BLUETHUMB CANVAS SUCCESS PROJECT
2023-10-15
Introduction
Creating a brand name in the art industry is indeed a challenge on its own. Bluethumb was a company established in 2012 with a mission to empower Australian Artists. They did not agree with the habit of having to wait for an artist to reach the level of having their own gallery or exhibition, instead started Australia’s online art gallery which today represents over 20,000 emerging and established artists (Bluethumb, 2023). I, myself, having the passion for art, stumbled upon their website which was very well made indeed. But as a new artist and user myself, there were some points I realized it lacked. They have an enormous customer base and an even larger artwork portfolio. Data science could revolutionize the way Bluethumb operates, if the website data could be utilized to provide insights into what are the current market trends and where do one’s artwork stands before the artist can start selling. This will enable emerging artists to increase the artwork sales potential by aligning with market trends whilst improving their own profiles to the level of those who are well established. For Bluethumb, this will not only increase their sales due to more artist engagement but also it will be a significant value add to their corporate social responsibility (CSR).
“Bluethumb Canvas Success Project” aims to provide insights on the market trends in the art industry within Australia to guide new artists to areas with high demand. Further summary insights into making better data driven decisions when it comes to sizing, texture, topic…etc. The artwork data and artists profile information coupled with these insights will be incorporated into developing an art growth score model to indicate an artwork’s potential to sell. The model initially will not take into account the image of the artwork itself but rather other variables that influence a buyer’s decision. For example, art style, topic, size, price, frame, artist popularity, follower, count..etc. It could be later extended to include image analysis in phase 2.
The objective of a growth score model is to provide a comparison with the artists who are selling and to highlight weak areas to improve on or to make better artwork wise decisions. As per Martin et al (2020), the moment the human brain encounters a mismatch between the goal and capacity it initiates a learning process. This development will provide new artists with a guide they can refer to in order to improve and focus their efforts into achieving the end goal of selling and becoming a more established artist. Whilst for Blue thumber it will drive more artist engagement which converts to better sales and increased CSR for the brand name. The combinations of advanced analytics for market analysis, artist comparison, propensity to buy mapped into an artist growth journey in a user friendly platform will be a novel data science initiative that stands out in the art industry.
Goal and Objectives
The goal of this project is to enable artists to drive growth which eventually converts to increase in sales. The project is developed particularly for Bluethumb. The objectives are briefly mentioned below:
Efficient use of resources As it will enable emerging artists to focus their efforts on making creatives in areas of market interest.
Drive growth As the artwork grade would change along with the artist profile and other variables which will act as a growth indicator.
Drive Artist Interaction The more an artist can see a quantified indication of results from their efforts and not just sales, the more motivated they are to continue interacting on Bluethumb.
Drive sales Support emerging artists into making their first sale and much more as they grow.
As of today, there are several challenges artists utilizing this online platform face, the project outcome on addressing them are shown below:
Business Model
Data Sources and Collection
The required for this project is already available on the Bluethumb website. When developing this project within Bluethumb it is a matter of tapping into their own website data utilizing APIs which will be enabled by the data engineer, data architect and the data scientist through collaboration. In order to develop a prototype the data is obtained from the website using a web scraper known as Octoparse. The format of the feed page is shown below where all published art is posted in the order it was published. The points that have been extracted are indicated by the red marker
Arts for sale page
Artwork page
Artwork details
Artist profile page
Octoparse Webscrapper
The data is extracted in three stages to speed up the process. Initially, the artwork url of all the artworks in the feed page is obtained after which, the workflow in the scraper is set to visit each artwork webpage and extract the relevant data. Finally, the url of each artwork is trimmed to obtain the artist page.
link from the extracted data and is fed into the scraper to visit each artist profile and obtain the artist data, after which it is stored in Artist_profile_data.csv. The three datasets are compiled into a single dataset named final_data.csv via excel and loaded into R Studio for data pre-processing and analysis. The octoparse workflows for each stage are shown below.
Data is extarcted from the
Bluethumb website and compiled into a single file named final_data.csv
and loaded into R Studio for data pre-processing and analysis. File
line: https://drive.google.com/file/d/1BMEo4PNLfArJddziqBzarlIscB4vpA7o/view?usp=sharing
Data collected summary
After the dataset was imported into R, the structure of the dataset was viewed via the glimpse() function . There are total of 17 columns in the dataset and 8,714 rows of unique artwork from which 253 are sold by now (Approx 3% conversion rate)
Data cleaning steps
# Install libraries
#install.packages("readr")
#install.packages("dplyr")
#install.packages("tidyr")
#install.packages("stringr")
#install.packages("visdat")
#install.packages("ggplot2")
#install.packages("wordcloud")
#install.packages("RColorBrewer")
#install.packages("wordcloud2")
#install.packages("tm")
#install.packages("tidymodels")
#install.packages("glmnet")
# Load libraries
library(readr)
library(dplyr)
library(tidyr)
library(stringr)
library(visdat)
library(ggplot2)
library(wordcloud)
library(RColorBrewer)
library(tm)
library(caret)
library(tidymodels)
library(glmnet)Load dataset
## Rows: 8714 Columns: 17
## ── Column specification ────────────────────────────────────────────────────────
## Delimiter: ","
## chr (16): title, size, price, artist, artwork, location, medium, hang, sold_...
## dbl (1): likes
##
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.## Rows: 8,714
## Columns: 17
## $ title <chr> "Waiting for the Airport Train", "'Viola Green Goddess…
## $ likes <dbl> 0, 1, 13, 0, 0, 0, 1, 10, 2, 2, 7, 0, 1, 3, 0, 0, 0, 4…
## $ size <chr> "122cm (W) x 91.4cm (H) x 3.8cm (D)", "92cm (W) x 102c…
## $ price <chr> "A$2,990", "A$1,250", "A$1,250", "A$290", "A$1,580", "…
## $ artist <chr> "Anna Mandoki", "Natalie Briney", "_ PEZ _", "Michael …
## $ artwork <chr> "15 artworks", "87 artworks", "158 artworks", "47 artw…
## $ location <chr> "Melbourne, Australia", "Margaret River, Australia", "…
## $ medium <chr> "Oil, acrylic, soil, bitumen and image transfer on can…
## $ hang <chr> "This artwork is currently stretched and ready to hang…
## $ sold_tag <chr> "free shipping and returns A$2,990 Add to Cart", "fr…
## $ sold <chr> "add_to_cart", "add_to_cart", "add_to_cart", "add_to_c…
## $ description <chr> "POLITICAL ART, PEOPLE & PORTRAIT ART, BIRD ART\n#bird…
## $ Page_URL <chr> "https://bluethumb.com.au/anna-mandoki/Artwork/waiting…
## $ artist_url <chr> "https://bluethumb.com.au/anna-mandoki", "https://blue…
## $ follow <chr> "9 followers | 1390 profile views", "10 followers | 13…
## $ featured_artist <chr> "featured-ico.6af4c6f5.svg", "featured-ico.6af4c6f5.sv…
## $ artwork_sold <chr> "Sold (1)", "Sold (37)", "Sold (98)", "Sold (11)", "Ar…Load dataset
## Warning: Expected 2 pieces. Missing pieces filled with `NA` in 2 rows [8505,
## 8628].## Warning: Expected 2 pieces. Missing pieces filled with `NA` in 3229 rows [4, 8, 15, 19,
## 22, 23, 28, 32, 35, 41, 42, 44, 45, 47, 48, 50, 51, 53, 58, 60, ...].## Warning: Expected 2 pieces. Missing pieces filled with `NA` in 303 rows [5, 12, 78, 89,
## 97, 132, 135, 179, 260, 264, 269, 335, 348, 399, 467, 492, 532, 635, 697, 717,
## ...].Load dataset
## Warning: There were 2 warnings in `mutate()`.
## The first warning was:
## ℹ In argument: `followers = as.numeric(str_trim(str_remove(string = followers,
## pattern = " follower")))`.
## Caused by warning:
## ! NAs introduced by coercion
## ℹ Run `dplyr::last_dplyr_warnings()` to see the 1 remaining warning.## Rows: 8,714
## Columns: 19
## $ title <chr> "Waiting for the Airport Train", "'Viola Green Goddess' …
## $ likes <dbl> 0, 1, 13, 0, 0, 0, 1, 10, 2, 2, 7, 0, 1, 3, 0, 0, 0, 4, …
## $ width <dbl> 122.0, 92.0, 170.0, 28.5, 112.0, 41.0, 75.0, 50.5, 194.0…
## $ height <dbl> 91.4, 102.0, 115.0, 41.0, 71.6, 30.5, 75.0, 40.5, 194.0,…
## $ diameter <dbl> 3.8, 4.0, 1.0, 0.3, 0.3, 0.4, 4.0, 3.5, 2.0, 3.0, 3.8, 0…
## $ price <dbl> 2990, 1250, 1250, 290, 1580, 630, 1600, 440, 1990, 2400,…
## $ artist <chr> "Anna Mandoki", "Natalie Briney", "_ PEZ _", "Michael Fe…
## $ artwork <dbl> 15, 87, 158, 47, 29, 25, 48, 69, 90, 48, 153, 29, 156, 6…
## $ location <chr> "Melbourne", "Margaret River", "Port Douglas", "Sydney A…
## $ medium <chr> "Oil, acrylic, soil, bitumen and image transfer on canva…
## $ hang <chr> "stretched and ready to hang.", "stretched and ready to …
## $ sold <chr> "add_to_cart", "add_to_cart", "add_to_cart", "add_to_car…
## $ category <chr> "POLITICAL ART, PEOPLE & PORTRAIT ART, BIRD ART", "NATUR…
## $ hashtag <chr> "birds, people, group, suitcase, travel, texture, light …
## $ followers <dbl> 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 23, 2…
## $ profile_views <dbl> 1390, 1390, 1390, 1390, 1390, 1390, 1390, 1390, 1390, 13…
## $ status <chr> "featured", "featured", "established", "0", "photograph"…
## $ artwork_sold <dbl> 1, 37, 98, 11, NA, 9, 15, 13, 11, 46, 79, NA, 110, NA, N…
## $ area <dbl> 11150.80, 9384.00, 19550.00, 1168.50, 8019.20, 1250.50, …Load dataset
Load dataset
## Rows: 8,714
## Columns: 19
## $ title <chr> "Waiting for the Airport Train", "'Viola Green Goddess' …
## $ likes <dbl> 0, 1, 13, 0, 0, 0, 1, 10, 2, 2, 7, 0, 1, 3, 0, 0, 0, 4, …
## $ width <dbl> 122.0, 92.0, 170.0, 28.5, 112.0, 41.0, 75.0, 50.5, 194.0…
## $ height <dbl> 91.4, 102.0, 115.0, 41.0, 71.6, 30.5, 75.0, 40.5, 194.0,…
## $ diameter <dbl> 3.8, 4.0, 1.0, 0.3, 0.3, 0.4, 4.0, 3.5, 2.0, 3.0, 3.8, 0…
## $ price <dbl> 2990, 1250, 1250, 290, 1580, 630, 1600, 440, 1990, 2400,…
## $ artist <chr> "Anna Mandoki", "Natalie Briney", "_ PEZ _", "Michael Fe…
## $ artwork <dbl> 15, 87, 158, 47, 29, 25, 48, 69, 90, 48, 153, 29, 156, 6…
## $ location <chr> "Melbourne", "Margaret River", "Port Douglas", "Sydney A…
## $ medium <chr> "Oil, acrylic, soil, bitumen and image transfer on canva…
## $ hang <chr> "stretched and ready to hang.", "stretched and ready to …
## $ sold <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,…
## $ category <chr> "POLITICAL ART, PEOPLE & PORTRAIT ART, BIRD ART", "NATUR…
## $ hashtag <chr> "birds, people, group, suitcase, travel, texture, light …
## $ followers <dbl> 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 23, 2…
## $ profile_views <dbl> 1390, 1390, 1390, 1390, 1390, 1390, 1390, 1390, 1390, 13…
## $ status <chr> "featured", "featured", "established", "0", "photograph"…
## $ artwork_sold <dbl> 1, 37, 98, 11, 0, 9, 15, 13, 11, 46, 79, 0, 110, 0, 0, 0…
## $ area <dbl> 11150.80, 9384.00, 19550.00, 1168.50, 8019.20, 1250.50, …Load dataset
check for outliers
Before Outlier Treatment
## `geom_smooth()` using formula = 'y ~ x'
Outlier Treatment
After Outlier Treatment
## `geom_smooth()` using formula = 'y ~ x'
Confusion Matrix
## price area height width artwork_sold
## price 1.000000000 0.64799132 0.61046742 0.602554969 0.03385989
## area 0.647991316 1.00000000 0.89376734 0.908309380 0.16601886
## height 0.610467420 0.89376734 1.00000000 0.704326149 0.11068518
## width 0.602554969 0.90830938 0.70432615 1.000000000 0.16311725
## artwork_sold 0.033859893 0.16601886 0.11068518 0.163117255 1.00000000
## followers 0.020284306 0.00519591 0.01827813 0.008518096 -0.02248001
## profile_views -0.001025278 0.05519091 0.03338166 0.048208127 0.16552747
## likes 0.077608513 0.05818019 0.05433367 0.057522276 0.11761012
## followers profile_views likes
## price 0.020284306 -0.001025278 0.07760851
## area 0.005195910 0.055190910 0.05818019
## height 0.018278126 0.033381663 0.05433367
## width 0.008518096 0.048208127 0.05752228
## artwork_sold -0.022480007 0.165527466 0.11761012
## followers 1.000000000 -0.090547033 -0.01017243
## profile_views -0.090547033 1.000000000 0.01479253
## likes -0.010172434 0.014792530 1.00000000scatterplot of price vs area across different artist status
## `geom_smooth()` using formula = 'y ~ x'
Price vs Artwork Area
## `geom_smooth()` using formula = 'y ~ x'
Artwork decription word cloud - Not sold portfolio
Artwork decription word cloud - sold portfolio
Artwork topic wise - Conversion rate
## word freq.x freq.y percent
## 11 city 14 147 0.09523810
## 1 aerial 11 186 0.05913978
## 30 patterns 38 683 0.05563690
## 21 interiors 21 396 0.05303030
## 16 fashion 6 139 0.04316547
## 37 space 5 116 0.04310345
## 3 architecture 11 267 0.04119850
## 7 body 7 181 0.03867403
## 15 fantasy 16 425 0.03764706
## 44 water 15 446 0.03363229
## 31 people 33 1015 0.03251232
## 34 portrait 33 1018 0.03241650
## 36 seascape 30 1034 0.02901354
## 4 beach 30 1037 0.02892960
## 24 men 4 145 0.02758621
## 12 culture 13 483 0.02691511
## 32 places 10 406 0.02463054
## 38 still 17 697 0.02439024
## 23 life 17 700 0.02428571
## 22 landscape 56 2367 0.02365864Medium word cloud- art portfolio
## Warning in tm_map.SimpleCorpus(corpus, content_transformer(tolower)):
## transformation drops documents## Warning in tm_map.SimpleCorpus(corpus, removePunctuation): transformation drops
## documents## Warning in tm_map.SimpleCorpus(corpus, removeNumbers): transformation drops
## documents## Warning in tm_map.SimpleCorpus(corpus, removeWords, stopwords("en")):
## transformation drops documents
Medium word cloud- sold art portfolio
## Warning in tm_map.SimpleCorpus(corpus, content_transformer(tolower)):
## transformation drops documents## Warning in tm_map.SimpleCorpus(corpus, removePunctuation): transformation drops
## documents## Warning in tm_map.SimpleCorpus(corpus, removeNumbers): transformation drops
## documents## Warning in tm_map.SimpleCorpus(corpus, removeWords, stopwords("en")):
## transformation drops documents
Medium conversion rates
## word freq.x freq.y percent
## 31 board 333 25 0.07507508
## 215 rag 224 14 0.06250000
## 60 cotton 345 18 0.05217391
## 151 media 469 17 0.03624733
## 88 framed 392 14 0.03571429
## 207 print 364 12 0.03296703
## 183 panel 200 6 0.03000000
## 155 mixed 482 14 0.02904564
## 212 quality 438 12 0.02739726
## 179 painting 550 14 0.02545455
## 168 oil 1599 31 0.01938712
## 186 pastel 223 4 0.01793722
## 284 wax 223 4 0.01793722
## 177 paint 1366 20 0.01464129
## 13 archival 343 5 0.01457726
## 2 acrylic 3879 49 0.01263212
## 251 stretched 1009 11 0.01090188
## 87 frame 379 4 0.01055409
## 185 paper 1381 14 0.01013758
## 275 varnish 299 3 0.01003344Scaling data between 0-1
## Rows: 7,637
## Columns: 8
## $ sold <fct> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,…
## $ likes <dbl> 0.000000000, 0.002450980, 0.000000000, 0.000000000, 0.00…
## $ area <dbl> 0.72393871, 0.60922105, 0.07579170, 0.52060509, 0.081115…
## $ price <dbl> 0.92675159, 0.37261146, 0.06687898, 0.47770701, 0.175159…
## $ artwork <dbl> 0.0065975495, 0.0405278040, 0.0216776626, 0.0131950990, …
## $ followers <dbl> 0.001006711, 0.001118568, 0.001342282, 0.001454139, 0.00…
## $ profile_views <dbl> 0.01046112, 0.01046112, 0.01046112, 0.01046112, 0.010461…
## $ artwork_sold <dbl> 0.0004852014, 0.0179524503, 0.0053372149, 0.0000000000, …Logistic model training
## # A tibble: 8 × 3
## term estimate penalty
## <chr> <dbl> <dbl>
## 1 (Intercept) -2.08 0
## 2 likes -3.11 0
## 3 area -1.30 0
## 4 price 1.21 0
## 5 artwork -3.08 0
## 6 followers -5.45 0
## 7 profile_views 22.3 0
## 8 artwork_sold -2.94 0## # A tibble: 6 × 4
## sold .pred_class .pred_0 .pred_1
## <fct> <fct> <dbl> <dbl>
## 1 0 0 0.871 0.129
## 2 0 0 0.907 0.0927
## 3 0 0 0.871 0.129
## 4 0 0 0.894 0.106
## 5 0 0 0.923 0.0766
## 6 0 0 0.910 0.0896model performance
## Truth
## Prediction 0 1
## 0 1492 25
## 1 2 9## # A tibble: 1 × 3
## .metric .estimator .estimate
## <chr> <chr> <dbl>
## 1 accuracy binary 0.982## [1] 0.8928571## [1] 0.5813953Derive probabilities for the full dataset using the model
Derive artwork growth level
## [1] (-0.000999,0.1] (0.1,0.2] (0.2,0.3] (0.3,0.4]
## [5] (0.4,0.5] (0.5,0.6] (0.6,0.7] (0.7,0.8]
## [9] (0.8,0.9] (0.9,1]
## 10 Levels: (-0.000999,0.1] (0.1,0.2] (0.2,0.3] (0.3,0.4] ... (0.9,1]## Rows: 7,637
## Columns: 20
## $ title <chr> "Waiting for the Airport Train", "'Viola Green Goddess'…
## $ likes <dbl> 0, 1, 0, 0, 0, 1, 10, 2, 7, 0, 1, 3, 0, 0, 0, 4, 0, 1, …
## $ width <dbl> 122.0, 92.0, 28.5, 112.0, 41.0, 75.0, 50.5, 90.0, 28.0,…
## $ height <dbl> 91.4, 102.0, 41.0, 71.6, 30.5, 75.0, 40.5, 120.0, 35.0,…
## $ diameter <dbl> 3.8, 4.0, 0.3, 0.3, 0.4, 4.0, 3.5, 3.0, 3.8, 0.3, 5.5, …
## $ price <dbl> 2990, 1250, 290, 1580, 630, 1600, 440, 2400, 440, 1260,…
## $ artist <chr> "Anna Mandoki", "Natalie Briney", "Michael Fernandes", …
## $ artwork <dbl> 15, 87, 47, 29, 25, 48, 69, 48, 153, 29, 156, 6, 3, 29,…
## $ location <chr> "Melbourne", "Margaret River", "Sydney Australia", "Top…
## $ medium <chr> "Oil, acrylic, soil, bitumen and image transfer on canv…
## $ hang <chr> "stretched and ready to hang.", "stretched and ready to…
## $ sold <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0…
## $ category <chr> "POLITICAL ART, PEOPLE & PORTRAIT ART, BIRD ART", "NATU…
## $ hashtag <chr> "birds, people, group, suitcase, travel, texture, light…
## $ followers <dbl> 9, 10, 12, 13, 14, 15, 16, 18, 19, 20, 21, 23, 24, 25, …
## $ profile_views <dbl> 1390, 1390, 1390, 1390, 1390, 1390, 1390, 1390, 1390, 1…
## $ status <chr> "featured", "featured", "0", "photograph", "featured", …
## $ artwork_sold <dbl> 1, 37, 11, 0, 9, 15, 13, 46, 79, 0, 110, 0, 0, 0, 69, 1…
## $ area <dbl> 11150.80, 9384.00, 1168.50, 8019.20, 1250.50, 5625.00, …
## $ artwork_growth <chr> "Level_02", "Level_01", "Level_02", "Level_02", "Level_…Visualizing model coefficent importance
References:
https://www.digitalocean.com/community/tutorials/normalize-data-in-r
https://www.datacamp.com/tutorial/logistic-regression-R
https://stackoverflow.com/questions/53357700/cleaning-a-column-in-a-dataset-r
https://towardsdatascience.com/create-a-word-cloud-with-r-bde3e7422e8a