Marketing Analytics Done Right: Data Granularity and Sampling Tips with Windsor.ai

Getting the granularity of your data just right means picking the level of detail that actually makes sense for your analysis. Simple enough, but that’s only half the battle. Because if the data itself is incomplete or “sampled,” even the sharpest, most finely tuned report can end up leading you astray.

This guide takes you through both of those concepts. I’ll show you how to manage granularity without breaking a sweat and, more importantly, how the Windsor.ai platform keeps your data complete and trustworthy. With that kind of foundation, building reports that are not only powerful but genuinely useful becomes a whole lot easier.

What is data granularity?

Think of data granularity like looking at a map. A high-granularity view is like zooming all the way in to see individual streets, buildings, and even streetlights. It’s incredibly detailed. A low-granular view is like zooming out to see the entire country, showing you the major highways and city locations.

How does this apply to your data?

• High granularity example: Viewing the number of clicks per hour for a specific ad.
• Low granularity example: Viewing the total number of clicks per month for an entire campaign.

High granularity is perfect for in-depth analysis and troubleshooting, while low granularity is your best friend for spotting long-term trends.

How data granularity affects your dashboards

The level of granularity you choose has a big impact on how useful and efficient your dashboards are.

• Readability: A daily chart covering two years of data can quickly become overwhelming and noisy. By switching to weekly or monthly granularity, you smooth out the fluctuations and reveal clearer, more actionable trends.
• Accuracy: Metrics like cost per conversion or ROAS depend heavily on the right level of detail. If your filters or calculations are applied at the wrong granularity, you risk ending up with numbers that look precise but don’t actually reflect reality.
• Performance: Pulling transaction-level data into tools like Looker Studio might work for small datasets, but at scale, it slows everything down. Aggregating your data at daily or weekly intervals keeps dashboards responsive without sacrificing insight.

Recommendations for using time periods, grouping, and filters

When working in Windsor.ai’s Data Preview, the way you set time periods, grouping, and filters has a direct impact on both the accuracy and performance of your dashboards. Here’s how to approach each:

Time periods

• Daily: Use when monitoring active campaigns, budget pacing, or experiments.
• Weekly: Best for identifying meaningful performance trends without daily fluctuations.
• Monthly/quarterly: Ideal for executive reporting and long-term strategy reviews.

Grouping (via dimensions)

Grouping in Windsor.ai is controlled by the dimensions you select in the Field Picker.

Examples: 

• Select only Campaign for a high-level view.
• Add Ad Group or Keyword if you need to drill into specifics.

Fewer dimensions means simpler, faster dashboards. More dimensions mean deeper insights, but also more complexity.

Filters

It’s recommended to apply filters directly in Windsor.ai using Advanced Filters rather than only in your BI tool.

You can use filters to:

• Limit data to relevant date ranges.
• Exclude inactive campaigns or test accounts.
• Focus on specific channels, geographies, or product categories.

Filtering early ensures your connected dashboards run faster and stay unsampled.

Summary: Choose a time period that matches your analysis, only group by dimensions you need, and filter early in Windsor.ai to keep dashboards lean, fast, and trustworthy.

Data sampling

Imagine your e-commerce site had 500,000+ visitors last month, and you want to know how many of them came from your summer sale email campaign. Instead of looking at all sessions, a sampled report might only analyze 50,000 of them and then estimate the total. This is, in essence, what data sampling is.

Platforms like Google Analytics 4 use this technique to process huge volumes of data quickly. When you create a custom report or look at a long date range, they often analyze just a sample of your data to provide a fast response.

But what happens if, by chance, the 50,000 visitors in the sample included a higher-than-average number of people from your email campaign? Your final report would overestimate its impact.

How data sampling affects your dashboards

Data sampling usually introduces uncertainty. When you make decisions based on a sample, you’re working with an educated guess, not the full story. This can lead to:

• Incorrect conclusions: You might think Campaign A is outperforming Campaign B, but in reality, the sampled data is misleading you, and there’s no significant difference.
• Bad budgeting decisions: Shifting your budget based on incomplete data can lead to significant financial losses.
• Loss of trust: When your team can’t trust the data, it undermines your entire analytics culture.

A common scenario: data sampling in Google Analytics 4

If you’ve ever spent time building reports in Google Analytics, you’ve likely run into data sampling. It’s a common frustration for marketers who need to dig deep into their data. While standard, out-of-the-box reports are generally unsampled, the moment you try to customize your analysis, you risk hitting a sampling threshold.

When GA4 doesn’t sample data:

• Standard reports: All built-in reports in the Reports section remain unsampled, even if you add filters, segments, or secondary dimensions.

When GA4 samples data:

Sampling in GA4 is most often triggered when you:

• Use explorations: Funnels, path analysis, cohort analysis, or segment overlap may be sampled if your dataset is large.
  
• Run large or complex queries: If the query involves more than about 10 million event rows, GA4 may sample the results.
  
• Apply multiple breakdowns: Adding several filters, segments, or custom combinations in explorations can trigger sampling.
  

How to spot data sampling:

You can easily check if your data is being sampled by looking at the top of your Google Analytics report.

When you see a small green checkmark in the top-right corner of your report, it means the data is unsampled. All results are based on the full dataset, and you can trust the numbers without concern for approximation.

If a yellow triangle with an exclamation mark appears, GA4 is telling you that the report is sampled. You’ll also see a short message above the chart or table (e.g., “This report is based on X% of available data”). This means the figures shown are calculated from only part of your dataset, so they may not exactly match your true totals.

Why is data sampling a problem in practice?

Imagine you’re comparing the performance of two ad campaigns. You add a secondary dimension to see which device users are on, and the yellow triangle appears. The report tells you that Campaign A has a conversion rate of 10.5% and Campaign B has a rate of 8.3%. You might be tempted to shift your budget to Campaign A.

However, because the report is based on a small sample, the reality could be that there is no statistically significant difference between the two. The small slice of data that GA4 analyzed just happened to have a few more conversions for Campaign A.

Making a budget decision based on this misleading, sampled data could be a costly mistake.

Windsor.ai: Your solution to data sampling

This is where Windsor.ai helps. Rather than relying on those sampled reports, Windsor.ai connects directly to the APIs of your marketing and analytics platforms. It then pulls the data in smaller chunks and pieces it back together, so you get the most complete version of your data. Instead of partial or estimated data, you get the complete view every time with full insights you can actually trust, which means:

• Complete data accuracy: Your reports are based on all your data, not just a fraction of it.
• Confidence: You can make critical decisions with full confidence in the numbers you’re seeing.
• True granularity: You can drill down into the most granular details of your data, knowing that what you see is real and not an estimation.

Practical examples of using granularity and sampling in Looker Studio & Google Sheets

Here are four distinct, scenario-based examples that show how both granularity and sampling can have a real-world impact on your decisions.

Example 1

The goal: You’ve just launched a new Google Ads campaign, “Campaign C,” with a strict target Cost Per Acquisition (CPA) of $50. After 30 days, you need to decide: should you scale the budget or kill the campaign?

Monthly view (low granularity):

Your first step is to check the overall performance. You build a simple scorecard in Looker Studio showing the average CPA for the last 30 days. The result is $65. Based on this single number, the campaign looks like a failure. It’s 25% over your target, and the obvious decision seems to be to pause it immediately.

Weekly view (medium granularity):

But before you do, you get a little curious. What if the monthly average is hiding something? You build a time series chart and set the date granularity to “ISO Year Week”. Now you see something interesting: the CPA was very high in the first two weeks, but it has been trending down. Last week was actually under your $50 target.

Daily view (high granularity):

For the final step, you duplicate the weekly chart and change the date granularity to ‘Date.’ Now the picture is crystal clear. The chart shows that the campaign had a learning period where the CPA was high, but your optimizations have worked. For the last 10 days, the CPA has been consistently and profitably below your $50 target. The campaign has found its footing and is now a winner.

The takeaway: Relying only on the low-granularity monthly average would have caused you to kill a successful campaign. By increasing the granularity, you uncovered the true trend and made the right decision to scale the budget.

Example 2

The goal: You need to identify your top 5 organic landing pages from the last six months to decide where to focus your SEO and content improvement efforts.

Report A:

You open Looker Studio and use the native Google Analytics connector. Because you’re looking at a long date range, you know the data will be sampled. You build a simple table with ‘Landing Page’ as the dimension and ‘Sessions’ as the metric. The table gives you a clear list of what you believe are your top 5 pages.

Report B:

Just to be sure, you build the exact same table on a new page, but this time you use the Windsor.ai connector as your data source. Because Windsor.ai provides complete, unsampled data, the numbers look a little different. In fact, when you sort the table by sessions, you realize that the true list of top 5 landing pages is different from what the sampled report showed you. One of your “top” pages from the first report isn’t even in the top 10.

The takeaway: The sampling in the native analytics hid the true performance of your landing pages. Making strategic decisions based on that flawed report would have led you to waste time and resources optimizing the wrong content.

Example 3

The goal: You’re looking at your Google Ads performance, and you see that your “Generic” campaign has a low Return on Ad Spend (ROAS).

Low granularity view:

You start by creating a pivot table with “Campaign” as your rows and “ROAS” as your values. This confirms the problem: the “Generic” campaign is indeed underperforming.

Medium granularity view:

To find out why, you add “Ad Group” as a second dimension to your pivot table’s rows. Now you can expand the “Generic” campaign and see all the ad groups inside it. It immediately becomes clear that the “Running Shoes” ad group is the one dragging down the campaign’s performance.

High granularity view:

For the final step, you add “Keyword” as a third dimension. Now you can expand the “Running Shoes” ad group and see the exact keywords that are spending money with a low ROAS.

The takeaway: You’ve successfully moved from a high-level problem to a specific, actionable list of keywords to pause or optimize.

Example 4

The goal: You need to find your top 10 keywords by conversions from a high-volume campaign over the last quarter.

Poor approach: You use a standard Google Analytics add-on for Sheets. The query returns sampled data. You create a pivot table and filter for your top 10 keywords by conversions. You now have what you think is your list of winners.

Effective analysis: You run the exact same query, but this time using the Windsor.ai Google Sheets add-on. When you build the same pivot table, the conversion counts for many of the keywords are different. The unsampled data reveals that two of the keywords from your original “top 10” list aren’t even in the top 20. The sampling skewed the numbers, giving you a flawed list.

The takeaway: Relying on the sampled data would have led you to put more budget and effort behind the wrong keywords, while your true top performers would have been ignored.

Conclusion

Data granularity and sampling aren’t just nerdy details; they actually make a huge difference in how useful your marketing insights are. The way you slice your data and whether you’re looking at every single point or just a sample, changes what your dashboards show, how fast they load, and how much you can trust what you see.

With Windsor.ai, you don’t really have to pick between speed and accuracy. You can tweak granularity right in the Data Preview, picking the dimensions that actually matter. As you can pull full, unsampled data from your marketing platforms, you don’t have to worry about missing stuff or being misled by a sample. In the end, your dashboards end up fast, reliable, and actually telling the story of your campaigns.

Next time you’re building a report, just pause and think, “Am I looking at enough detail? Is my data really complete?” That’s usually enough to steer your choices, and it’ll help you turn raw numbers into insights you can actually use.

Ready to see your data in full detail and make decisions with complete confidence? Try Windsor.ai for free today and unlock unsampled, fully granular insights for all your campaigns.