There are many variables that influence the implementation phase. What is the breadth of the analytics need? What performance requirements are needed? How many business stakeholders will be targeted? What analytic features are a priority? These are just a few questions that can be targeted based on established objectives.
Blue Orange works with clients on each step of this process. We have worked across a wide range of tools, data sets, and projects, and we aim to identify the correct solution for each client. It’s helpful for our planning team to understand the business climate so we can take this into account when we’re generating a roadmap. It’s common that business requirements will shift standard implementation plans. Fundamentally, we aim to bring transparency and clarity to data projects.
Step 1: Discovery
Before we can provide a project estimation, we first need to understand your business and the key challenges that can be improved by modern analytics. This discovery phase includes detailing business objectives and current limitations. This also consists of an audit of how decisions are currently made. A successful analytics project requires tight integration with key decision makers.
In the initial Discovery meeting, we like to start broad, then hone in on project specifics. Key initial topics include:
- High-level opportunities that can be enabled by data and analytics
- State of Data: Existing data vs external data requirements
- Data structure and performance requirements
- Project roles, responsibilities, and ownership
- Data migration and integration
- Security & privacy
- Regulatory and licensing
- Project communication
Step 2: Strategy
A well-scoped project ideally has a set of actions that the organizations are taking now that can be better informed using data science. This isn’t limited to improving existing activities. Often, we end up creating a new set of actions as well. Generally, it’s a good strategy to first focus on informing existing actions instead of starting with entirely new measures that the organization isn’t familiar with implementing. Enumerating the set of actions to directly inform will allow the outcome of the project to help the organization achieve its goals.
Enumerating the set of actions to directly inform will allow the outcome of the project to help the organization achieve its goals.
Once we understand your objectives, resources and the current state of data, we can put together a project strategy. This initial outline will try to take into account the business needs identified in the Discovery phase. We build our strategy in stages leaving room for evolution in the project. As with most analytics projects, findings in each staging will have to impact the proceedings.
It’s important to note that accelerating certain features to accommodate business goals, though often necessary, can accrue technical debt. As we identify these potential pitfalls, we’ll work with the Project Manager to outline the challenges and to create a plan that balances technical hurdles.
Once drafted, we’ll review this strategy with stakeholders to ensure we’ve captured objectives outlined in the Discovery.
Step 3: Estimation
Data science and analytics services are high-value investments whose costs and benefits can be difficult to estimate in advance. The primary factors we consider when providing clients with project estimates are as follows:
- Quantity: How much data is available? External enrichment requirements?
- Format: What is the current state data? How many resources are to be consumed?
- Collaboration: The skillset and amount of internal capability already present
- Analytical Complexity: What is the analytical detail and complexity you’re looking for?
- Model Complexity: Does the project require advanced algorithms (AI, Machine Learning, etc.) or traditional analytic techniques?
- Integration: How will business stakeholders consume insights?
- Support: What ongoing needs are required?
- User Breadth: How many teams/users will require?
Step 4: Implementation
Once the planning phase has been completed, we’ll outline implementation details. This includes:
- - Project Management tools and process
- - Internal communication methods
- - Update cadences and format
- - Access and infrastructure
- - Request workflow and approvals
- - Integration and licensing
- Analytics Dashboards
- Data warehouse creation and Data cleaning projects
- Predictive models and AI & Machine Learning
- Marketing Automation
A typical analytics project results in some useful findings, and an analytical interface takes those findings and operationalizes them throughout the business. The creation of such an app can touch many different parts of a business, ranging from the front-line to logistics, finance and more. We build both custom analytics applications as well as integrate into existing visualization tools (Tableau, Power BI, Quicksight, etc.). Depending on the project requirements, we help clients identify the best solution that both answers specific business questions while being flexible enough to allow for future inquiry. At project inception, it’s preferable to scope and detail specific analytics needs to determine the right data solutions. Analytics projects are iterative. Initial tracking and analysis should not be delayed while waiting for a final, more elaborate analytics tool.
A few factors that determine the price of an analytical dashboard:
- What analytic features we’re looking to track?
- What integrations or third-party resources are required?
- Scope of user interfaces and analytic complexity required?
- Data update frequency
- User provided data requirements
Project estimation: ~$100k and up.
Data warehouse creation and Data cleaning projects
Factors that determine the price of creating a data warehouse:
- the number of data sources you want to include
- the messiness of the data
- the level of involvement and experience of your team
- what you’re trying to do with the data warehouse
- how much logic you’ll need to include to get the result you want
An energy company wanted unified access to disparate internal data sources and then wanted to enrich the data with third-party weather data. The project involved dozens of massive data sets updated by the millisecond. Teams using this streaming data warehouse included product, internal data science, and IT. This system had to be fault tolerance with high availability to support operating teams.
Project estimation: ~$150k and up
Predictive models and AI & Machine Learning
Machine Learning and AI is an incredible tool when applied to specific and scoped problems. On larger analytics projects, these problems tend to reveal themselves as general analytics projects progress. Essential in any Machine Learning modeling is a well-scoped problem set with clearly defined objectives.
Factors that determine the price of a predictive model or AI application:
- How much data wrangling will be involved
- Computing requirements
- The complexity of the algorithms (Supervised vs Unsupervised, etc.)
- How much change will need to occur within the models and/or AI application over time
- Problem complexity
- How does it need to be operationalized
Predictive model example:
Using a company’s customer spending habits, we identified the best times, channels, and offers to attract customers to re-engage with the brand. The cross-channel outreach increased revisiting customer conversion by 6% and reduced overall spending.
Project estimation: ~$70k
Artificial Intelligence example:
We worked with a startup to build a recommender to process and analyze resume data and find skill matches to optimize internal talent. The Blue Orange team workshopped with the product team and built the prototype recommendation engine that their infrastructure team then rolled out across the platform. The machine provided recommendations through a custom built API.
Project estimation: ~$60k
Factors that determine the price of marketing automation help:
- How much data will we be working with and from how many sources
- Determine complexity of marketing automation (Reduce CPA, increase segmentation conversion, identify new verticals, etc)
- The current state of available customer data
- User base size (How many people need to be trained and their expertise levels?)
Blue Orange worked with a client in the event ticket space that had online and email functionality. They engaged us to improve customer targeting through segmentation and cross-sell events to existing theatergoers. We built a data-science pipeline that supported the marketing department data analysis and integrated third-party market automation software.
Project estimation: $25k per month