Marketing Analytics: Metrics and Dashboards That Matter

Marketing generates more numbers than almost any other business function: impressions, clicks, sessions, followers, open rates, conversions, costs. Most are useful for someone, but only a few tell a marketing leader whether money is being spent well. Marketing analytics is about finding those few, defining them precisely, and comparing channels by what they deliver rather than by how much activity they create.

The metrics, defined precisely

Vague definitions cause most marketing reporting arguments, so it helps to use the definitions of the platforms that report them:

Beyond these, most businesses also track customer acquisition cost across all marketing spend, and customer lifetime value, the revenue a customer brings over the relationship. Comparing the two answers the question behind all the others: is marketing buying customers worth more than they cost?

A metric hierarchy

The metrics form a chain from attention to money:

1. Reach and attention: impressions, CTR.
2. Engagement: sessions, engagement rate.
3. Action: conversions, conversion rate.
4. Efficiency: CPA, ROAS.
5. Value: revenue, lifetime value, retention.

Leadership decisions depend on levels 4 and 5. Levels 1 to 3 explain why those numbers moved. A dashboard that leads with impressions and followers, and buries cost and value, is optimised for looking busy; vanity vs actionable metrics explains how to spot that.

Comparing channels by what they deliver

The Online Shoppers Purchasing Intention dataset records 12,330 shop sessions with an anonymised traffic source and whether each ended in a purchase (UCI: Online Shoppers Purchasing Intention). The five largest sources:

import io
import urllib.request
import zipfile
import pandas as pd

url = "https://archive.ics.uci.edu/static/public/468/online+shoppers+purchasing+intention+dataset.zip"
sessions = pd.read_csv(zipfile.ZipFile(io.BytesIO(urllib.request.urlopen(url).read())).open("online_shoppers_intention.csv"))
by_source = sessions.groupby("TrafficType").agg(sessions=("Revenue", "size"), purchases=("Revenue", "sum"))
by_source["share of sessions"] = by_source["sessions"] / by_source["sessions"].sum()
by_source["conversion rate"] = by_source["purchases"] / by_source["sessions"]
by_source["share of purchases"] = by_source["purchases"] / by_source["purchases"].sum()
print(by_source.nlargest(5, "sessions").round(3).to_string())
print("overall conversion rate:", round(sessions["Revenue"].mean(), 3))

             sessions  purchases  share of sessions  conversion rate  share of purchases
TrafficType                                                                             
2                3913        847              0.317            0.216               0.444
1                2451        262              0.199            0.107               0.137
3                2052        180              0.166            0.088               0.094
4                1069        165              0.087            0.154               0.086
13                738         43              0.060            0.058               0.023

overall conversion rate: 0.155

Source 2 brings 32% of sessions but 44% of purchases, converting at 21.6%. Sources 1 and 3 together bring over a third of sessions but under a quarter of purchases, and source 13 converts at under 6%. A report ranking channels by traffic would put sources 1 and 3 near the top; a report ranking by purchases per session tells a different story. Add each channel's cost, and CPA and ROAS follow directly: the question becomes whether source 13's cheap traffic is still worth buying at a 6% conversion rate.

Accounting for timing

Channel comparisons also need to account for when traffic arrives:

by_month = sessions.groupby("Month")["Revenue"].agg(["size", "mean"]).rename(columns={"size": "sessions", "mean": "conversion rate"})
print(by_month.sort_values("conversion rate", ascending=False).round(3).head(4).to_string())

       sessions  conversion rate
Month                           
Nov        2998            0.254
Oct         549            0.209
Sep         448            0.192
Aug         433            0.176

November sessions convert at 25.4%, well above the 15.5% average. A channel that runs mostly in November will look better than one that runs mostly in spring, even if both are equally good. Compare channels within the same period, or report seasonally adjusted figures. The same caution applies to before-and-after comparisons of a campaign: a launch in October will look like a success partly because October and November convert better anyway.

A marketing dashboard layout

• Top row: spend, conversions, CPA and ROAS, each against target and previous period.
• Middle: a channel table with sessions, conversion rate, cost, CPA and ROAS, sorted by spend so the biggest decisions are at the top.
• Bottom: a monthly trend of conversions and spend, and the conversion funnel; funnel analysis covers building it.
• Footnote: the attribution model used, and which conversions are counted.

Common mistakes in marketing reporting

• Mixing definitions across platforms. An ad platform, a web analytics tool and the order database each count conversions differently. Put them side by side only with the differences explained.
• Comparing channels over different periods, which rewards channels active in strong months, as the November figures show.
• Averaging rates across channels instead of recomputing them from totals. A blended conversion rate must be total purchases divided by total sessions, not the average of each channel's rate.
• Reporting spend without outcomes, or outcomes without spend. Efficiency needs both.
• Ignoring small samples. A channel with 50 sessions and five purchases has a 10% conversion rate that could easily be 5% or 15% next month.

Marketing KPIs work best as a small tree with value at the top and drivers underneath; how to create effective KPIs shows how to build one.

Attribution: the trap behind channel numbers

Most customers interact with several channels before buying. Which one gets the credit depends on the attribution model: last click, first click or a model that shares credit. Channel CPA and ROAS change, sometimes dramatically, when the model changes. State the model, keep it consistent over time, and treat big budget shifts based on attributed numbers with care. Where possible, test channel value directly by varying spend in comparable regions or periods and measuring the overall effect.

Marketing campaign analytics is one of the projects in the Data Analytics course.

Sources

• Clickthrough rate (CTR): Definition (Google Ads Help) — Google Ads Help
• Conversion rate: Definition (Google Ads Help) — Google Ads Help
• About Target CPA bidding (Google Ads Help) — Google Ads Help
• About Target ROAS bidding (Google Ads Help) — Google Ads Help
• GA4 engagement rate and bounce rate (Analytics Help) — Google Analytics Help
• Online Shoppers Purchasing Intention Dataset (UCI Machine Learning Repository) — UCI Machine Learning Repository