Restaurant Business Analytics
Cenacle
The Challenge
The primary challenges faced by Restaurant Businesses, like any other business are:
- Customer Retention
- How do I make my customers come back?
- Cost Reduction
- How do I increase my profit margins (without increasing the prices) ?
While there could be more challenges and goals apart from the above, usually they all can be expressed in one or both of the above two. Surprisingly, these above two are not necessarily independent. What we observed in the course of our many Big-Data Analytics solution deployments for Businesses is, fixing one usually fixes the other automatically.
The Solution
The key to developing a successful solution for the above challenges is:
- Understanding the customer behaviour
- Who are the customers most likely to stay / leave?
- What are the factors that affect a customer to stay / leave?
The primary driver for achieving Customer Retention is: Customer Satisfaction, which is driven by:
- Service / Product Quality
- Response Time
- Expectation (mis)match
There are further secondary drivers such as: Price, external factors (weather …)
- What if tomorrow rains? Would it affect my Sales?
- Rather what if tomorrow is Sunny?
- What if the Christmas falls on Sunday? How would it affect my sales?
The primary drivers for Cost Reduction on the other hand are:
- Waste Elimination
- Produce right levels to match the consumption
- Optimal work-labor costs
- Right labor schedules that match the demand
We can eliminate waste if we know the consumption / demand before hand. But How can one know the consumption / demand before hand?
Enter Predictive Analytics
Cenacle offers a comprehensive predictive analytics solution that can analyze your business data in-depth and answer questions such as those raised above. Some of the features of our Predictive Analytics Solution for the Restaurant Business are:
- Interactive exploratory analysis tools
- Robust forecasting algorithms
- Analysis on multiple levels (item-wise, weekday-wise, hour-wise etc.)
- Forecasting results from multiple models (weather based, temperature based, time-series based …)
- Multiple output formats for auxilary data
Model
- Order Prediction
- How many orders expected in the next 7 days?
- Results from multiple prediction algorithms
- Association Rules
- Which items are being frequently ordered together?
- Can lead to better pricing strategy
Features
- Item-wise sale predictions
- Cost-cutting
- Drop the low sale items from menu for specific days
- Inventory management
- Right levels of stock reduces waste, improves quality
- Labor schedules
- Improves customer satisfaction, increases work-life balance for employees
- Customer churn analysis
- Who are the customers most likely to stay, who would deflect?
- Design custom loyalty programs to increase sales
- When is the next customer visit most likely?
- Price analytics: what would be the effect of price increase on each?
- Exploratory Analysis
- What are the top-3 drivers for my sales during Holiday Season?
- Which section needs more labor and during which period?
- What are the effects of Weather on my business?
- If I drop an item from my product list, how would it affect other item sales?
- Social Network Sentiment Analysis
- What are people talking about my brand?
- How many people are feeling positive about my products?
- How are my competitor’s products faring against my products?
- What is the most (dis)liked feature about my product/service?
- If I release new product, how many people are likely to buy it?
Data Analysis
After cleansing the data to remove NA and fill missing dates, the data exhibited the following:
##
## found Factors: holiday fname discname tblgrp table weather
## found Numerics: amount guests temp
## found UniqueIDs: id
## found Date: date
## found Time: time
Basic statistics of the supplied data is as below:
## date holiday
## Min. :2013-01-01 00:00:00 Valentine's Day : 165
## 1st Qu.:2013-04-19 00:00:00 Washington's Birthday: 133
## Median :2013-08-12 00:00:00 New Year's Day : 115
## Mean :2013-08-12 06:43:06 Mother's Day : 101
## 3rd Qu.:2013-11-30 00:00:00 Groundhog Day : 99
## Max. :2014-03-31 00:00:00 (Other) : 1583
## NA's :17108
## fname discname tblgrp amount
## Deva :8082 Ramasubbu:8082 Bar : 459 Min. : 0.00
## Shanker:4184 Krishna :4184 Catering : 7 1st Qu.: 20.90
## Guru :3719 Govind :3719 NULL : 1017 Median : 32.40
## Parthi :1627 Chin :1627 PRIVATE ROOM: 442 Mean : 41.07
## Jey : 802 waiter : 802 Section 1 :10431 3rd Qu.: 48.70
## James : 434 Pereira : 434 Section 2 : 3500 Max. :2600.00
## (Other): 456 (Other) : 456 TAKEOUT : 3448
## guests weather temp weekday
## Min. : 0.00 Sunny :10202 Min. :12.00 Length:19304
## 1st Qu.: 1.00 Rain : 4104 1st Qu.:37.00 Class :character
## Median : 2.00 Fog : 1552 Median :54.00 Mode :character
## Mean : 2.33 Snow : 1311 Mean :54.93
## 3rd Qu.: 3.00 Fog-Rain : 803 3rd Qu.:73.00
## Max. :250.00 Rain-Snow: 448 Max. :98.00
## (Other) : 884
## week month quarter year
## Min. : 1.00 Length:19304 Length:19304 Min. :2013
## 1st Qu.: 9.00 Class :character Class :character 1st Qu.:2013
## Median :19.00 Mode :character Mode :character Median :2013
## Mean :22.61 Mean :2013
## 3rd Qu.:36.00 3rd Qu.:2013
## Max. :53.00 Max. :2014
##
## hour
## Min. : 0.00
## 1st Qu.:12.00
## Median :16.00
## Mean :15.65
## 3rd Qu.:19.00
## Max. :22.00
## Forecasting
The supplied restaurant business data is subjected to various time-series analysis methods to derive forecasting for the next 7 days subjected to the conditions of weather, temperature, holiday events etc and the results are as below:
Method1
## Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
## 66.00000 36.49739 19.35912 53.63566 10.286674 62.70811
## 66.14286 37.84754 20.56807 55.12701 11.420870 64.27420
## 66.28571 39.68802 22.26360 57.11244 13.039665 66.33637
## 66.42857 66.28861 48.71549 83.86173 39.412842 93.16437
## 66.57143 60.43428 42.70873 78.15983 33.325396 87.54316
## 66.71429 49.65886 31.77717 67.54055 22.311169 77.00655
## 66.85714 33.03198 14.99043 51.07352 5.439816 60.62414
## 67.00000 37.08341 18.67096 55.49585 8.924005 65.24281
## 67.14286 38.43355 19.85577 57.01134 10.021282 66.84582
## 67.28571 40.27403 21.52727 59.02080 11.603326 68.94474
## 67.42857 66.87462 47.95525 85.79400 37.939937 95.80931
## 67.57143 61.02030 41.92471 80.11588 31.816124 90.22447
## 67.71429 50.24487 30.96951 69.52024 20.765745 79.72401
## 67.85714 33.61799 14.15928 53.07670 3.858465 63.37752
Method2
## Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
## 66.00000 37.21038 18.87964 55.54112 9.175935 65.24482
## 66.14286 39.48814 20.89990 58.07638 11.059878 67.91640
## 66.28571 40.99597 22.40049 59.59144 12.556646 69.43529
## 66.42857 54.04485 35.44918 72.64053 25.605220 82.48449
## 66.57143 54.16538 35.56970 72.76106 25.725738 82.60502
## 66.71429 47.36280 28.76712 65.95848 18.923157 75.80244
## 66.85714 38.27826 19.68257 56.87394 9.838613 66.71790
## 67.00000 37.13959 17.76499 56.51419 7.508698 66.77048
## 67.14286 40.69062 21.29444 60.08680 11.026719 70.35452
## 67.28571 40.61518 21.21839 60.01197 10.950345 70.28001
## 67.42857 52.06673 32.66992 71.46354 22.401872 81.73159
## 67.57143 51.05693 31.66012 70.45374 21.392074 80.72179
## 67.71429 45.11795 25.72114 64.51476 15.453087 74.78281
## 67.85714 39.02473 19.62792 58.42154 9.359874 68.68959
Method3
## Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
## 66.00000 36.89334 23.05747 50.72921 15.733198 58.05348
## 66.14286 36.40287 22.39283 50.41290 14.976369 57.82937
## 66.28571 39.89813 25.71608 54.08019 18.208547 61.58772
## 66.42857 67.48151 53.12949 81.83353 45.531989 89.43103
## 66.57143 59.46346 44.94346 73.98345 37.257044 81.66987
## 66.71429 51.97358 37.28753 66.65962 29.513211 74.43394
## 66.85714 28.70711 13.85687 43.55735 5.995626 51.41859
## 67.00000 36.89334 21.88070 51.90598 13.933490 59.85319
## 67.14286 36.40287 21.22957 51.57617 13.197310 59.60843
## 67.28571 39.89813 24.56586 55.23041 16.449440 63.34683
## 67.42857 67.48151 51.99189 82.97113 43.792176 91.17084
## 67.57143 59.46346 43.81807 75.10884 35.535903 83.39101
## 67.71429 51.97358 36.17396 67.77319 27.810152 76.13700
## 67.85714 28.70711 12.75476 44.65946 4.310090 53.10412
Exploratory Analysis
Amount of income generated on hourly basis arranged in weekday order is as below:
Number of orders observed on hourly basis arranged on weekday order is as below:Interactive Exploration
You can explore the data on the fly with various tools provided in the below. Try dragging and dropping few other fields and changing the chart type.
A sample of the original input data is as below:## date holiday fname discname tblgrp amount guests
## 1: 2013-01-01 New Year's Day Deva Ramasubbu Bar 17.90 2
## 2: 2013-01-01 New Year's Day Deva Ramasubbu Bar 35.80 4
## 3: 2013-01-01 New Year's Day Deva Ramasubbu Bar 15.00 1
## 4: 2013-01-01 New Year's Day Deva Ramasubbu TAKEOUT 10.95 0
## 5: 2013-01-01 New Year's Day Deva Ramasubbu Section 2 10.95 1
## ---
## 19300: 2014-03-31 NA Deva Ramasubbu Section 1 25.90 2
## 19301: 2014-03-31 NA Deva Ramasubbu Section 1 51.20 5
## 19302: 2014-03-31 NA Parthi Chin NULL 27.90 0
## 19303: 2014-03-31 NA Deva Ramasubbu Section 1 45.30 2
## 19304: 2014-03-31 NA Deva Ramasubbu Section 1 69.70 4
## weather temp weekday week month quarter year hour
## 1: Sunny 35 Tuesday 1 January Q1 2013 0
## 2: Sunny 35 Tuesday 1 January Q1 2013 0
## 3: Sunny 35 Tuesday 1 January Q1 2013 0
## 4: Sunny 35 Tuesday 1 January Q1 2013 11
## 5: Sunny 35 Tuesday 1 January Q1 2013 11
## ---
## 19300: Rain 39 Monday 13 March Q1 2014 20
## 19301: Rain 39 Monday 13 March Q1 2014 21
## 19302: Rain 39 Monday 13 March Q1 2014 21
## 19303: Rain 39 Monday 13 March Q1 2014 21
## 19304: Rain 39 Monday 13 March Q1 2014 21
At Cenacle we build Big Data Analytics Engines for
- Automotive
- E.g.: Predictive Maintenance (CBM) systems
- Healthcare
- E.g.: Personalized Medicine Regimes and Decision Support Systems
- Retail
- E.g.: Real-time Recommendation Engines
- Energy
- E.g.: Demand Planning using Smart Metering Systems
- BFSI
- E.g.: Usage-based Insurance Systems