See exactly what your model sees: Introducing Model Center

A mechanic can tell you your car is fine and expect you to nod along, or they can pull up the diagnostic, point to the exact reading, and show you why. One of those builds trust while the other asks you to just take their word for it.

Marketers have spent years in the second category with their MMM. The model says a channel drove a certain amount of revenue, that a campaign is saturated, that a new budget plan will outperform the current one, and there's rarely a way to check the model's reasoning. You either trust it or you don't, and that's not much of a choice.

Being able to see how a model reached its conclusion matters because it helps you defend a change to your CMO, your CFO, or your own team when they ask the obvious follow-up question: how do you know? Model Center gives you that visibility without asking you to become a data scientist first.

Key takeaways

• Model Center brings every model you run into one place, instead of buried in a settings tab with unclear names.
• Model Accuracy gives you a single 0-100 score for how well a model reflects your real results.
• The Model Health tab shows backtests and confidence bands, so you can see how the model performed against outcomes it wasn't trained on.
• The Attribution tab shows where the model says your results are actually coming from, by channel.
• The Priors tab shows the assumptions a model started with, and how those assumptions have changed as it's learned from your data.
• The Config tab keeps a full record of what changed in a model and when, so a shifted recommendation never has to be a mystery.
• None of this requires waiting on a call with your CSM to walk you through it.

You've been asked to trust the model. Now you can see it.

MMMs have always been somewhat mysterious to the people using them. That's not because the modeling behind them is weak. In Prescient's case, it's the opposite: the methodology is the most sophisticated part of the platform, built on years of research into how marketing actually behaves in the real world, not how a textbook says it should. The problem was never the modeling but that you, our clients, had no real way to see it.

That gap creates a strange dynamic. A model can be right and still get overridden or ignored because nobody could explain why it said what it said. Model Center closes that gap by putting the model's work, not just the output, in front of the people making decisions with it.

A home base for every model you run

Before Model Center, finding your models meant digging through a settings page with no quick sense of which model was actually healthy. That's a rough way to start a planning conversation, and an even rougher way to explain your measurement approach to someone outside your team.

Model Center replaces that with a single overview page showing every active model you run, side by side. Each model gets its own card with its accuracy, its spend, and the channels feeding it, so you can scan the health of your entire modeling setup in seconds. If something looks off, you know exactly where to look next instead of guessing which model needs attention.

A number that tells you how much to trust a forecast

Not every marketer wants, or needs, to understand the statistics behind a model. What they do want is a fast, reliable way to know if a model's output is worth acting on. That's what Model Accuracy is for.

Model Accuracy is a single score from 0 to 100, built from a standard forecast-error measure called SMAPE. A higher score means the model's predictions have tracked closer to what actually happened. You don't need to know how SMAPE is calculated to use the score, you just need to know that a strong score means you can move forward with more confidence, and a weaker one means it's worth digging deeper before you shift budget based on that model's recommendation.

Seeing the model's work in Model Health

A single accuracy score is useful, but it doesn't tell the whole story on its own. The Model Health tab is where you see the work behind that number so you can understand its context.

Model Health shows a backtest chart comparing your actual results against what the model would have forecasted, including a holdout period the model was tested against but never trained on. That holdout comparison is the closest thing to a fair test of whether a model actually works, rather than one that just memorized your past data and called it a win. Alongside the backtest, confidence bands show how certain the model is around its own forecast. Wider bands mean more uncertainty, and narrower bands mean the model is more confident in what it's telling you.

Attribution: where the model says your results are coming from

Knowing a model is accurate is one thing. Knowing what it's actually attributing your results to is another, and that's where the Attribution tab comes in.

Attribution breaks down how the model is crediting your results across channels for the model and time period you're looking at. Instead of taking a recommendation at face value, you can trace it back to the channel-level story behind it and decide for yourself whether that story lines up with what you're seeing on the ground. If a recommendation doesn't match your own intuition, this is where you go to understand why the model sees it differently.

Priors: the assumptions your model started with

Every model starts somewhere. Before a model has learned anything from your data, it starts with a set of assumptions about how each channel is likely to behave. Those starting assumptions are called priors, and they've historically been invisible to clients, tucked away as an internal detail rather than something anyone outside the data science team could review.

The Priors tab shows those assumptions for every channel, along with a timeline of how they've changed as the model has run and learned more about your business over time. If a model's output shifts, this is often where the "why" actually lives. Instead of wondering whether the model just changed its mind for no reason, you can see the exact assumption that moved and roughly when it happened.

Config: a record of what changed and why

A model's configuration, the settings and structure behind how it runs, can shift over time as your business, channels, and data change. Without a visible record of those shifts, it's easy to assume a model just started behaving differently on its own.

The Config tab keeps a full history of every configuration change, so you can see exactly what was adjusted and when it happened. When a recommendation looks different than it did last month, you can check whether something about the model actually changed, instead of guessing or assuming the model is simply less reliable than it was before.

Wrapping it up…

Model Center is a direct extension of what Prescient has always believed: that the strength of an MMM comes down to the rigor of its methodology, and that rigor is worth showing. Model Accuracy, backtests, priors, and configuration history all exist so you don't have to take a model's word for it. You can see the reasoning for yourself, and bring that same clarity into every budget conversation you have going forward.

The Model Center is now live in the platform for clients, at no additional cost.