re-cell-linear-regression

Context

The market for used and refurbished phones and tablets, once confined to a few online platforms, has expanded significantly in the last decade. According to a forecast by the International Data Corporation (IDC), the used phone market is expected to reach a value of $52.7 billion by 2023, growing at a compound annual growth rate (CAGR) of 13.6% from 2018 to 2023. This surge is largely due to increasing consumer interest in these devices, which offer substantial cost savings compared to new models.

Both individual consumers and businesses find that refurbished and used devices are an economical alternative for technology purchases. The benefits of buying used extend beyond just cost savings. These devices often come with warranties and can be insured with a proof of purchase. Retailers like Verizon and Amazon, among others, frequently run attractive promotions on refurbished items. Furthermore, purchasing second-hand devices contributes to environmental sustainability by extending the lifespan of technology, which aids in recycling efforts and reduces waste. The onset of the COVID-19 pandemic has likely accelerated this trend, as economic uncertainty has led consumers to reduce unnecessary spending and focus on essential purchases like phones and tablets for immediate use.

Objective

The growing promise of the somewhat overlooked market for used and refurbished devices is sparking interest in adopting machine learning to craft a flexible pricing strategy. ReCell, a startup eager to capitalize on this opportunity, has brought me on board as a data scientist. My task is to delve into the data they've gathered, construct a linear regression model to accurately predict the pricing of used phones and tablets, and pinpoint the key factors that impact these prices.

Data Description

The dataset features a range of attributes for used and refurbished phones and tablets, gathered throughout 2021. Below is a comprehensive data dictionary to help you understand the specifics of each attribute:

• brand_name : The manufacturer's brand name.
• os : Operating system the device runs on.
• screen_size : The diagonal size of the screen measured in centimeters.
• 4g : Indicates whether the device supports 4G connectivity.
• 5g : Indicates whether the device supports 5G connectivity.
• main_camera_mp : The resolution of the main (rear) camera in megapixels.
• selfie_camera_mp : The resolution of the front-facing (selfie) camera in megapixels.
• int_memory : The device's internal storage capacity in gigabytes.
• ram : The device's random access memory (RAM) in gigabytes.
• battery : The battery capacity in milliampere-hours (mAh).
• weight : The weight of the device in grams.
• release_year : The year the device model was initially released.
• days_used : The total number of days the device has been used.
• normalized_new_price : The standardized price of a new device of the same model, expressed in euros.
• normalized_used_price : The standardized price of the used or refurbished device, also in euros.

Skills

• Exploratory data analysis
• Data preparation for modelling
• Checking linear regression assumptions
• Performing linear Regression

Insights

• The model is able to explain ~84% of the variation in the data and within 4.6% of the normalized_used_price on the test data which is good.

• If screen size increases by 0.0217 units, used price increases by 1 unit, all other variables held constant

• If main camera size increases by 0.0229 units, used price increases by 1 unit, all other variables held constant

• If selfie camera size increases by 0.0130 units, used price increases by 1 unit, all other variables held constant

• If ram increases by 0.0172 units, used price increases by 1 unit, all other variables held constant

• If new price increases by 0.4277 units, used price increases by 1 unit, all other variables held constant

• If years since release increases by 0.0267 units, price decreases by 1 unit, all other variables held constant

• Price decreases if the OS is in the not windows, ios or Android

Key conclusions from our analysis

• Used devices gain more value when equipped with higher-spec components.

• Devices featuring larger screens, enhanced camera capabilities, and more RAM generally fetch higher resale values.

• The presence of popular operating systems adds to the resale value of used devices.

• Devices operating on well-known platforms such as Windows, iOS, and Android see higher resale prices.

• Brands like Nokia, Asus, and Xiaomi positively impact the resale prices of their devices.

• Devices supporting 4G connectivity generally exhibit higher resale values compared to those with 5G, likely because 5G technology was still relatively new and not widely adopted at the time the data was collected. This trend may shift as 5G becomes more prevalent in the future.

• Generally, the more time that has elapsed since a device's release, the more its value depreciates.

Recommendations for Business

1. Focus on High-Spec Devices: Prioritize acquiring and stocking used devices that are equipped with high-performance components. These include models with higher processing power, RAM, and superior camera functionalities. Promoting these features can attract buyers looking for premium devices at a lower cost.

2. Emphasize Larger Screens and Better Cameras: Highlight devices with larger screen sizes and enhanced camera capabilities in your marketing efforts. These features are significant selling points that can command higher resale values.

3. Stock Popular Operating Systems: Ensure a good inventory of devices that run on popular operating systems like Windows, iOS, and Android. These systems are more familiar to consumers and generally promise better user support, thereby enhancing resale value.

4. Leverage Brand Value: Capitalize on devices from brands like Nokia, Asus, and Xiaomi, which are perceived to hold their value better. Positioning these brands prominently can leverage brand loyalty and influence purchasing decisions.

5. Optimize 4G Device Sales: While the shift to 5G is ongoing, continue to offer and promote 4G devices, which currently have better resale values due to wider current usage and compatibility. Keep an eye on market trends to anticipate when to increase focus on 5G devices as they become more mainstream.

6. Manage Inventory Based on Device Age: Actively manage your inventory to avoid overstocking older models that depreciate faster. This might involve more dynamic pricing strategies to ensure older stock is sold in a timely manner, preventing significant loss in value.

7. Educational Marketing: Use educational content to inform potential customers about the benefits of purchasing used devices, especially highlighting how top-spec features on slightly older models offer great value compared to buying new.

8. Flexible Trade-In Options: Encourage customers to trade in their older models for newer ones. This not only ensures a continuous supply of used devices but also keeps your offerings fresh and more appealing.

9. Watch Market Trends for 5G: Stay updated on the adoption rates and network expansions of 5G. Be prepared to adjust your sales strategy to start emphasizing 5G devices once they begin to demonstrate widespread appeal and stability in resale value.