What geographic data does Paramark need before designing a geo experiment?

Last updated: September 29, 2026

If you want Paramark to help design a geo experiment, the most important starting point is not the market list itself. It is the historical outcome data at the same geographic level where the test will run.

In practice, that means we usually need your target metric broken out by state or DMA over time before we can confidently recommend test and control markets.

What data we typically need

Before a geo experiment design starts, Paramark usually looks for four things:

1. A geo-level target metric. Examples include free signups, purchases, qualified leads, or new paid customers, broken out by state or DMA.
2. Historical depth. We typically want several months of clean historical data so we can see how markets move over time and find strong test/control matches.
3. Consistent definitions. The metric should mean the same thing throughout the history. If tracking logic changed, that can weaken the design.
4. Enough market coverage and volume. Geo tests work best when there are enough eligible markets and enough conversion volume in those markets to detect lift.

If one of these is missing, that does not always mean the experiment is impossible. It usually means we first need a feasibility check, a different metric, a different geography level, or a different test structure.

What “geo-formatted” data means

When teams say they want data “in a geo format,” they usually mean that the outcome metric can be reported by geography and by time period.

For example, instead of only knowing total weekly free signups across the whole business, we would want something closer to:

Weekly free signups by state
Weekly new paid customers by state
Or, for some channels, the same metrics by DMA

If you are considering new paid customers as the target metric, it is also worth confirming exactly what is included before the design work starts. For example, teams often need to verify whether enterprise or sales-led customers are included, whether any channels or customer segments are excluded, and whether that definition has stayed consistent over the historical period. A metric can look viable at first glance, but if its scope changed over time or omits part of the business the campaign is meant to influence, that can weaken the design or point us toward a different primary metric.

This structure is what allows Paramark to compare how different geographies behaved historically and build a design where test and control markets track closely before launch.

State vs. DMA: how to choose the right geography level

The right answer is usually the most granular geography where the outcome metric is both consistently available and high-volume enough to support a credible design.

Use DMA-level data when the business can report the target metric cleanly at DMA level and there is enough historical volume in each market.
Use state-level data when DMA is unavailable, too sparse, or too inconsistent over time.

In practice, Paramark often evaluates both options during the feasibility review. The better level is not always the smaller geography — it is the one that gives the cleanest historical signal for matching test and control markets.

Before you use “new paid customers” as the target metric

If your team wants to use new paid customers as the decision metric, make sure the definition is explicit and stable before the experiment is designed.

A few questions matter right away:

Does the metric include every acquisition path you care about? For example, if enterprise or sales-led customers are strategically important, confirm whether they are included or excluded.
Is that inclusion logic consistent historically? A changing definition can make the metric look like it moved when the instrumentation changed instead.
Is the metric available at the same geography level as the test? A national metric is not enough for a geo experiment.

If those answers are still unclear, Paramark may recommend treating new paid customers as a secondary metric until the definition is confirmed and the geo-level series is validated.

How to choose the right geo-level metric

Not every business metric makes an equally strong experiment target. A good geo-level target metric is usually:

Close enough to marketing that changes in media can realistically move it during the experiment window
High-volume enough that real signal can be detected above normal market noise
Consistently tracked by geography across the historical period
Available quickly enough that the experiment can produce a timely answer

That is why teams often discuss both a more down-funnel metric and an earlier-funnel metric together.

For example:

New paid customers may be closer to the real business outcome
Free signups may arrive faster and in higher volume

In many cases, one of these becomes the primary decision metric, while the other is used as a secondary or contextual metric.

What if we want to evaluate multiple metrics for the same campaign?

That is common, especially for launches or seasonal campaigns.

The best practice is usually to define:

One primary metric that answers the main business question
One or two secondary metrics that help explain what happened higher up or lower down the funnel

For example, a team could frame the experiment like this:

Primary metric: new paid customers
Secondary metric: free signups

Whether that is the right setup depends on the volume, lag, and stability of each metric by geography. If the down-funnel metric is too sparse or too delayed, Paramark may recommend using the earlier metric as the primary readout.

How seasonality affects the design conversation

Seasonal campaigns often make teams wonder whether geo testing is still feasible. Usually, the answer is yes — but seasonality raises the importance of having clean geo-level historical data.

In a geo experiment, both test and control markets experience the same broad seasonal period at the same time. That helps the comparison remain valid. But if the chosen outcome metric is highly volatile, inconsistently tracked, or too sparse at the geo level, seasonal noise can still make the design weak or expensive.

That is why the first design conversation often focuses on the data itself before anyone finalizes the exact market list.

What to expect in the first design discussion

Once geo-level data is available, Paramark can usually help assess:

Whether the metric is viable for testing
Whether state or DMA is the better geography level
Whether there are enough eligible markets
What tradeoffs exist between budget, duration, and detectable lift
Whether the proposed primary and secondary metrics make sense

In other words, the first step is often not “here are your final geos.” It is “do we have the right geo-level data to produce a credible design?”

What if the experiment plan changes after the first draft?

That is normal. A geo experiment design is only as good as the channel plan and market constraints behind it.

If your team changes the planned holdout, shifts spend between channels, introduces launch timing changes, or rules certain geographies in or out, Paramark will usually revisit the design rather than forcing the original market split.

In practice, that can mean revisiting:

Which geographies are still eligible for test and control
Whether the primary metric is still the right one
Whether the revised spend plan still creates enough signal to measure
Whether a different geography level or test structure is now more appropriate

A refined plan is helpful, not disruptive. The more clearly the final channel plan and market constraints are defined before launch, the more credible and execution-ready the experiment design will be.

Bottom line

To begin a geo experiment design, Paramark usually needs historical outcome data broken out by geography and time. If you are deciding between metrics like free signups and new paid customers, we can usually evaluate both — but the primary metric should be the one that best balances business relevance, volume, speed, and data quality at the geo level.

If that data is available, the next step is a design and feasibility review. If it is not yet available, the next step is usually to instrument or export the metric in a consistent geo format first.