Marketing Cloud reporting is the work of getting trustworthy numbers out of the platform and reconciling them with the rest of the business — through data views, tracking extracts, Marketing Cloud Intelligence and Data 360. Twopir Consulting builds that measurement layer so marketing can report contribution rather than channel metrics, and so the numbers hold when finance looks at them.
Most Marketing Cloud reporting problems are routing problems: the number exists, in a place nobody looked, in a form nobody trusted. This is the routing table we work from.
| The question | Where the answer lives | How it gets out | The caveat |
|---|---|---|---|
| Did this send perform? | Tracking data views inside the platform. | Standard reports for a quick look; a data view query for anything repeatable. | Fine for a single send. Assembling a trend by hand from standard reports is where reporting time disappears. |
| How is engagement trending over months? | Tracking data, but outside the platform where it can be held long enough to see a trend. | Automation Studio extract and file transfer to a warehouse or SFTP destination. | Retention inside the platform is finite. A trend nobody exported is a trend that quietly stops being available. |
| What did marketing contribute to pipeline? | Both systems — engagement in Marketing Cloud, outcomes in CRM. | Send and engagement data written back as CRM activity, plus campaign association. | This is where the two systems disagree. See the next section, which is the whole answer. |
| Which audience is actually active? | Tracking data views, aggregated per subscriber rather than per send. | A query that computes engagement recency and frequency into its own data extension. | Almost never available as a standard report, and almost always the most useful number in the platform. |
| What is our cross-channel performance? | Marketing Cloud Intelligence, formerly Datorama. | Connectors into Intelligence, with harmonized dimensions across channels. | Value depends entirely on the quality of what feeds it. It will not repair inconsistent definitions upstream. |
| Who is this customer, across everything? | Data 360, formerly Data Cloud. | Identity resolution across sources, then calculated insights. | A platform program rather than a reporting task. Worth it when the answer must serve sales and service too. |
Scroll the table sideways to see every column
Capabilities verified against Salesforce documentation Last reviewed 23 September 2026
They are counting different things, over different periods, about different populations — and each is internally correct. The mismatch is not a bug and it is not usually a broken integration; it is four definitional differences that nobody wrote down.
Once they are written down, the two can be reconciled to an agreed variance and reported together. Until they are, every monthly review spends its first twenty minutes arguing about whose number is right.
Each of the four below has a specific remedy, and none of them requires new tooling.
Marketing Cloud counts subscribers and contacts, which may include records that exist in no CRM object — imported lists, app users, event registrants. CRM counts leads and contacts, which may include records marketing never messages. Neither population is wrong; they simply do not overlap completely, and until somebody states which population a report is about, the totals cannot agree.
A marketing report usually attributes an outcome to the send that preceded it. CRM campaign influence models attribute across a window, sometimes to several campaigns at once, sometimes weighted. Both are defensible; averaged together without being reconciled, they produce a number that is neither.
Send and engagement data written back into CRM is not instantaneous, and the write-back cadence is a configuration choice. A report run on the first of the month against a window that closed on the last of the previous one will be missing the tail of the activity, every time, and by a consistent amount nobody has quantified.
Whether a bounce counts as a send, whether a resend counts twice, whether an unsubscribe belongs to the send that prompted it or the month it occurred in. These are choices, and two systems configured by two teams made them differently. Writing one definition per metric and applying it in both places is unglamorous and it is the entire fix.
Dashboards come last. A dashboard built before the definitions are agreed is a faster way to distribute numbers people do not trust, and it is where most reporting projects begin.
One definition per metric, signed off by marketing and by whoever will challenge the number later. Dull, unbillable-feeling, and the single highest-leverage hour in a reporting engagement — most reporting disputes are definitional rather than technical.
Data views answer questions inside the platform; extracts move engagement data somewhere it can be held long enough to show a trend and joined to everything else the business measures.
Engagement data flowing back into CRM as activity, campaign association configured deliberately, and a documented variance between the two systems so that the difference is expected rather than argued about.
Built last, in whatever the business already uses. An executive view answers three questions on one screen; an operational view lets a marketer diagnose without asking anyone. They are different artefacts and conflating them produces something nobody reads.
Each of these is common, none is malicious, and all four erode trust in marketing numbers faster than any single wrong figure would.
If reporting is being used to steer performance rather than to describe it, the optimization page covers what to do with the numbers once they are trustworthy.
Leading with open rate — Mail clients pre-fetch images on the recipient's behalf, which means an open can be recorded without a human involved. Open rate still has directional uses, but as a headline metric it moves for reasons that have nothing to do with the campaign — and everyone in the meeting now knows that.
Reporting averages that hide the shape — A stable average engagement rate can conceal an active segment shrinking while a dormant one grows. Distribution is the useful view: what proportion of the list is genuinely active, and which way that proportion is moving.
Rebuilding the monthly report by hand — Someone exports, pastes, pivots and formats for two days every month. It is not only the cost — a hand-built report has no audit trail, so when a number looks wrong nobody can reconstruct how it was produced.
Measuring journeys by message metrics — A journey has a purpose, and Journey Builder lets you configure a goal that represents it. A journey reported on click rates with no goal configured is being measured on whether people engaged with it, not on whether it worked.
The symptom is rarely a wrong number. It is that the number stops being used — decisions get made on instinct while the dashboard is politely ignored, and nobody says so directly.
What to do with the numbers once they are trustworthy — and why deliverability comes before everything else.
Write-back to CRM, and the cadence that decides how stale a report can be.
Tracking extracts and file transfers: the standard route for getting engagement data out.
When the reporting question is really an identity question spanning more than marketing.
The practice page: architecture, edition selection and how a larger engagement runs.
A worked example of the architecture-first approach on a real client engagement.
Because they are counting different populations, attributing over different windows, receiving engagement data on a delay, and applying definitions nobody wrote down. Each system is internally consistent; the two are simply not answering the same question.
The fix is not technical. It is agreeing one definition per metric, stating the attribution model explicitly, aligning the write-back cadence to the reporting cycle, and then documenting the residual variance so it is expected rather than disputed.
For anything repeatable, a tracking extract in Automation Studio, transferred to a warehouse, an SFTP destination or a BI tool on a schedule. Data views answer questions inside the platform; the APIs retrieve data on demand for another system.
Which one you need depends on who is consuming it and how fresh it has to be. What matters most is that it leaves at all — engagement history that is never exported quietly stops being available once retention expires.
It earns its place when you have meaningful spend across several channels that has to be reported as one number, and when the data feeding it is already consistent. Intelligence harmonises and visualises; it does not repair inconsistent definitions upstream.
If the actual problem is that email reporting is untrustworthy, Intelligence will present that untrustworthy data more attractively. Fix the definitions and the extraction first, then decide.
Directionally, at most, and never as a headline. Mail clients pre-fetch images on the recipient's behalf, so an open can be recorded without a person having read anything — which means the metric moves for reasons unrelated to the campaign.
We report delivery and placement as the health metrics, click and downstream action as the engagement metrics, and goal completion as the journey metric. Open rate is kept where it is genuinely useful, such as comparing two sends to the same audience on the same day, and dropped everywhere else.
Yes, where the integration supports it — which means engagement data reaching CRM as activity, campaign association configured deliberately, and an attribution model everyone has agreed to before the first report is built.
The honest caveat is that it will be an attributed number rather than a proven one, and that any partner presenting marketing contribution as causal is overselling. Stated as attribution under a named model, it is exactly the number a budget conversation needs.
Retention inside the platform is finite and depends on your configuration, which is precisely why the extract layer matters. Anything you expect to analyze as a multi-year trend needs to live outside the platform.
This is also the most common irreversible mistake we find during a migration: the reporting baseline was never exported, and by the time anybody wants it the source is gone.
Tell us which number is currently disputed and who disputes it. That usually identifies the definitional gap faster than any review of the reporting itself.
Serving: US | Canada | UK | UAE | Australia | New Zealand
