Trend analysis: How to turn imperfect data into marketing wins
- Sarah Crooke
- May 28
- 10 min read
Authors: Sarah Crooke & Myriam Jessier
Marketers love data. We obsess over numbers, reports, and dashboards. But here’s the thing—we’re never working with a complete dataset. Perfect data is a myth and chasing it can paralyze your decision-making. Instead, embrace data pragmatism. Leverage the data you do have instead of worrying about the data you don’t.
While transactional data needs to be 100% spot-on, analytical data can already be sampled, incomplete data. Just as human behavior is fuzzy, so is the behavioral data we collect.
Why? First, we miss things—some users block tracking, some actions aren’t recorded, and some behaviors happen outside our analytics tools. Then, even when we do collect data, platforms like GA4 sample it, meaning what we see is just an approximation.
This isn’t a conspiracy—it’s just the reality of big data. Processing every single data point in real time is expensive and slow, so platforms give us the best version they can, fast.
The truth is, no matter how hard you try, you are never going to get it all. Consider this: studies show that a sample size of just 70 times the number of variables can yield valuable insights, and you’re likely working with more data than some medical studies.
It’s time to shift your perspective from data perfection to data pragmatism—time to rely on trend analysis to make informed decisions faster, optimize campaigns more effectively, and ultimately deliver results.
Table of contents:
Data pragmatism: Your path to meaningful insights
Stop chasing perfect data accuracy and vanity metrics—both of which can lead you astray over the long run. Instead, by understanding the underlying trends and patterns that truly impact business outcomes, you can shape those outcomes in your favor.
Nike decided to do the exact opposite when it pivoted to a 'data-driven' approach focusing solely on digital direct-to-consumer sales—a very costly mistake…
This case exemplifies Goodhart’s Law: “When a measure becomes a target, it ceases to be a good measure.” Or, as Becky Simms puts it, “If it is easy to measure, it probably isn’t that valuable.”
By obsessing over easily quantifiable metrics, you risk ruining your business model.
A prime example of Goodhart’s Law in SEO is the obsession with keyword rankings. When you focus on this, you ignore zero-click searches and other elements. Another example in PPC: fixating on Quality Score can result in over-optimizing ad relevance at the expense of conversion intent.
Your objective as an analyst is to gain actionable insights, even if the data isn’t perfect. Here’s how:
Practical solutions — Prioritize widely comprehensible data over complex, specialist-only metrics (e.g., instead of building an intricate customer lifetime value (CLV) model that factors in dozens of variables, use average order value (AOV) and repeat purchase rate—these are easier to track, explain, and optimize).
Business-centric analysis — Just because you can measure it doesn’t mean you should. Align your data strategy with specific business needs, context, and objectives.
Efficient resource utilization — Streamline processes for rapid insights, minimizing obstacles.
Actionable communication — Present data in a clear, compelling manner that drives action and decision-making.
Measure vs. Target — Avoid Goodhart’s law at all costs. Extract insights from sampled data; resist vanity metrics that compromise genuine value.
TL;DR: Stop chasing perfect accuracy or complex metrics that don’t contribute to business objectives. Focus on measuring what matters.
Embracing data pragmatism is key to driving meaningful business outcomes. The best way to extract meaningful insights from imperfect data sets is trend analysis. Focusing on trends rather than absolute numbers means you can make informed decisions faster, without worrying about whether the data is sampled or not.
Trend analysis crash course
To get marketing insights from sampled data, you should aim for directional reporting by, as Dana DiTomaso advises, “focusing on the data that you do have and what you can do with it to generate . . . insights so that you know what to do in order to improve your marketing, your sales—ultimately whatever goal you're trying to accomplish with your website and your business.”
To that end:
Stop worrying about sampling
Beware of biased data
Use the three types of trend analysis appropriately
Follow best practices when using sampled data
Stop worrying about sampling
Whether we’re looking at 100% or 80% of our data, the trends hold steady. It’s not about having every single data point—it’s about seeing the big picture (which looks the same both ways).
Beware of biased data
While sampled data is acceptable, biased data can derail your analysis. Be aware of inherent biases, such as under-representation of privacy-conscious users or Safari users.
To detect bias:
Check against other data sources — Does the percentage of groups seem roughly the same (e.g., people from certain locations, devices, interests)?
Benchmark against industry standards — Major discrepancies may indicate tracking issues.
Use the ‘sniff test’ — If your data seems too good (or bad) to be true, it might be faulty or biased.
Takeaway: Instead of obsessing over individual keyword rankings, focus on metrics that indicate overall organic search health and user satisfaction.
Pro tip: Clearly document any assumptions you make during the sampling process and how they might affect your conclusions.
3 main types of trend analysis
To effectively leverage trend analysis in SEO, it’s essential to grasp the three main types of analysis:
Analysis type | Definition | Example |
Comparative analysis | Examine trends across different categories to identify emerging opportunities. | Use People Also Asked data to swiftly adapt to spikes in search behavior caused by news or seasonal events. |
Regression analysis | Explore relationships between variables to understand how changes in one area might affect another. | Historical searches for [COVID] have evolved from identifying symptoms to managing policies and symptoms, highlighting the need for content realignment. |
Time-series analysis | Track data points over time to spot long-term trends and seasonal patterns. | [Best running shoes] reveals a change over time. From questions about brand comparisons to an increasing focus on sustainability and performance. |
Key takeaway: Agencies can enhance their SEO efforts by leveraging keyword trendspotting strategies to anticipate market shifts, align content with evolving user intents, and optimize regional campaigns.
Tips on using sampled data
What to do | Details | Why it's important |
Define the question first | What are you trying to answer with the data? What action will you take if it goes up or down? | Staring at streams of data can lead to analysis paralysis. A clear question can focus your effort and keep you on track. |
Focus on percentages | Use percentages instead of whole numbers when comparing data. | Percentages are easier to compare across different groups and help show what went up and what went down clearly and easily. |
Analyze trends over time | Look at how data changes between different time periods. | Harness data pragmatism: Are your numbers going up or down over time? That’s the best way to understand performance and pinpoint what needs to be done. |
Examine averages carefully | Calculate averages but also use histograms to see the full distribution. | Looking at averages of numbers, like order value, time on site, and scroll depth, can help put a data story together. Warning: Averages can hide outliers. For example, if 99 people spend $10 and 1 person spends $1,000, your average order value is $19.90, which isn’t really telling the full story. |
Use visualization tools | Create scatter plots or box plots to visualize your data. | Dashboards help you quickly spot patterns, outliers, and relationships that might be missed in raw numbers. It's like having a bird’s eye view of your data. |
Consider the context | Interpret results with the full story behind it. | Numbers alone don’t tell the whole story. Understanding the context helps you determine if your findings are practically significant—not just statistically significant. Knowing a sale was occurring or that the mobile app was updated at a certain date can really help you contextualize the numbers. |
Mitigate inherent data bias by relying on multiple sources
You cannot expect one data source to capture all aspects of the user journey, and most have some bias toward the platform itself. Just as chasing the wrong metric can create destructive feedback loops, relying on a single data source also comes with its own problems.
For agency owners, marketers, and analytics specialists, integrating multiple data sources can mitigate the limitations of individual methods and reveal hidden patterns that a single source might miss. This comprehensive approach allows you to more accurately measure customer happiness and frustration.
The power of diverse data sources
Start by creating a measurement plan, which is what actually matters to the company (though revenue might also be key). What need is the brand fulfilling? And, how are you measuring that you are meeting that need?
Figure out the metrics first and then pick the right data source. This holistic view helps identify areas for improvement and enhances customer satisfaction.
Examples of data sources include:
Internal data — CRM systems, website analytics, sales data
External data — Social media analytics, market research reports, and customer feedback surveys
Qualitative data — Customer interviews, focus groups
Quantitative data — Transaction records, usage statistics
Tools and techniques for data integration
Leveraging robust tools and techniques is crucial to effectively combining data from multiple sources. One tool I highly recommend is BigQuery. BigQuery allows marketers to:
Access unanalyzed, large datasets quickly and efficiently.
Integrate data from various sources, such as Google Ads, Google Analytics 4, CRM systems, and more.
Run complex queries to uncover insights and trends.
Use built-in machine learning capabilities to predict customer behavior and sentiment.
By utilizing BigQuery, marketers can seamlessly integrate diverse data sources, providing a comprehensive view of customer happiness and frustration. This, in turn, enables more informed decision-making and strategic planning. At the end of the day, a client wants to know, ‘If I do X, Y, Z, will it get me more money or customers?’ You can use trend analysis to help answer this.
Custom metrics: Change the conversation, not the data
“Humans don’t necessarily measure their experiences in a finite way.” — Tom Haczewski
Custom metrics allow you to tailor your data analysis to your needs, offering more relevant and actionable insights.
One thing to keep in mind when creating custom metrics is to focus on metrics that align with your company’s unique values and goals. Start away from the tracking tools to plan custom metrics that are consistent with your business goals. Do not let them influence you.
Here’s how to build metrics that matter:
Articulate the challenge
Define custom metrics
Assemble your data sources
Identify patterns
Communicate trends and drive action
01. Articulate the challenge: Define a metric to track users’ happiness
Forget the average order value for a minute. Focus on customer happiness. What actions signal a great shopping experience? To figure it out and create a ‘happiness index’, you should:
Talk to your customers about what makes them happy.
Gather any customer feedback through surveys or stakeholders.
Review your Google Analytics or other on-site analytics to find insights on what users do on the site that show they are engaging with the brand.
02. Define custom metrics
For an eCommerce brand, this isn’t just feel-good fluff. It’s a direct line to customer orders and brand loyalty. Here are the five key metrics of our digital happiness index:
Page load speed
High-value orders
Returned purchases
Add to wishlist
Five-star reviews
03. Assemble your data sources
Combine data from various sources to mitigate their inherent limitations. This could include:
Google Ads & GA4 data
Google Merchant Center data
Algolia
etc.
04. Identify patterns
Look for trends and correlations in your assembled data.
Growth isn’t linear—it’s about understanding user behavior shifts and adapting accordingly. The index won’t tell you the solution but will guide you on where to look. A 31% drop in the Happiness Index means something changed.
Key data points (from the example above):
High-value orders fell from 40% to 10% (-75%)
Additions to the wishlist dropped from 20% to 5% (-75%)
Other metrics (page load speed, return purchases, five-star reviews) stayed relatively stable
Possible causes and tests based on behavioral funnels
Hypothesis | Short-term action | Long-term strategy |
Users are skipping the wishlist but converting faster. | Check conversion rate for first-time visitors. If they’re buying without wishlisting, the drop isn’t a problem. | Analyze whether wishlist adds are still a valuable KPI or worth replacing with a more relevant metric. |
A pricing shift (cheaper products) caused fewer high-value orders. | A/B test “Spend $X for free shipping” to nudge users toward larger purchases. | Examine pricing strategy over time. Adjust high-value item visibility if needed. |
Traffic sources changed (e.g., more low-intent visitors). | Compare conversion rates across different traffic sources to see if the audience mix changed. | Adjust SEO and paid traffic strategies to focus on high-intent keywords and audiences. |
Users are dropping off due to checkout friction. | Run session replays and heatmaps to identify where users abandon the checkout process. | Optimize checkout flow by reducing unnecessary steps and offering trust-building signals. |
Site performance (speed, UX) isn’t the issue. | Check core metrics to confirm no technical degradation has occurred. | Continue optimizing UX by tracking behavioral shifts and ensuring a seamless journey. |
A shift in product discovery behavior led to fewer wishlist adds. | Run a quick survey asking users if they use wishlists and why. | Introduce alternative engagement mechanisms (e.g., personalized recommendations). |
For example, the image above clearly shows that (compared to last month) high-order values and wishlist additions are down. The first step is to look into these specific issues. Was there a push for cheap products this month? Did fewer people add to the wishlist but more convert on the first visit?
The index won’t tell you the solution but will guide you on where to look. The goal for next month would be to implement marketing or site changes to increase these percentages. If high-value orders were down, you could offer free shipping for orders of a certain value.
05. Communicate trends & drive action
You also want to know what is negatively impacting performance. Use a similar process to find out what frustrates your customers.
This could include the opposite of your happiness metrics (though it should not be all the opposite), for example:
Page load speed (slow)
Rage clicks
Cart abandonment
One- and two-star reviews
Abandoned on-site searches
Tracking and analyzing these provides a clear next step for what elements of your user experience to audit.
Trend analysis = Insights + patterns + action
Embrace the messy reality of marketing data and get comfortable being uncomfortable. You can’t hide behind pristine Google Analytics numbers anymore—Goodhart’s Law is in full effect, turning our success metrics into targets that have lost their meaning.
Instead, your mission is to:
Dive deep into the data chaos, be pragmatic, and uncover those juicy trends that actually drive action.
Focus on the data points impacting your bottom line and banish the information overload.
Don't get tunnel vision with your data sources—mix it up! Blend insights from social media chatter, brutally honest customer feedback, and those nitty-gritty transaction records.
Remember, perfect data is a myth. Your job isn’t to chase flawless numbers but to be the trend-spotting, insight-generating marketing analyst your team needs.
With 20+ years of working in digital, from developer to account manager, Sarah works for some of the big-name brands in Australia and several not-for-profit and charity organizations. Her consultancy, Meliorum, works with clients in implementation, reporting, and analysis.
Myriam Jessier is an SEO consultant and trainer with 15+ years of experience. She loves to share her knowledge because it helps everyone make the web a bit more human and bot-friendly. Twitter | Linkedin