Introduction
After years of running campaigns and building dashboards in both Adobe Analytics and Google Analytics, I’ve stopped believing the edge comes from how much data you collect or which tool you use. The edge is in reading the numbers correctly and turning that into a marketing decision. Most teams think bad analytics means they need more data or a better tool. In my experience, that is rarely the real problem.
Three challenges come up again and again, and solving them in order is what separates teams that trust their numbers from teams that argue about them. The rest of this article walks through each one, in the order I would tackle them with a new client.
TL;DR
Your edge in analytics comes from reading the numbers correctly, not from collecting more of them. Three challenges decide whether your numbers hold up, and each one builds on the last:
- Read the core metrics correctly. Users, sessions, and page views measure different things, so treating them as the same leads to confident but wrong calls.
- Label your marketing at the source. Disciplined tracking codes are what make all the data downstream worth trusting.
- Aggregate carefully outside the tool. The hard part is combining analytics data with leads, pipeline, or budget once the totals stop matching the platform.
The first two are foundation. The third is where the real work lives, and where most of this article spends its time.
Challenge 1: How to Read the Data
Users, sessions, and page views often get treated as three versions of the same number. In reality, each one answers a different question. Mix them up, and you can make a confident decision on the wrong basis.
Users and Visitors
Users map most closely to real people and real audience size. When a client asks how big their audience is, they are really asking about users. Most of the decisions that matter, like budget, reach, and audience sizing, come back to that number. One thing helps when you work across both platforms. What Google Analytics now calls a user used to be called a unique visitor, and Adobe still uses that older term. Either way, it counts a person once over a period, so someone who visits daily for a month still counts as one. There is an honest caveat worth sharing with clients. User counts are built on cookies and device IDs, not verified identities, so they can overcount when someone clears cookies or uses more than one device.
Visits and Sessions
Sessions tell you how often that audience comes back and engages. I watch session duration and pages per visit to judge whether a channel drove a real visit or just a click that bounced. The bigger payoff is using visits to map the journey people actually take through the site. Once you can see the common paths, you can shape them toward the actions you care about, like form fills.
A few practical moves come straight out of this. When a page is pulling a lot of traffic, add a clear CTA so that attention has somewhere to go. When a blog post draws steady visits, link it to the product pages it naturally leads into. That gives readers an easy path from interest to intent. And when a CTA is not converting, A/B test its placement before you assume the offer itself is the problem. In my experience, small changes to where and how you ask tend to move form fills more than a full redesign does.
Page Views and Hits
Page views and hits sit at the most detailed level, which cuts both ways. They are great for diagnosing a single page and easy to over-read at the same time. A hit is the smallest unit, a request that fires on page load or on a tracked action. A single page load can fire several hits, so you can have more hits than page views, but never the other way around. These are used when running very focused tests on a page. For instance, trying to understand if a particular CTA placement needs to change or not. Running A/B test and comparing hit level tracking will help inform such decisions.
Challenge 2: Setting Up Marketing Channels for Proper Tracking
None of these metrics are trustworthy unless the marketing effort behind them is labeled correctly as it enters the system. That is what tracking codes do. It is also where most of the mess in a marketing stack begins.
Tracking Codes and Measuring Experiments
Tracking codes are what make an experiment measurable in the first place. Sloppy tagging does more than make a report hard to read. It quietly ruins the results before anyone opens a dashboard. In GA4, any UTM value that does not match an expected pattern gets pushed into an “(Other)” bucket and drops out of your channel report. The most common self-inflicted version is not implementing tagging before launching a campaign. Teams are left guessing about the traffic spike on a particular page is probably because of the campaign that was launched.
Centralized, Granular Tracking
Tools like Claravine make their case on enforcing a taxonomy at the point of entry. They auto-generate clean UTMs and campaign codes right down to the creative or keyword level, so your structure survives contact with real campaigns built by real people on deadline. I treat vendor claims as a vendor’s pitch, but the core idea holds up. When messy data enters the system, the platform combines it faithfully. The problem is that it combines it faithfully wrong, because it has no way to know the data was bad to begin with.
Is a Google Sheet Really That Bad?
Here is an opinion that surprises people. A spreadsheet full of tracking codes is not the villain everyone makes it out to be. What matters is having a central, automated way to build the tags, so the same structure gets enforced every time, no matter who creates the link. A sheet with strict validation can take you a long way. Think dropdowns of approved values, rules that force lowercase and block spaces, and formulas that build the final tagged URL for you. That is governance through automation, handled up front rather than cleaned up later. For teams just starting out, this is more than enough. In the end, the tool is just where the process runs, and the process itself is the real work.
Challenge 3: Aggregating Data Outside the Web Analytics Tool
This is the hardest of the three, and the one I tell every team to budget the most time for. It shows up the moment you try to connect web analytics data to leads, pipeline, or budget inside a full-funnel dashboard. Suddenly the numbers stop matching what the platform shows on its own screen. The first two challenges are about getting clean data in. This one is about what happens to that data once it leaves the safe walls of the tool, and it is where most of a marketing report’s credibility is won or lost.
Why the Numbers Don’t Match
Build that dashboard, and the totals almost never match what you see when you log into the platform directly. This is structural, built into how these tools work, so there is no bug to chase. Unique metrics, and users are the clearest example, are non-summable. The platform does not count unique users exactly at scale. It estimates them. So totals pulled from different slices of the same data will not line up, and the gap grows as volume grows. A few other things pile on top. Underneath all of it sits sampling, where the platform scales a sample up to be roughly right rather than exact. Put all of this together, and the real surprise is that anyone expected the numbers to match in the first place.
I have learned to raise this with clients before the first dashboard ships. Once you explain that the mismatch is expected, the conversation changes. It moves to “which of these numbers do we actually trust for this decision.” That is the useful one, and it is much harder to have after a client has already spotted the gap on their own.
Raw Data, API Duplication, and Identity
Getting user-level data out of the platform means handling a huge volume of raw records. That volume alone is where most teams get stuck. Native API connections often make it worse, because they tend to duplicate records on export, which inflates every count you build downstream.
Adobe’s data feed shows what “raw” really means. It is an unsampled, hit-level export delivered as batch files, one row per hit, and it needs heavy transformation before you can use it. The gap between that raw feed and the in-tool number can be large. On Adobe’s own forums, people have reported feeds running about 20% off the reported figure. Most of that comes from hits the interface hides by default, like bot traffic. To reconcile the two, you have to reapply the same filters the interface runs for you automatically.
Now add a CRM join on top, and you run into identity resolution. This is probably the root cause under this whole section. An anonymous cookie ID and a CRM email address are not the same kind of identifier, and matching them is genuinely messy. It is not a solved problem you can shortcut.
The mistake I see most often is treating identity resolution as a technical task to finish rather than a modeling choice to own. No export setting solves it. You choose the rules yourself: how sure a match has to be before you count it, and what you do with the records that never match. Those rules belong to whoever owns the report, not to the connector. The teams that get burned are usually the ones who assumed the tool made that call for them. They find out it didn’t when someone asks why the same lead shows up twice.
Do They Even Need to Match?
After all that, the more useful question is whether the numbers need to match at all. Used directionally, web analytics tells you which way a trend is moving and which channels are pulling their weight. For most marketing decisions, that is enough to act on. Chasing exact reconciliation between the platform and an outside dashboard usually costs more in engineering and analyst time than the extra precision is worth.
Marketers are not data accountants, and reporting built around contribution and trend serves them better than reporting built around one perfect, fully reconciled number. I think that is largely right, but it needs a counterpoint I would rather include than skip. Directional data is not always enough, especially for numbers going in front of stakeholders who expect precision. Too much visible uncertainty in a report can freeze a decision instead of supporting one.
The real skill is knowing, decision by decision, which numbers can run on direction and which ones truly need the harder work of reconciliation. A channel budget shift can run on the trend. A revenue figure in a board deck, the kind a forecast gets built on, usually cannot. Reading which is which is most of the job. Part of that is being honest with the client about the difference, rather than passing off an estimate as if it were exact.
Conclusion
The real value in analytics comes from getting clean, well-labeled data out of the tool and connecting it to outcomes the business actually cares about. The tool itself matters far less. Reading the metrics correctly, labeling your marketing effort at the source, and accepting directional truth over precision you cannot afford are three parts of the same discipline. The teams whose numbers are trusted most have usually done the unglamorous work of mastering the seams between their systems.
That matters far more than the size of their data warehouse or the price of their platform.