Analytics Across Business Functions

Introduction

Business analytics creates value when evidence is connected to a business decision. Although analytical techniques are often taught as general methods, their meaning depends strongly on the function in which they are applied. A marketing analyst may estimate campaign response, a finance analyst may assess profitability and liquidity, an operations analyst may improve capacity utilisation, and a human-resources analyst may investigate employee turnover. The same data principle—comparing observed outcomes with an appropriate benchmark—can therefore support very different decisions.

This chapter examines the application of analytics across marketing, sales, finance, operations, supply chains, customer service and human resources. It focuses on practical metrics, descriptive and predictive models, decision rules, interpretation and governance. The central theme is that a useful analysis must connect four elements:

1. a clearly defined business objective;
2. reliable and relevant data;
3. an appropriate analytical method; and
4. an action whose result can be evaluated.

Analytics should not be treated as a substitute for managerial judgement. It improves judgement by making assumptions visible, identifying patterns and quantifying uncertainty.

Learning Objectives

By the end of this chapter, you should be able to:

• distinguish descriptive, diagnostic, predictive and prescriptive analytics in functional settings;
• select suitable metrics for major business functions;
• apply basic formulas for marketing, sales, finance, operations, supply chain, customer service and HR;
• interpret segmentation, forecasting, classification, experimentation and optimisation results;
• recognise data-quality, ethical, privacy and governance risks; and
• recommend an analytics-supported action and an appropriate performance-monitoring plan.

1. A Functional View of Business Analytics

A business function is a specialised area of activity with its own objectives, processes, data and performance measures. Functional analytics translates organisational goals into measurable questions.

A metric is a defined measure, while a key performance indicator (KPI) is a metric judged to be strategically important. A good KPI has a precise numerator, denominator, time period, population and data source. For example, “customer satisfaction” is vague unless the organisation specifies the survey population, scale, response rule and reporting period.

A useful analytical workflow is:

Business question → data definition → data preparation → analysis or model → decision → measurement of outcome.

This workflow prevents a common failure: producing an accurate model that does not answer a decision-maker’s question.

2. Marketing Analytics

Marketing analytics evaluates how customer needs, communications, channels, prices and campaigns influence behaviour and financial outcomes. It combines customer, transaction, web, advertising and survey data.

2.1 Core marketing metrics

The conversion rate measures the proportion of a defined audience completing a target action:

The denominator must be defined consistently. A website conversion rate and a sales-lead conversion rate are not interchangeable.

Customer acquisition cost (CAC) estimates the cost of acquiring a new customer:

CAC may include advertising, agency fees, sales salaries and campaign technology, depending on the organisation’s accounting policy. It should be compared with customer lifetime value (CLV), not merely with first-purchase revenue.

A simplified contribution-based CLV for a customer with constant periodic contribution is:

where m is contribution margin per period, r is the retention probability per period, and i is the discount rate per period. This expression assumes constant margin, constant retention, a sufficiently long horizon and contributions received at the end of each period. It excludes acquisition cost unless CAC is subtracted separately.

For a finite planning horizon, a more explicit model is:

Here T is the number of periods, P(Active_t) is the probability that the customer remains active in period t, and all monetary values must use the same currency and time basis.

Return on advertising spend (ROAS) is:

ROAS is not the same as profit return. A campaign can have high ROAS but low profitability if product margins are small or fulfilment costs are high.

2.2 Segmentation and experimentation

Segmentation divides customers into groups with meaningful differences in needs, behaviour or value. Common approaches include demographic, geographic, behavioural, psychographic and value-based segmentation. Behavioural segmentation may use recency, frequency and monetary value (RFM):

• Recency: how recently the customer purchased;
• Frequency: how often the customer purchased; and
• Monetary value: how much the customer spent.

Segmentation should support a different action. A cluster that cannot be reached, served or measured differently has limited practical value.

A/B testing compares a treatment group with a control group. If the conversion rates are p₁ and p₀, the absolute lift is:

and the relative lift is:

A valid experiment requires a defined hypothesis, random assignment where possible, a stable outcome measure, adequate sample size and protection against premature stopping. Correlation from observational campaign data does not by itself establish that the campaign caused the outcome.

2.3 Attribution limitations

Attribution assigns credit for a conversion to marketing touchpoints. First-touch, last-touch, linear and position-based rules are simple but can be misleading. A last-touch rule may over-credit a search ad that occurred immediately before purchase while ignoring earlier brand exposure. Incrementality tests, holdout groups and carefully designed experiments provide stronger evidence of causal impact.

3. Sales Analytics

Sales analytics supports territory planning, pipeline management, lead qualification, forecasting and account prioritisation.

Important measures include win rate, average deal size, sales-cycle length, quota attainment and pipeline coverage. Pipeline coverage can be expressed as:

If the historical win rate is w, a rough required pipeline value is:

This estimate assumes that the historical win rate applies to the current pipeline and that deal values are comparable. It is not a guarantee.

A sales funnel identifies movement between stages such as lead, qualified opportunity, proposal and closed sale. Stage conversion rates reveal where prospects are lost. However, stage definitions must be operationally consistent; otherwise, a change in sales staff behaviour may appear to be a change in customer demand.

Forecasting can use historical averages, weighted pipeline values, time-series methods or salesperson judgement. Weighted pipeline forecasting is commonly expressed as:

The probabilities should be calibrated from historical outcomes. A probability copied from a generic CRM template may produce false precision.

Sales models also raise fairness and accountability issues. Lead-scoring systems should be checked for unequal treatment, data leakage and inappropriate use of protected characteristics or their proxies.

4. Finance Analytics

Finance analytics connects operational activity to revenue, costs, cash, risk and value creation. It requires careful distinction between accounting measures and economic measures.

4.1 Profitability and break-even

Contribution margin is sales revenue less variable costs:

Contribution margin per unit is:

For fixed costs F and unit contribution c, the break-even volume is:

This formula assumes a constant selling price and unit variable cost, fixed costs within the relevant range, and a single product or a constant sales mix. If a business sells multiple products, weighted-average contribution or a product-mix model is required.

Worked example. A service sells for $80 per unit, has a variable cost of $50, and incurs fixed costs of $60,000. Unit contribution is $30. Therefore:

Break-even units = $60,000 ÷ $30 = 2,000 units.

At 2,500 units, operating profit before financing and tax is:

($80 − $50) × 2,500 − $60,000 = $15,000.

4.2 Cash and working capital

The cash conversion cycle measures the time between paying suppliers and collecting cash from customers:

Days sales outstanding (DSO), for example, can be estimated as:

The credit-sales denominator and time period must correspond to the receivables balance. A lower cycle is not always automatically better: excessively low inventory may cause stockouts, while delaying supplier payment may damage relationships.

4.3 Financial modelling discipline

Financial dashboards should reconcile to the general ledger where appropriate, distinguish actuals from forecasts and document assumptions. Analysts should perform sensitivity analysis because results may change substantially with price, volume, exchange-rate or cost assumptions. A model should show ranges or scenarios rather than presenting uncertain forecasts as exact facts.

5. Operations Analytics

Operations analytics improves the transformation of inputs into goods or services. It addresses capacity, productivity, quality, waiting time, scheduling and maintenance.

5.1 Capacity and utilisation

Capacity utilisation is:

Design capacity is the theoretical maximum; effective capacity reflects realistic constraints such as maintenance, staffing and product mix. Analysts must state which capacity concept is used.

Productivity compares output with an input:

Output and labour hours must be measured consistently. A productivity increase that results from lower quality or unsafe work is not a genuine operational improvement.

5.2 Quality and process stability

The defect rate is:

First-pass yield measures the proportion that meets requirements without rework. Control charts distinguish common-cause variation from unusual assignable causes. A process should not be adjusted after every random fluctuation; excessive intervention can increase instability.

5.3 Bottlenecks and queues

A bottleneck is a resource whose capacity constrains system output. Improving a non-bottleneck may create work-in-process without increasing throughput. Little’s Law links average work-in-process (L), throughput rate (λ) and average flow time (W):

The units must be compatible—for example, orders, orders per hour and hours. The relationship assumes a stable system over the measurement period. It is descriptive, not a complete queueing model.

6. Supply-Chain Analytics

Supply-chain analytics coordinates suppliers, inventory, transportation, warehouses and customers. Its objective is usually to balance service, cost, resilience and working capital.

6.1 Inventory measures

Inventory turnover is:

Days inventory on hand is approximately:

A higher turnover may indicate efficient inventory use, but it may also indicate insufficient stock and lost sales.

For stable demand, the economic order quantity (EOQ) is:

where D is annual demand in units, S is ordering cost per order, and H is annual holding cost per unit. EOQ assumes relatively constant demand, known parameters, instantaneous replenishment, no stockouts and no quantity discounts. These assumptions often require modification in practice.

Safety stock protects against uncertainty. A simplified reorder point is:

If daily demand is 100 units, lead time is 5 days and safety stock is 200 units, the reorder point is 700 units. The calculation assumes that demand and lead-time units are aligned.

6.2 Service and resilience

The order fill rate measures the proportion of demand supplied immediately from available stock. On-time-in-full (OTIF) requires both delivery by the promised time and the complete quantity. Supply-chain dashboards should also monitor supplier concentration, lead-time variability, single-source exposure and recovery time after disruption. Optimising only purchase price can increase total cost through poor quality, delays or risk.

7. Customer-Service Analytics

Customer-service analytics examines contact demand, response speed, resolution quality, channel performance and customer experience.

Common metrics include average handle time (AHT), first-contact resolution (FCR), abandonment rate, service level and customer satisfaction (CSAT). AHT is often calculated as:

AHT should not be maximised blindly. Short calls may increase repeat contacts if issues are not resolved. FCR is:

Net Promoter Score (NPS) is the percentage of promoters minus the percentage of detractors, based on a defined recommendation question. It ranges from −100 to +100 and is not an interval-scale measure of individual customer value. Organisations should use it alongside behavioural and operational measures.

Text analytics and speech analytics can classify contact reasons, detect emerging problems and identify sentiment. Automated systems require monitoring for language bias, incorrect escalation and privacy risks. Customers should have a clear route to human assistance where automated decisions affect access, refunds or complaints.

8. Human-Resources Analytics

HR analytics applies evidence to workforce planning, recruitment, development, performance, wellbeing and retention. It is especially sensitive because employee data concerns identifiable people and power relationships.

8.1 Workforce metrics

Turnover rate can be estimated as:

The organisation should specify whether voluntary and involuntary separations are combined. Absence rate may be calculated as lost workdays divided by scheduled workdays, but leave policies and part-time schedules must be handled consistently.

Time-to-fill measures recruitment speed, while quality-of-hire attempts to assess subsequent performance, retention or capability. A fast hire is not necessarily a good hire.

8.2 Predictive HR analytics and fairness

A churn model may estimate the probability that an employee leaves within a specified period. The model should be evaluated using an outcome definition, validation data and metrics such as precision, recall, calibration and false-positive rates. A model can be statistically accurate yet unfair if errors are distributed unequally across groups.

HR analytics should avoid using sensitive information without a legitimate purpose. Employees should understand what data is collected, how it is used and how decisions can be reviewed. Predictive scores should support conversations and organisational improvement, not label employees as inevitable leavers.

9. Cross-Functional Integration

Functional metrics can conflict. Marketing may seek more leads, sales may seek higher-quality leads, finance may seek lower acquisition cost, operations may face capacity limits, and customer service may experience increased contact volume. A dashboard that optimises each function separately can therefore harm the organisation as a whole.

A balanced performance framework links:

• outcomes: revenue, profit, retention and customer value;
• drivers: conversion, quality, productivity, service and employee capability;
• constraints: cash, capacity, risk, compliance and sustainability; and
• learning measures: experimentation, innovation and process improvement.

Data definitions are especially important. “Customer,” “order,” “active employee,” “revenue” and “resolved case” should have shared definitions across systems. A semantic layer or governed data catalogue can reduce conflicting reports.

10. Illustrative Integrated Case: Northstar Home Appliances

Background. Northstar Home Appliances sells through its website and retail partners. Revenue increased during a promotional quarter, but profit declined. Customer complaints about late delivery rose, and employee turnover in the service centre increased.

Problem. Management initially proposed increasing advertising because website traffic was below target. The analytics team instead investigated the full customer journey.

Analysis. Marketing data showed a conversion rate of 3.0%, up from 2.6%, but the campaign attracted a high proportion of low-margin products. Finance reported that contribution per order had fallen from $42 to $31. Sales data showed that promotional leads were numerous but had a lower repeat-purchase rate. Supply-chain data showed that the primary warehouse operated at 96% effective capacity, while stockouts affected popular products. Customer-service data showed FCR falling from 78% to 68% and AHT increasing because agents handled delivery-status contacts. HR data showed higher absence and turnover among employees working peak shifts.

Approach. The team created a cross-functional contribution measure rather than evaluating the campaign on revenue alone. It linked campaign source, product margin, fulfilment cost, delivery performance, repeat purchase and service contacts. A holdout group was used for the next promotion to estimate incremental sales. Operations used demand forecasts to reserve capacity for high-margin products, while customer service introduced proactive delivery notifications. HR adjusted shift planning and reviewed workload and training.

Outcome. In the next test period, incremental orders increased modestly rather than dramatically, but contribution per order recovered to $39. Delivery-related contacts declined, FCR improved to 75%, and peak-shift absence decreased. The campaign was retained for high-margin segments and redesigned for low-margin segments. The result illustrates that the best decision was not “spend more” or “spend less,” but “target demand that the organisation can profitably and reliably serve.”

Learning points. First, revenue is not equivalent to value. Second, customer experience can be an operational constraint rather than merely a service issue. Third, functional metrics must be connected through a common customer and order definition. Finally, causal testing is preferable to assuming that observed sales growth was caused entirely by advertising.

11. Governance, Ethics and Implementation

Analytics across functions creates several risks:

• Privacy: personal, transaction, employee and behavioural data may be collected beyond what individuals reasonably expect.
• Security: unauthorised access, weak retention controls and insecure data transfers can expose sensitive information.
• Bias: historical decisions may encode discrimination into recruitment, credit, targeting or service models.
• Opacity: managers may not understand why a model produced a score or recommendation.
• Metric gaming: employees may optimise a measured target while harming the underlying objective.
• Data drift: customer behaviour, economic conditions and processes change over time.

Good practice includes purpose limitation, data minimisation, access controls, documented definitions, model validation, human oversight, audit trails and periodic monitoring. Analysts should report uncertainty and limitations, not only favourable results. Where applicable, organisations must comply with privacy, employment, consumer-protection, financial-reporting and sector-specific regulations.

Artificial intelligence can assist with forecasting, recommendation, anomaly detection, natural-language classification and automated reporting. Its use is appropriate only when the business objective, data rights, validation method and accountability structure are clear. Generative AI outputs require human review because they may contain fabricated explanations, confidential-data leakage or inconsistent reasoning.

12. Best Practices for Professional Analysts

1. Begin with a decision, not a dataset.
2. Define the population, time period, numerator, denominator and unit for every KPI.
3. Reconcile important financial figures with authoritative records.
4. Separate correlation, prediction and causation.
5. Use a baseline, comparison group or control where feasible.
6. Segment results when averages conceal important differences.
7. Test assumptions through sensitivity and scenario analysis.
8. Present operational implications, not merely statistical significance.
9. Monitor unintended consequences and fairness.
10. Assign an owner and review date to every implemented recommendation.

Chapter Summary

Analytics becomes valuable when it links data to decisions across the organisation. Marketing uses conversion, CAC, CLV, attribution and experimentation; sales uses funnel conversion, pipeline coverage and calibrated forecasts; finance uses contribution, break-even and cash-cycle measures; operations uses capacity, productivity, quality and flow; supply chains use inventory, reorder-point and service measures; customer service uses AHT, FCR and experience indicators; and HR uses turnover, absence, recruitment and fairness measures. Each metric depends on clear definitions and appropriate assumptions.

Cross-functional analysis is essential because local optimisation can damage total performance. Reliable analytics requires governed data, causal awareness, ethical use, model monitoring and action-based evaluation. The most effective analysts combine quantitative competence with process understanding, communication and professional judgement.

Key Terms

A/B test: A controlled comparison between treatment and control conditions.

Average handle time: Average time spent handling a customer contact, including defined related activities.

Break-even point: Output or sales level at which total contribution equals fixed costs.

Capacity utilisation: Actual output expressed as a proportion of available capacity.

Customer acquisition cost: Average acquisition expenditure per new customer.

Customer lifetime value: Expected discounted contribution from a customer over a defined or modelled relationship.

First-contact resolution: Proportion of contacts resolved during the first interaction.

Forecast calibration: Agreement between predicted probabilities and observed frequencies.

Funnel: Sequential stages through which a prospect or transaction progresses.

Inventory turnover: Cost of goods sold divided by average inventory.

Key performance indicator: A strategically important, formally defined performance measure.

Little’s Law: Relationship stating that average work-in-process equals throughput rate multiplied by average flow time in a stable system.

Pipeline coverage: Qualified pipeline value relative to a sales target.

Reorder point: Inventory level at which replenishment should be initiated.

ROAS: Attributable advertising revenue divided by advertising expenditure.

Safety stock: Inventory held to protect against demand or supply uncertainty.

Segmentation: Division of a population into groups with meaningful similarities.

Turnover rate: Employee separations relative to average employment over a period.

Review Questions

Short Answer Questions

1. Distinguish descriptive, predictive and prescriptive analytics.
2. Why should CAC be compared with contribution-based CLV rather than only first-purchase revenue?
3. State two assumptions of the EOQ model.
4. What does first-contact resolution measure?
5. Why can a high inventory turnover ratio be interpreted cautiously?

Long Answer Questions

1. Explain how marketing analytics can combine segmentation, experimentation and customer lifetime value to improve campaign decisions.
2. Discuss the relationship between sales-funnel metrics, pipeline coverage and forecast reliability.
3. Explain how finance and operations analytics can jointly evaluate a capacity-expansion decision.
4. Compare the main performance measures used in supply-chain and customer-service analytics.
5. Discuss ethical and governance requirements for predictive HR analytics.

Analytical / Application Questions

1. A company has fixed costs of $90,000, a selling price of $120 and variable cost of $75 per unit. Calculate break-even units and operating profit at 2,500 units.
2. A warehouse has average throughput of 40 orders per hour and average flow time of 3 hours. Use Little’s Law to estimate average work-in-process, and state the assumption needed for the calculation.
3. A campaign generates $180,000 attributable revenue from $30,000 advertising expenditure, but contribution margin is 20% of revenue. Calculate ROAS and advertising contribution before other costs. Should the campaign automatically be expanded? Explain.

Multiple Choice Questions with Answers

Question 1

Which measure is most directly designed to compare acquisition expenditure with the number of new customers?

A. Customer acquisition cost

B. First-contact resolution

C. Inventory turnover

D. Capacity utilisation

Correct Answer: A. Customer acquisition cost

Question 2

A product sells for $80, has a variable cost of $50, and fixed costs are $60,000. What is the break-even volume?

A. 1,500 units

B. 2,000 units

C. 2,500 units

D. 3,000 units

Correct Answer: B. 2,000 units

Question 3

Which condition is required for the simple EOQ model described in the chapter?

A. Demand is completely unpredictable

B. Replenishment is instantaneous

C. Stockouts are mandatory

D. Holding cost is zero

Correct Answer: B. Replenishment is instantaneous

Question 4

A qualified pipeline is worth $400,000 and the sales target is $100,000. What is the pipeline coverage ratio?

A. 0.25

B. 2

C. 4

D. 5

Correct Answer: C. 4

Question 5

According to Little’s Law, if throughput is 40 orders per hour and average flow time is 3 hours, average work-in-process is:

A. 13.3 orders

B. 37 orders

C. 120 orders

D. 400 orders

Correct Answer: C. 120 orders

Question 6

Which practice provides stronger evidence that a marketing intervention caused an observed outcome?

A. Last-touch attribution alone

B. A randomised holdout experiment

C. A larger descriptive dashboard

D. A higher advertising budget

Correct Answer: B. A randomised holdout experiment

Question 7

What does first-contact resolution measure?

A. The average cost of a customer contact

B. The proportion of contacts resolved in the first interaction

C. The time required to recruit a service agent

D. The percentage of customers who recommend a firm

Correct Answer: B. The proportion of contacts resolved in the first interaction

Question 8

If annual cost of goods sold is $600,000 and average inventory is $100,000, inventory turnover is:

A. 0.17 times

B. 6 times

C. 60 times

D. 600 times

Correct Answer: B. 6 times

Question 9

Why should a predictive HR model be checked for group-level false-positive differences?

A. To increase advertising reach

B. To assess potential unequal treatment

C. To calculate inventory holding cost

D. To determine website conversion rate

Correct Answer: B. To assess potential unequal treatment

Question 10

A campaign produces $180,000 attributable revenue from $30,000 advertising expenditure. Its ROAS is:

A. 0.17

B. 3

C. 6

D. 30

Correct Answer: C. 6