How to Write an Analytics Insight (Not Just a Number)
By Emily Redmond, Data Analyst at Emilytics · April 2026
TL;DR: An insight has four parts: the number, context (comparison), hypothesis (why), and implication (what we do). Write all four. Everything else is noise.
The Difference Between Data and Insight
Data: "Organic traffic is 8,420 sessions this month."
Insight: "Organic traffic is 8,420 sessions this month, up 16% from March. The increase correlates with our new long-form content strategy—we published four guides targeting high-intent keywords. This suggests our audience prefers detailed content over short-form articles. We should keep investing in this format."
The second one is an insight. It has context, explanation, and implication. The first is just noise.
The Insight Framework
Every insight has four components. Writing the same piece of data four different ways:
Component 1: The Number (Observation)
"Conversion rate is 3.2%"
Just the fact. No interpretation.
Component 2: Context (Comparison)
"Conversion rate is 3.2%, up 0.4 points from last month and above our 3% target."
Now it has context. Is this good or bad? You know.
Component 3: Hypothesis (Causation)
"Conversion rate is 3.2%, up 0.4 points from last month. This likely correlates with our redesigned checkout flow, which we launched mid-March. Mobile users experienced the biggest improvement (2.1% to 2.8%), suggesting the mobile-first checkout was the main driver."
Now we have causation. Why did it move? We have an educated guess.
Component 4: Implication (Action)
"Conversion rate is 3.2%, up 0.4 points from last month, likely due to our redesigned checkout flow. Mobile conversion improved 33%, desktop 8%. This validates our hypothesis that mobile was our conversion bottleneck. We should continue investing in mobile experience before optimizing desktop (which is already performing above benchmark)."
Now it's actionable. What do we do about it? We have a direction.
The Insight Formula
[Number] [Comparison] [Hypothesis] [Action]
Example:
"Email signups are 320 this month (↓ 12% from last month, 18% below target). This drop correlates with our reduced publishing cadence (4 articles vs. 6 last month). Lower page traffic means lower signup opportunity. We should either increase publishing cadence back to 6 articles weekly or test different CTA placement on lower-traffic pages."
How to Write Insights at Different Levels
Junior Level Insight
"Traffic is up 12%."
Missing: context, hypothesis, implication. This is data, not insight.
Intermediate Level Insight
"Traffic is up 12% MoM (to 8,420 sessions) and up 18% YoY. This is above our 10% growth target."
Better. Added comparison. Still missing: why did it move? What should we do?
Senior Level Insight
"Traffic is up 12% MoM (8,420 sessions, 18% YoY). Organic driven this growth (traffic up 18%), while paid remained flat. This organic lift correlates with our new keyword strategy—we expanded to 120 target keywords (from 80). This suggests we're not hitting saturation yet on our target keywords and should continue keyword expansion. However, position on existing keywords slipped 0.3 points, so we may also want to check content freshness and backlink quality."
Expert-level. Full picture: the number, comparison, multiple hypotheses, and specific next steps.
Common Mistakes in Insight Writing
Mistake 1: Correlation = Causation
"Traffic spiked 30% on March 15. We launched a campaign on March 15. The campaign caused the spike."
Maybe. Could be Google algorithm update. Could be competitor going down. Could be seasonality. Use evidence: "70% of the traffic spike came from our new keyword targets, and search position improved 0.5 points on 12 of our target keywords. This suggests the campaign was the primary driver."
Mistake 2: No comparison
"Conversion rate is 3.2%."
Compared to what? Last month? Last year? Your goal? Add context: "Conversion rate is 3.2% (up from 2.8% last month, above our 3% target)."
Mistake 3: Vague language
"Revenue is doing well." "Traffic looks good." "Engagement seems up."
"Well" is not a metric. Use numbers: "Revenue is $45,200, up 12% MoM and 8% above our monthly target."
Mistake 4: Forgetting the implication
"Churn is up 2 points to 12%."
This is bad. But what do we do? "Churn is up 2 points to 12%. Highest churn is in the first 30 days (onboarding cohort). We should audit our onboarding flow and test improvements before the end of Q2."
Now it's actionable.
💡 Emily's take: I edited insights for years. The most common problem was analysts reporting numbers without implications. They'd say "mobile conversion dropped 8%," then move on. I'd ask "so what?" They'd stare at me like I'd asked a weird question. But "so what" is the analyst's job. Report the number, but answer "so what" or nobody will act on it.
Insight Writing by Report Type
Weekly Report Insights (1–2 sentences)
"Paid CAC increased 8% to $45 this week. This correlates with iOS 15 tracking changes limiting our audience targeting. We're testing new audience segments on Google Ads this week to maintain efficiency."
Monthly Report Insights (3–4 sentences)
"Organic traffic grew 16% to 8,420 sessions, driven by new long-form content (4 guides published, 3 ranked in top 3 position within two weeks). Average time-on-page for guide traffic is 4m 30s, vs. 2m 10s for other content. This suggests our audience prefers in-depth content. We're investing 30% more budget into long-form guides next month to capitalize on this trend."
Quarterly Report Insights (5–6 sentences)
"Q1 revenue grew 22% YoY to $420K, driven by both new customer acquisition and expansion revenue. New customers are up 18% QoQ, with CAC stable at $38 and LTV up 12% due to feature adoption in our Pro tier. Churn remained stable at 2.1%. The biggest opportunity: Enterprise tier customers have 3.2x higher LTV than Pro tier but only represent 8% of base. If we shift 10% of Pro customers to Enterprise, we could increase revenue 18% without increasing CAC. We should build out dedicated Enterprise support starting Q2."
How to Improve Your Insight Writing
1. Always include comparison. Number + benchmark = context. Always.
2. Support hypotheses with evidence. Don't guess. "Traffic spiked likely because X" should be backed by data: "70% of the spike came from our target keywords."
3. End with a direction. Every insight should hint at what comes next.
4. Use plain language. "Our new feature is driving higher engagement" not "the new feature increased the engagement velocity metric."
5. Keep it short. One paragraph per insight. Two max.
Frequently Asked Questions
Q: How confident do I need to be in a hypothesis to write it?
A: 70%+ confidence is reasonable for reporting. "Likely due to" or "probably caused by" is honest language. "Definitely caused by" only if you have proof. Readers appreciate intellectual honesty.
Q: Should I include negative findings in insights?
A: Yes. "Traffic dropped 5% due to outage Wednesday, recovered Thursday. Net impact: 400 lost sessions." Being transparent builds credibility.
Q: How do I write an insight when the data is surprising or contradictory?
A: Flag it. "Email conversion is up 15% this month, but signup volume is down 8%. This suggests email is reaching higher-intent audiences, but overall list quality may be declining. We should analyze email list segment performance before scaling sends."
Q: Should I always have a hypothesis for every number?
A: No. "Sessions: 8,420" might not warrant investigation if it's normal. Save insights for the numbers that matter—the ones that drive business outcomes.
The Bottom Line
Insight writing is the skill that separates analysts from analysts who get promoted. Anyone can read a number. Analysts who translate numbers into implications move businesses.
Start small: add one comparison and one implication to every number you report. That's the minimum. Build from there.
For more on reporting, see weekly analytics reports or analytics writing for executives.
Emily Redmond is a data analyst at Emilytics — AI analytics agent watching your GA4, Search Console, and Bing data around the clock. 8 years experience. Say hi →