Each signal answers its own question, and they all work alongside the other measurement methods you already use. What changes is where they show up: at the point of decision, instead of glued together from separate reports after the fact.
The signal is only half of it, though. The inputs are still on you. Define your outcomes clearly. Know how your data is collected and labeled. Decide which signal belongs to which decision.
In an agentic world, owning your data strategy isn't optional. Every definition you leave to a platform default, every report you accept pre-blended, is a molecule your agents will have to guess their way through.
AI will make the factory run faster. It won't decide what the factory should build. People still set the objective, and people still have to look at what comes out and ask: does this actually make sense?
The next advantage in marketing AI won't come from automation alone. Everyone will have the electricity. It will come from systems that understand what actually creates value: better signals, connected to better decisions, and made actionable across the entire campaign lifecycle.
Marketing in an agentic era becomes intelligent one atomic signal at a time.

Maor is the CEO & Co-Founder at INCRMNTAL. With over 20 years of experience in the adtech and marketing technology space, Maor is well known as a thought leader in the areas of marketing measurement. Previously acting as Managing Director International at inneractive (acquired by Fyber), and as CEO at Applift (acquired by MGI/Verve Group)