1. Hi, Florence. First, let me introduce myself.

I am Henry, has about 5 years’ experience

in forecasting since graduate from school. Thank you for giving me this opportunity to
take this interview.

2. Now let me start to explain my analysis for this case. I also save the excel file in one
drive as backup. You can find all processes, and function I used and graph in that file.

3. In part A. I used VLOOKUP to find the information.

NPA I is the first three numbers of subscriber (səbˈskraɪbər/) number, and NXX is the
middle three numbers of subscriber number. We can get brand, city and province from
Table 2 and table 3 after we know NPA and NXX
Question 2 to 4, We can build a pivot table to get the answers, I inserted Year and
month as column, brand, province, city and account number as Row, and Activation
amount and deacts amount as value in pivot table.
Finally. We can get the answer as I show in this PowerPoint.

4. In part B, question 1. I also use pivot table which build from table 4. Year and month is
rows, and sum of segment 1 and segment 2 are values. Finally, group month in quarter.
Like January to march is the first quarter. However, we can group other continue
months if we need fiscal year as show, you know each company’s fiscal year is
different. like November to January is quarter 1.
by the way, I find the original data (Year and month) is not date as show in table 4, so
we need convert data to real date data in table 4.

5. For question 2. Actually, I am not sure I under the question request. I guess we need
to assume some definition to answer the question. Therefore, I assume Sale Channel:
Sales channels are the methods or pathways businesses use for selling their products
and services to consumers. Channel A is sale by Store in different locations. Channel B
is sale Online and Channel C is sale by Phone. Also, I assume Segments are two rogers
internet plan for customer which is both provide in Ontario and Quebec. They are plan
A and Plan B.
After these two hypotheses (Haiposis), I got three observations (/ˌɑbzɜːrˈveʃən/).
 From the first graph, Plan B is more popular in Ontario, while Plan A is slightly
more popular in Quebec. When we consider the overall sales volume in each
province, we see that Ontario's sales volume is three times that of Quebec,
which is in line with the population ratio. This also means that the sales
proportion of Rogers' services in both provinces is similar. As a result, the
competitive pressure in these provinces is also similar.
 From the second graph, Comparing Plan A and Plan B, it's easier to recommend
Plan A in Channel C. This suggests that telemarketers may have more
authority /əˈθɔːrəti/ to offer customers relatively larger discounts for Plan A.
Additionally, we have noticed that customers tend to place orders by phone or
in-store. Only a minority (/maɪˈnɔːrəti/) of people prefer to order directly
online, possibly due to the online discounts not being attractive enough.
  In the third chart, Service cancellations primarily happened in Channel B. This
might be due to people's preference for deactivating services online to avoid
direct communication. It could be beneficial for the company to restrict (rɪ
ˈstrɪkt/) cancellations through the internet.
There are three part I find based on my assumption, but not only three absolutely.

6. For question three, I find channel A had the strongest performance in 2023 Q1 by pivot
table which only set 2023 Q1 as time filter.
The reasons for better sales in physical stores can vary and depend on multiple factors,
including the type of products, geographical location, market competition, and
marketing strategies. Here are some potential reasons:
• Immediate Gratification: Shopping in physical stores allows customers to
receive their purchases immediately, without the need to wait for delivery.
• Interpersonal Interaction: Physical stores provide the chance for communicate
with sales personnel and other customers, which can assist customers in
making informed purchasing decisions.
• Location: The geographic location of a physical store is importanrt. Stores in
busy commercial areas or high-traffic shopping centers are more likely to
attract more customers.
• Marketing Strategies: Marketing strategies, such as promotions, discounts, and
special sales events, can draw in more customers to physical stores.

7. If let me discover further, I will ask my business partner more information to support
the above results. like
• The time of activation and deact, I need to know what time is peak time which
can support our sale team to make right decision.
• Promotions with each order in each channel and segment.
• Competitors' promotion at the same period, like bell’s, TELUS’s and so on. It can
help us make adjustment in time.
• How long does it take to fulfill orders in each channel, it can save labor cost in
further?
• Dose the customer talk about competitors‘plan, which can optimize our words
in sale process.
• New customers or loyalty customers, get them relative promotion.
• Customers’ birthday, gender, ethnicity (/eθˈnɪsəti/), postcode……make the
forecasting more actual.

8. For Part C. I used time series analysis to come up the forecast.

maybe I can use forecasting analysis function directly in Excel, but I dont have a
window system, it led my excel loss this function. I only have a MacBook at home.
Sorry with that. So my method maybe looks a liitle stupid because it need do lots of
calculation. Let me introduce my way now.
i. Data Collection and Preprocessing: Choosing the useful date with pivot
table from Table#5. e.g. Use the province (ON) and segment
 (Segment_1) as filter to get history data of Segment_1 in Ontario's
information
ii. Adding the secondary columns to calculate necessary coefficient (
/ˌkoʊɪˈfɪʃ(ə)nt/) for time series analysis. (MA12,
CMA12,St,It,Tt,Deseasonalize), all the formula are show in excel file
iii. Get forecasting by using coefficient (intercept, t and actual data in
history)
iv. Comparing actual data and forecasting data to make sure the trend is
right.

9. Let see the result. Red line is actual in history, blue is trend which I got with time
series, and green line is forecasting including history and future. And I filled further
number in the chart. May to December in 2023.
In the graph Segment 1 in Ontario. There are some gaps between actual and
forecasting. I guess there maybe some reasons cause it. the details are not enough, I
only have history number in each month, but we are not sure what happened in
history, like some unexpected factors also can cause these gaps.

10. Same as segment 1 in Quebec. Therefore, segment 1 maybe happened some big
change from 2019 to 2023. For example, sometimes segment 1 has big promotion or
no promotion suddenly in next month.

11. Let see segment 2 in Ontario and Quebec. The forecasting of history can match the
actual of history. Therefore, our forecasting for further 7 months are more accurate
than segment 1 in Ontario and Quebec.

12.

13. I just add the result to get total in Ontario and Quebec

14. For question 2

I applied the same approach to different provinces and segments.
Reason:
• we only have access to monthly and yearly segment data, and there are
no additional details available that could influence the results.
• Using different approaches for forecasting may indeed lead to
confusion, especially when dealing with the four different segments. It's
important not to aggregate (/ˈæɡrɪɡət/) data from these different
methods in total table, as doing so could result in inaccuracies. It's best
to keep the data separate and analyze each one according to the specific
method used.
• If we had more relevant forecasting data, linear regression could be a
suitable method for improving our forecasts, not time series analysis.
 15. To achieve a more reasonable forecast, consider incorporating
the following internal and external data and factors: I list each in each
parts

Internal Data:

1. Historical Sales Data: Analyze past sales trends, seasonality, and

patterns to make informed forecasts.
2. Customer Data: Understand your customer behavior, demographics,
and purchasing history to predict future preferences.
3. Inventory Levels: Keep track of your product inventory, as stockouts or
overstock can significantly impact sales.
4. Marketing and Promotions: Consider the impact of marketing
campaigns, discounts, and promotions on sales figures.
5. Website Traffic: Monitor web traffic, click-through rates, and conversion
rates for online sales.
6. Customer Feedback: Incorporate customer feedback and reviews to
gauge product satisfaction and improve forecasting.

External Data:

1. Economic Indicators (/ˈɪndɪkeɪtərz/): Keep an eye on economic factors

like GDP, unemployment rates, and inflation, as they can influence
consumer spending.
2. Competitor Data: Analyze your competitors' sales data and market
share to anticipate market trends.
3. Industry Trends: Stay updated on industry trends, emerging
technologies, and changing consumer preferences.
4. Seasonal Effects: Consider how seasons, holidays, and special events
impact sales.
5. Demographic (/ˈɪndɪkeɪtərz/)Data: Factor in demographic information
related to your target audience.
6. External Events: Assess how external events, such as natural disasters
(/ˈɪndɪkeɪtərz/)or geopolitical (/ˌdʒiːoʊpəˈlɪtɪkl/) changes, may influence
demand.

16. Thank you for listening.

MA12: the Average of values for 12 consecutive months
CMA12: the Average of values for 2 consecutive MA12
St, It: other nonrandom sources of variations of series.（Total/CMA12）
St: describe the movement along the term, the average of each month’s St, It
Deseasonalize: Seasonal adjustment or deseasonalization is a statistical method for removing
the seasonal component of a time series. (Total/ St)Total/CMA12）
St: describe the movement along the term, the average of each month’s St, It
Deseasonalize: Seasonal adjustment or deseasonalization is a statistical method for removing
the seasonal component of a time series. (Total/ St)