Evaluating options with business analytics

Accounting Ahead | Original study resource | accountingahead.com.au

Evaluating options with business analytics

MODULE 5 | 5.3.2 | STUDENT COMPANION

Business analytics finds patterns in data and communicates what they mean. Data comes from inside

and outside the organisation: sales records, customer and staff surveys, research on rivals and the

industry, economic statistics and financial results. Statistical models reveal patterns and estimate the

probability of outcomes, testing qualitative judgements against hard numbers.

Three types of business analytics

1. Descriptive

What happened? Historical data

explains past successes and failures

Peak: weekday mornings run at a

quarter of capacity; most cancellations

come in month three

2. Predictive

What is likely to happen? Models

estimate the probability of future

events

Peak: which members are likely to

cancel next month

3. Prescriptive

What should we do? Recommends

actions and estimates their outcomes

Peak: a free coaching session where it

pays; an $18 weekday morning price

A prescriptive pricing test: a typical weekday morning

Now: 50 climbers x $25 $1,250

Model's price: 90 climbers x $18 $1,620

Extra revenue per weekday morning $370

The walls were standing idle, so most of the extra revenue is profit

GLOBAL STRATEGY AND LEADERSHIP | M5 | 5.3.2 | STUDENT COMPANION

1

Accounting Ahead | Original study resource | accountingahead.com.au

Customer lifetime value at Peak

Average monthly fee $90

Contribution per month at a 70% gross margin $63

Average membership length 20 months

Customer lifetime value: $63 x 20 $1,260

Cost of a free coaching session (retention offer) $20

Extra margin if members stay 2 more months: $63 x 2 $126

Net gain per member: $126 - $20 $106

Numbers turn 'look after our members' into an option that can be compared with others

Other common uses: measuring how well an advertising campaign works, and predicting how likely customers

are to switch to a competitor if a product changes.

Analytics before buying a rival gym (imagined)

Traditionally a buyer relied on the target opening its books. Today buyers also form their own view from external

and internal data before they negotiate.

Question Data Peak could use

How many members does the rival have? Public reviews, social media activity, class timetables

How do its prices and hours compare? Its published prices and opening hours against Peak's

What would the site cost to run? Peak's own cost data, applied to the rival's site

What will its walls and equipment cost? Age and condition; maintenance and replacement costs

The result A sharper negotiating position, and a better plan for combining the

businesses

GLOBAL STRATEGY AND LEADERSHIP | M5 | 5.3.2 | STUDENT COMPANION

2

Accounting Ahead | Original study resource | accountingahead.com.au

Data lake or data swamp?

A well-governed data lake

Stores raw data, structured and unstructured

Transforms copies for each analysis; originals stay

intact

Labelled, owned and quality-checked

Collected with consent and protected under privacy

law

A data swamp

Vast and chaotic, with no organisation

No labels or descriptions

No rules about quality

Nobody can find anything useful in it

Peak's lake: booking records, app activity, online reviews, coaching notes, wall sensors

Governance, ethics and law

Collect customer data with consent, and protect it under privacy law.

Check models for bias: a model trained on a narrow group of past customers may make poor or unfair

predictions about new ones.

Give data clear owners and descriptions, so the lake doesn't become a swamp. CPAs, with professional

obligations of their own, should raise these risks early.

Case note: Zillow Offers

Zillow, the US property website, launched Zillow Offers in Phoenix in April 2018, using its pricing algorithms to

buy homes directly, renovate them and resell them. By 2021 its Zestimate valuation served as the opening cash

offer on some homes, and in the third quarter of 2021 it bought 9,680 homes. But its models misjudged where

prices were heading. On 2 November 2021 Zillow announced it would close Zillow Offers and cut about a quarter

of its staff, after a write-down of about US$304 million on homes it had paid too much for; the home-buying

business lost US$881 million before tax in 2021. Chief executive Rich Barton said the unpredictability in

forecasting home prices 'far exceeds what we anticipated'.

Sources: Zillow Group third-quarter 2021 release and shareholder letter (2 November 2021); fourth-quarter 2021 results (10 February

2022); Zillow release, 12 April 2018.

Judgement over models

Leaders need to know what a model assumes, how accurate it has been and how badly things go wrong when it

errs. A model that is right on average can be ruinous if its errors all point the same way at once. If Peak's model

recommends closing its oldest gym because it earns the lowest margin per square metre, judgement adds what

the model can't see: it is where Peak began, and it hosts the competitions that attract new members to all three

sites. Questions a CPA should ask: what is the error rate, what happens in a downturn, and how much could

we lose if the model is wrong?

GLOBAL STRATEGY AND LEADERSHIP | M5 | 5.3.2 | STUDENT COMPANION

3

Accounting Ahead | Original study resource | accountingahead.com.au

Practice questions

MODULE 5 | 5.3.2 | TRY THESE BEFORE THE WORKED ANSWERS

Question 1 - Classify the analyses

Analysis Descriptive, predictive or

prescriptive?

(a) A chart of last year's sales by month

(b) A model estimating which customers will switch to a rival next quarter

(c) A system that sets tomorrow's hotel room prices to maximise revenue

(d) A report comparing this year's staff turnover with last year's

(e) A forecast of demand for flu vaccines next winter

(f) A tool that recommends which stores should receive extra stock, and

how much

Question 2 - Customer lifetime value

(a) Calculate customer lifetime value. (b) Calculate the net gain from the offer per customer. (c) If only 40% of

customers who receive the offer actually stay a month longer, is it still worthwhile?

Item FreshCrate (imagined meal kits)

Average monthly spend $260

Gross margin 30%

Average subscription 8 months

Retention offer A $25 discounted box that keeps the average customer one month

longer

GLOBAL STRATEGY AND LEADERSHIP | M5 | 5.3.2 | STUDENT COMPANION

4

Accounting Ahead | Original study resource | accountingahead.com.au

Question 3 - When the model says 'buy'

A model at an imagined hardware chain recommends buying a rival chain for $40 million, predicting $6 million a

year in extra profit. How should the board respond?

GLOBAL STRATEGY AND LEADERSHIP | M5 | 5.3.2 | STUDENT COMPANION

5

Accounting Ahead | Original study resource | accountingahead.com.au

Worked answers

MODULE 5 | 5.3.2 | WORKED ANSWERS - ATTEMPT THE QUESTIONS FIRST

Question 1

Analysis Type

(a) Last year's sales by month Descriptive: what happened

(b) Customers likely to switch Predictive: what is likely to happen

(c) Tomorrow's room prices Prescriptive: recommends an action

(d) Staff turnover comparison Descriptive

(e) Flu vaccine demand forecast Predictive

(f) Stock allocation tool Prescriptive

Question 2 - FreshCrate

(a) Contribution per month: $260 x 30% $78

Lifetime value: $78 x 8 $624

(b) Extra margin: one more month $78

Net gain: $78 - $25 $53

(c) Expected extra margin: 40% x $78 $31.20

Expected net gain: $31.20 - $25 $6.20

Still worthwhile, but only just: the result is very sensitive to the assumptions

Question 3 - The board's response

Treat the recommendation as a starting point, not a decision.

Ask what the model assumes: customer retention after the deal, synergies, prices and costs.

Ask how accurate the model has been, and how good the data on the rival is (much of it external).

Test the downside: a downturn, lost customers, integration problems, culture clashes.

Commission due diligence, and have a CPA test the $6 million against the $40 million price, for example with

net present value and a range of scenarios.

GLOBAL STRATEGY AND LEADERSHIP | M5 | 5.3.2 | WORKED ANSWERS

6