Aim of the project
Telecom wants to increase CAPEX intensive service distribution and wants to build business case, as well as sequence the investment based on cost/benefit. The aim is to determine which geographic regions to focus on first, because penetrating service is investment heavy. Both geographic conditions, customer demography, wealth and estimated demand (influenced by existing competition) are important factors in business case calculations.
Methodology
- Triple pay product distribution shall be analyzed along with factors that coexist in situations where product penetration is high
- Collecting information on real estate information. Most recent proxy for price can be crawled over the web , this will be the best indicator for dynamically changing real estate prices, -think economic crisis- after duplicate advertising is cleaned up. As a next step geographic databases and statistical sources on various dimensions of apartment information, cost of living, offering of competitors, shall be collected and centralized. These databases are available by national statistic offices, ministries, and economic research institutions. Finally, not only the geography but the structure of the real estate is a significant factor. Whether it is family house, residential condominium, or panel block of apartments are going to further differentiate the value and therefore expected demand
- Estimate the price points and expected demand from combination of competitive presence and other clustering information. Determine the dependency of price on competitors, as well as other significant factors
- Create decision tree to explain if Triple Play will be predicted to have high distribution, and at which price point (identify outlier values from current factbase and explain the differences from expectations, e.g. external factors such as past advertising)
- These methodologies are applicable also in countries where there are less number of transactions or there is a significant time lag between actual prices and publishing by statistics organizations
Figure: DBSCAN Differentiating between real estate categories
Some areas are mostly concrete-residential blocks, however with some landed house in between, as well as condominiums. A DBSCAN algorithm efficiently differentiates these areas from one another. This information along with latest advertising pricing crawled from web, past transactions and demographics from public datasources, as well as gathering regional competitor presence and price offering will contribute to accurate estimation of potential profitability of service introduction.
Results
- Profit scenarios after product introductions in function of mid and long term market demand evaluation
DOs and DON’Ts
Do
- All public data sources have to be aggregated
- Good data quality is recommended to ensure the best results
- Some data are must (basis area, plot area, location, etc.)
- Some data are nice to have (e.g. condition of the property, type of heating, etc.)
- Include household level demographic information
|
|
Don’t
- Don’t use existing legacy classification of counties, districts but more micro-segmentation variables
- Don’t use only street level information as nature of buildings can have serious impact for cost/benefit
|
Why we are different
- Approach is applicable in any geographic context, in combination of existing and public data sources
- We create a more accurate territorial clustering according to population density, type of real estate and geo-location, incumbent competitors, and therefore we also derive scenarios for future barrier to entry of potential competitors
|
