How Data Analysts Add Real Business Value

The description of analyst work in occupational databases centres on reports: summarising business data for stakeholders and maintaining the dashboards that deliver it (O*NET: business intelligence analysts). Reports are necessary, but they are not where most value comes from. Value comes when an analysis changes a decision: a budget moved, a problem fixed, a risk avoided, a product changed. Analysts who understand that do different work, and are treated differently, from analysts who see their job as filling requests.

1. Frame the decision behind the request

"Can you pull sales by region?" is rarely the real question. Behind it is a decision: where to add a salesperson, which region's targets to revise, whether a campaign worked. Ask what the answer will be used for, what would change the decision, and by when it is needed. Microsoft's introduction to the analytics process makes the same point from the other direction: when a business question is not getting the right answer, the problem is usually the approach (Microsoft Learn: the analytics process). Reframing a request often turns a table nobody reads into an answer somebody acts on.

Three examples of reframing in practice:

• "Send me last quarter's churn numbers" becomes "which customers are most likely to leave next quarter, and is it worth contacting them?", a question about where to spend the retention budget.
• "Build a dashboard of website traffic" becomes "which traffic sources bring customers who actually buy?", a question about where to spend on marketing.
• "How many support tickets did we get?" becomes "which product issues generate the most tickets, and what would fixing them save?", a question about what to fix first.

In each case the data work is similar, but the output answers the decision instead of describing the data.

2. Size opportunities before anyone invests

Managers constantly choose between ideas with limited time and money. An analyst who can say roughly what each idea is worth makes those choices far better. Sizing is simple arithmetic on real data plus explicit assumptions. An example on a UK online gift retailer's transactions (UCI: Online Retail): how much would a win-back campaign for lapsed regular customers be worth?

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

url = "https://archive.ics.uci.edu/static/public/352/online+retail.zip"
with zipfile.ZipFile(io.BytesIO(urllib.request.urlopen(url).read())) as archive:
    raw = pd.read_excel(archive.open("Online Retail.xlsx"), dtype={"InvoiceNo": str, "StockCode": str})
services = {"POST", "DOT", "M", "m", "C2", "D", "S", "B", "BANK CHARGES", "AMAZONFEE", "CRUK"}
sales = raw[~raw["InvoiceNo"].str.startswith("C") & ~raw["StockCode"].isin(services)
            & ~raw["StockCode"].str.startswith("gift_") & (raw["Quantity"] > 0) & (raw["UnitPrice"] > 0)
            & raw["CustomerID"].notna() & (raw["InvoiceDate"] < "2011-12-01")]
sales = sales.assign(revenue=sales["Quantity"] * sales["UnitPrice"])
customers = sales.groupby("CustomerID").agg(last=("InvoiceDate", "max"), orders=("InvoiceNo", "nunique"),
                                            revenue=("revenue", "sum"), first=("InvoiceDate", "min"))
days_quiet = (pd.Timestamp("2011-12-01") - customers["last"]).dt.days
lapsed_regulars = customers[(customers["orders"] >= 3) & (days_quiet > 90)]
months_active = ((lapsed_regulars["last"] - lapsed_regulars["first"]).dt.days / 30.4).clip(lower=1)
monthly_value = (lapsed_regulars["revenue"] / months_active).median()
print(f"regular customers (3+ orders) silent for over 90 days: {len(lapsed_regulars)}")
print(f"median revenue per active month while they were buying: {monthly_value:,.0f}")

regular customers (3+ orders) silent for over 90 days: 270
median revenue per active month while they were buying: 223

Then the sizing, under three stated assumptions about how many would come back:

for win_back_rate in (0.05, 0.10, 0.20):
    recovered = len(lapsed_regulars) * win_back_rate * monthly_value * 12
    print(f"if {win_back_rate:.0%} return for a year: about {recovered:,.0f} in revenue")

if 5% return for a year: about 36,091 in revenue
if 10% return for a year: about 72,183 in revenue
if 20% return for a year: about 144,366 in revenue

The answer is a range, roughly 36,000 to 144,000 a year, not a single confident number, and that is the point. It tells a manager whether a campaign costing, say, 10,000 is worth trying, and which assumption matters most: the win-back rate. The honest next step is a small test on a random half of these customers to measure that rate, rather than arguing about it.

3. Measure whether changes worked

Businesses change prices, launch features and run campaigns constantly, and often never find out what happened. Analysts add value by designing the measurement before the change: the metric, the comparison group, the time window, and the size of effect that would matter. Where possible, that means a randomised test; where not, a careful before-and-after comparison with a similar control group. Without this, organisations repeat what feels successful rather than what works.

4. Find leaks, errors and risks

Some of the most valuable analyses are the least glamorous: finding invoices that were never billed, fees charged twice, discounts applied outside policy, stock recorded but not present, or a tracking error that has understated sales for months. These often have clear, immediate financial value, and they come from analysts who reconcile numbers carefully instead of accepting them.

5. Save people time

A report that three people rebuild by hand every Monday is a cost the organisation pays every week. Automating it, with a scheduled query and a dashboard, returns those hours to more valuable work, and removes a source of errors. Build self-service tools for recurring questions, so that analyst time goes to new questions instead of repeated ones.

Showing your impact

• Keep an impact log: the question, the decision it informed, and the result, with numbers where you have them.
• Lead with the decision and the money when presenting: "The win-back test could be worth 36,000 to 144,000 a year; a four-week test would tell us which end of that range is real."
• State assumptions and ranges openly; credibility comes from being right about what you do not know.
• Follow up after decisions are made to see what happened, and share the results.

What undermines an analyst's value

• Vanity reporting: dashboards of numbers that nobody uses to decide anything. Vanity vs actionable metrics covers how to tell them apart.
• False precision: reporting 72,183 as if it were certain, instead of a range.
• Answering exactly what was asked when the question was wrong.
• Being slow for the decision: a perfect analysis delivered after the decision has been made has no value.

Good KPIs are the foundation for much of this work; how to create effective KPIs covers them, and KPI scorecards for executives shows how to report them.

KPI analysis, sales and revenue analysis, and business reporting are part of the business analytics concepts module of the Data Analytics course.

Sources

• Business Intelligence Analysts (O*NET OnLine 15-2051.01) — O*NET OnLine, US Department of Labor
• Explore the analytics process that turns data into insights (Microsoft Learn) — Microsoft Learn
• Online Retail (UCI Machine Learning Repository) — UCI Machine Learning Repository