There is a version of marketing reporting that looks rigorous and isn't. The numbers are technically correct. The slide deck is well-designed. The executive summary is full of upward arrows. And somewhere in the language, if you know what to listen for, are the signs that what you're being told is a story rather than the truth.

Misleading marketing reports rarely involve outright fabrication. The more common problem is framing: presenting real numbers in ways that obscure what they actually mean, selecting metrics that flatter rather than inform, and using the language of rigour to make weak results harder to challenge. For e-commerce teams making commercial decisions on the back of this reporting, that gap between what the numbers say and what they mean is where the value is lost.

The phrases that signal this are consistent enough to be worth learning. Because the moment you can hear the difference between a result that has been interrogated and one that has been packaged, the reporting changes.

Why Misleading Marketing Reports Are So Hard to Spot

The challenge with misleading marketing reports is that the numbers in them are usually real. 

An agency reporting a 400% ROAS isn't lying about the revenue figure, they're lying about what caused it, or what it means for your actual commercial position. 

An AB testing report stating statistical significance at 95% confidence is telling you something real, but without adequate context to validate this is means anything beyond “we got lucky”.

This is what makes the language of digital reporting so effective as a defence mechanism. It borrows the authority of data without submitting to the discipline of analysis. The result is a version of commercial reality that encourages continued investment rather than honest evaluation, a version, in other words, that has been produced by people whose objective is your spend, not your success.

Platform bias operates the same way. Meta reports Meta's version of your results. Google reports Google's. Each platform is reporting it’s own success, applying its own attribution logic, and producing a number that reflects well on continued investment in that channel. When you receive reporting that draws directly from platform dashboards without independent verification, you are not receiving an analysis. 

You are receiving a pitch built from someone else's data.

The Phrases That Signal a Misleading Marketing Report

Certain phrases appear consistently in reporting that is technically defensible but commercially misleading. They are worth knowing not because they prove bad intent, most are used habitually rather than deliberately, but because each one signals a specific type of gap between the number presented and the commercial reality it is supposed to represent.

"Extrapolated revenue impact" is one of the most common. 

It appears in AB testing reports when the measured uplift is too small to justify the investment on its own, so the result is projected forward, often annualised, to produce a number worth presenting. The extrapolation may be mathematically sound. 

What it doesn't tell you is whether the test was adequately powered to detect a real effect, whether the result holds consistently across the full test period, or whether the uplift would sustain over the projection window. A 0.4% conversion uplift extrapolated to £800,000 annual revenue impact is a headline. Whether the 0.4% was real is a different question entirely.

"Statistically significant" is used to close down scrutiny rather than invite it. 

In practice, statistical significance at 95% confidence means one specific thing: if there were genuinely no difference between your control and variant, you would see a result this large by chance only 5% of the time. It does not mean the result is real, stable, commercially meaningful, or safe to act on. 

A test that reached significance because a single high-value transaction fell in the variant group is technically significant. It is also misleading. When significance is presented without any discussion of whether the result was consistent, whether it was called at the right time, or whether the effect size was meaningful, that is a signal to ask harder questions, not to stop asking them.

"Average order value impact" appears in reporting when conversion rate didn't move. 

AOV is a legitimate commercial metric, but it is also one of the most volatile, disproportionately sensitive to a small number of high-value transactions in a way that conversion rate is not. A single order at three times your typical basket size can manufacture a meaningful-looking AOV uplift with no underlying change in customer behaviour. 

When AOV is the headline metric in a test result and conversion rate is absent or buried, that alone is worth interrogating.

"Increased reach" or "impressions" as a primary success metric for paid advertising is one of the oldest deflections in digital marketing. 

Reach tells you how many people could theoretically have seen an ad. It says nothing about whether they engaged with it, visited your site, or bought anything. For an e-commerce team spending real budget on paid media, the question is not whether the ad was seen. It is whether the investment drove sales at a return that justifies the cost. 

Reach, presented without click data, without traffic quality analysis, and without revenue attribution, is a number that fills a slide without answering the question.

"The ROI of your investment is X" is the phrase that most confidently collapses several distinct problems into a single misleading figure. 

Platform-reported ROI compares an estimate of revenue uplift, calculated using the platform's own attribution model, against the media spend. It does not account for the overlap with other channels. It does not account for organic sales that would have happened anyway. It does not account for profitability, margin, or the cost of goods sold. 

A ROAS of 400% looks strong in a report. Whether it represents a profitable investment depends on information the platform's reporting is not designed to surface.

What Honest Marketing Reporting Actually Looks Like

The difference between misleading marketing reports and honest ones is not one of complexity. 

It is a difference of intent. 

Reliable reporting is built to disprove the result before it declares it. It applies independent verification, acknowledges the limitations of the methodology, and presents the confidence it has in a finding alongside the caveats that qualify it.

Honest AB test reporting sounds different. 

It states the observed uplift, the statistical significance, and the statistical power, the probability that the test was capable of detecting a real effect at the claimed magnitude. 

It accounts for high-value transaction anomalies, reports consistency of impact across the test window, and distinguishes between results that held throughout the test and those manufactured by two or three exceptional days. 

A result that passes all of those checks and still shows a 2% conversion uplift, even at 91% confidence is one worth acting on. 

A result that reports significance without any of those checks may be telling you something. It is harder to know what.

Honest paid media reporting applies a consistent attribution model across channels, rather than drawing from each platform's native dashboard. 

Last-touch attribution applied uniformly across paid search, paid social, and email produces a comparable picture across channels, not an accurate one, attribution is never perfectly accurate, but a consistent one that supports real commercial decision-making. 

Channel-level ROAS calculated on the same basis, reconciled against actual Shopify revenue rather than platform-reported revenue, is a single version of the truth. 

How Platform Bias Distorts the Commercial Picture for E-Commerce Teams

Misleading marketing reports are rarely produced by agencies acting in bad faith. 

The deeper problem is structural: the platforms that most e-commerce teams rely on for reporting have a commercial interest in the numbers they produce. Meta's attribution model credits Meta. Google's attribution model credits Google. Each platform's reporting is designed to justify continued investment in that platform, and when an e-commerce team accepts those reports as its primary view of commercial reality, it has handed ownership of the truth to the parties with the most to gain from distorting it.

This is not a conspiracy. It is an incentive structure, and it operates at the level of default settings, recommended attribution windows, and the metrics each platform chooses to surface by default. 

A 7-day view attribution window on Meta inflates the revenue credited to paid social. A target ROAS bidding strategy optimises for the metric Meta can influence most directly, platform-reported return, rather than the metric that actually matters, which is your bottom line after cost of goods, fulfilment, and returns.

The e-commerce teams that outperform their competitors are not the ones that trust their reports and dashboards most. 

They are the ones that have built an independent view of their commercial reality, one that brings all of their data together under one framework, with one objective: the truth, even when it's uncomfortable. That is not the same as having better reporting software. It is to have a different relationship with the data entirely.

How to Read a Marketing Report With Healthy Scepticism

The practical test for any piece of marketing reporting is a simple one: what has this analysis done to try to disprove its own conclusion? 

Next time you are presented with a QBR report, an experimentation update or an ask for more investment based on the level of success seen so far, a few questions you might want to have in your mind are below.

For AB test results:

Was the test adequately powered to detect the claimed effect? 

Was the result declared at the planned duration, or early? 

Does the daily performance show a consistent pattern, or was the aggregate result manufactured by a handful of exceptional days? 

Were high-value transactions checked for distortion? 

Does the same result hold when you look at conversion rate independently of revenue?

For paid media reporting:

Is the revenue figure drawn from the platform's attribution model or reconciled against Shopify? 

What attribution window is being used, and is it consistent across channels? 

What is the channel's contribution when you remove brand search and direct traffic, the customers who were going to buy anyway? 

Does the ROAS figure account for margin, or is it calculated on revenue gross of cost?

None of these questions are unfair. A reporting partner who finds them threatening is a reporting partner whose numbers cannot survive scrutiny. The ones worth working with are the ones who have already asked those questions before presenting the result, and who present the answer alongside the headline.

The Commercial Cost of Trusting Misleading Marketing Reports

The cost of misleading marketing reports is not visible test by test or campaign by campaign. 

It compounds. 

An e-commerce team that consistently acts on flattering but unreliable data makes investment decisions that look rational at the time and produce results that never quite match the reporting. The paid channel that was generating 400% ROAS turns out, when independently verified, to be running at 160%. The AB test that supposedly lifted conversion by 3% delivers 0.8% in production. The reach campaign that generated impressive impression numbers delivered no measurable revenue at all.

Over time, this dynamic and confident reporting yet disappointing reality, erodes the team's ability to make effective commercial decisions, because the data that should inform those decisions has been consistently wrong. The teams that find it hardest to grow are not usually the ones with the worst products or the smallest budgets. They are the ones whose commercial picture has been owned by platforms and partners whose objective was their investment, not their success.

Reclaiming that picture starts with a simple discipline: before you act on a result, ask what the analysis has done to try to prove itself wrong. If the answer is nothing, the result is not an insight. It is a pitch. 

To put it simply… being told you're making money doesn't mean you're making money.