Someone always chose the frame first
Every chart you didn’t build yourself came with a decision already baked in. Someone picked the date range, the axis scale, the metric that got highlighted, and the metric that got left out. By the time it reaches you, the chart isn’t raw data anymore. It’s an argument, dressed up to look like a fact.
That’s not necessarily dishonest. A payment processor showing you a “conversion uplift” chart isn’t lying, it’s selling. An ad platform showing “attributed revenue” isn’t lying either, it’s just counting things the way that makes its own product look useful. The bias isn’t always malicious. But if you read the chart the way it was designed to be read, you’ll walk away believing whatever the person who made it wanted you to believe, even if every number on it is technically accurate.
Running my own businesses off these numbers has taught me a specific skill: separating what a chart actually shows from what it’s trying to make me feel. It’s not about distrust. It’s about reading the picture, not the caption.
Start with the axes before the line
The fastest tell in any chart is the y-axis. If it doesn’t start at zero, a small change looks dramatic. A churn chart that moves from 4.8% to 5.3% over three months can look like a cliff if the axis only spans 4.5% to 5.5%. The same data plotted from 0% to 100% barely wiggles.
Neither version is wrong. They’re both accurate representations of the same numbers. But one makes you want to email your team in a panic and the other makes you shrug and keep working. Before you react to the shape of a line, check what the axis is actually measuring and where it starts. This takes five seconds and it’s the single highest-leverage habit in this whole article.
Same goes for the x-axis. “Last 30 days” and “last 90 days” can tell completely different stories from the same underlying trend, especially with anything seasonal like conversion rate on an ecommerce store around a holiday, or signups for a tool that gets used more at the start of a quarter.
Ask what’s missing from the frame
A chart can only show what it was built to show. A funnel chart from an ad platform showing you clicks to landing page to “conversion” is showing you their funnel, not your business’s funnel. It stops counting the moment their pixel fires. It has no idea if that conversion later refunded, chargebacks, or turned into a customer who churned in month two. If you’re looking at a chart built by a tool that only sees one slice of the customer journey, remember that the slice it’s missing is invisible to you unless you go looking for it elsewhere.
This is especially true with attribution charts. A last-click attribution chart will always make the channel closest to the sale look strongest, because that’s literally what it’s built to measure. It’s not wrong, it’s just answering a narrower question than “what caused this sale” even though it’s often presented as if it answers that broader question. If a vendor’s dashboard is showing you attributed revenue with no mention of the attribution model, that’s a gap worth noticing, not filling in with the assumption that’s most flattering to the chart.
Watch for averages standing in for the real shape
A lot of charts you’ll get handed, especially from software vendors reporting on your own account, show an average: average order value, average session length, average customer lifetime value. Averages are useful, but they smooth over the exact detail that usually matters for a decision.
Say a chart shows your average CLV climbing steadily. That’s a genuinely useful signal. But an average can climb because your typical customer is getting more valuable, or it can climb because a handful of large customers pulled the number up while most of your customers stayed flat or declined. Those are two very different businesses, and only a cohort or distribution view, not a single average line, tells you which one you’re actually in. If a chart only gives you the average and nothing about the spread underneath it, treat the average as a starting question, not a finished answer.
Correlation charts get dressed up as causation
This is the one I catch myself falling for the most, because it’s subtle. A chart that overlays two lines, say ad spend and revenue, climbing together over the same months, is visually persuasive. It looks like proof. It isn’t. It’s two things that moved in the same direction over the same period, which could mean the ad spend caused the revenue, or it could mean both were riding the same seasonal wave, or a new product launch happened in the middle of that window and did the real work.
A/B testing exists specifically because eyeballing two lines on a chart isn’t enough to separate cause from coincidence. If a chart is presenting a correlation as if it settles the question of what worked, that’s worth flagging even if you can’t immediately prove the alternative explanation. The honest move is to notice you don’t yet know, not to accept the story the overlay is telling you.
Check who benefits from you believing the chart
This isn’t cynicism, it’s just useful context. When a chart comes from inside your own business, from your own analytics setup that you configured and trust the plumbing of, you can usually take it closer to face value, though you should still sanity check the axis and the window. When a chart comes from someone with something to sell you, a vendor benchmark deck, a case study slide, an ad platform’s own performance report, ask a simple question before you internalize it: what does this chart want me to do next? Renew? Spend more? Feel behind competitors I’ve never actually seen the numbers for? That question alone often reveals which parts of the chart are the data and which parts are the pitch.
This applies to competitor and industry benchmark charts too. A “the average business in your category converts at X” chart is only as good as the businesses that got pooled into “average,” and you’re rarely told how that pool was built, how big it was, or how similar those businesses actually are to yours. Treat benchmark charts as a loose reference point, not a target you’re failing to hit.
A short checklist before you act on someone else’s chart
I run through something like this before I let any outside chart change what I do:
- Does the y-axis start at zero, and if not, how much is that stretching the visual difference
- What’s the actual time window, and is it long enough to rule out a seasonal blip
- Is this showing an average, and if so, do I know anything about the spread underneath it
- Does this chart’s own tracking stop somewhere short of the outcome I actually care about
- Is there a second explanation for this trend that the chart doesn’t rule out
- Who built this chart, and what do they want me to do after I look at it
None of these questions require distrust of the person who made the chart. Most of the time they built it in good faith, using the data they had, for the purpose they had. The point of the checklist isn’t to catch someone lying. It’s to make sure you’re answering your own question with their chart, instead of quietly answering theirs.
The chart is a starting point, not a verdict
None of this means charts are useless or that you should second-guess every dashboard you open. It means a chart is one person’s compressed summary of a much messier reality, and compression always throws something away. Reading it critically just means noticing what got thrown away before you build a decision on top of it. That takes practice, and it gets easier the more you’ve had to build your own charts and watch how many small choices go into making one number look bigger or smaller than another.
If you want more of this kind of plain-language walkthrough on the metrics and methods that actually run a small business, from cohort retention to funnel math to how AI tools fit into a one-person operation, you can find the rest of the collection here.
Get new guides and videos first — join the Telegram channel.