Data Science Services Turning Data Into Decisive Business Intelligence

Our data science services cover the eight services listed above, from data science consultation and data wrangling to predictive analytics, time series forecasting, operational analytics, business intelligence, an AI centre of excellence and integration management, run through the six-stage approach the page sets out. The sections below define data science in the words of NIST, the US National Institute of Standards and Technology, draw where each service sits, compare it with business intelligence and data engineering, and cover validation, the ways to hire, what moves the cost, what you receive and what to ask.

On this page

• What data science is, in NIST's words
• Where each data science service sits
• Data science, business intelligence or data engineering
• The data science lifecycle, stage by stage
• Validating a model before it reaches production
• Ways to hire data scientists, compared
• What moves the cost of a data science project
• What you receive
• What to ask any data science consulting company
• Questions buyers ask about data science
• More about data and AI work at Sthenos

What data science is, in NIST's words

Four definitions from NIST SP 1500-1r2, the first volume of NIST's Big Data Interoperability Framework, frame the eight services above. Each is quoted or closely paraphrased from that volume.

Where each data science service sits

Six of the eight services act on data as it moves from your systems to a decision; data science consultation sets the strategy above that path, and the AI centre of excellence sets the frameworks and governance models around it. Figure 1 draws the path using the services' own descriptions, with NIST's life cycle stages underneath.

Data science, business intelligence or data engineering: which one the work needs

Three purchases are often confused. Our software development services page sells big data engineering and data science together, for the pipeline and the model behind a reporting layer, and this page's Business Intelligence service is the reporting layer itself. Our team augmentation page makes the same split for hiring: data engineer, analytics engineer and machine learning engineer are three different jobs however often one advert merges them.

The three disciplines, side by side

The data science lifecycle, stage by stage

The strategy cards above set out six stages, and they apply in this order. The entry point is the one in the contact section at the foot of this page: a call at your convenience, a short, bounded discovery scoped per engagement, and a costed roadmap before committing to a build.

• Business understanding and problem definitionObjectives, success metrics and constraints are agreed with business stakeholders, so every data initiative is tied to measurable outcomes.
• Data discovery and assessmentData sources, quality, availability and governance readiness are evaluated, and gaps, dependencies and risks found before modeling begins.
• Data preparation and feature engineeringData is cleaned, normalized and transformed, and relevant features engineered, so datasets are analysis-ready.
• Model development and validationStatistical and machine learning models are designed, trained and tested, and validated for accuracy, bias, scalability and real-world applicability.
• Deployment and integrationValidated models are deployed into production and integrated with existing systems, applications and workflows, with a focus on reliability, security and low latency.
• Monitoring, optimization and governanceModel performance, data drift and operational impact are monitored after deployment, under governance frameworks for compliance and transparency.

Validating a model before it reaches production

Stage 4 validates models for accuracy and bias, and stage 6 monitors them after deployment. For the machine learning models among them, NIST's AI Risk Management Framework, intended for voluntary use, names seven characteristics of trustworthy AI systems, and says creating trustworthy AI requires balancing them based on the system's context of use. NIST notes that validity and reliability for deployed AI systems are often assessed by ongoing testing or monitoring, and that accuracy measurements should always be paired with clearly defined and realistic test sets. Our AI and machine learning page sets out the framework's four functions.

Ways to hire data scientists, compared

The services heading above names expert data scientists, and NIST's definition of the role spans business needs, domain knowledge, analytical skills and software and systems engineering. When you hire, say which of the three data roles you need, per our team augmentation page, then choose one of the four engagement models our software development services page publishes.

What moves the cost of a data science project

Data science work is billed at the published Sthenos rate band, and where the scope is clear we quote a fixed price for a defined outcome, with a costed roadmap before any build starts.

Within those bands, these are the questions that change an estimate. Each comes from this page or from our custom software development and AI and machine learning pages, and none carries a number here.

• The state of the data. AI projects stall on data, not only on models: inventory, access, quality, lineage and the pipelines that keep a feature fed after launch. Stage 2 finds the gaps before modeling begins.
• How settled the scope is. A written, agreed scope is a large variable in any estimate, and an unsettled one is the reason estimates move.
• The integration count, and whether the other side is documented. An integration into a modern documented API is routine; one into an undocumented system is a project of its own.
• The compliance evidence the contract requires. Producing the artifacts an auditor or a contracting officer will ask for is work, and it belongs in the estimate.
• What happens after launch. Monitoring for model performance and data drift is stage 6 of this page's approach. Get it priced up front.

What you receive

A process description says what a vendor does. This says what you end up holding.

• A data strategy aligned with business objectives, analytics and machine learning.
• Objectives, success metrics and constraints agreed before any modeling.
• Datasets cleaned, normalized and analysis-ready, with engineered features.
• Models validated for accuracy, bias, scalability and real-world applicability.
• Dashboards and reporting systems with real-time visibility into business performance.
• The repository and cloud accounts in your name, and monitoring of model performance and data drift after deployment.

What to ask any data science consulting company

These questions apply to all the data science consultants on a shortlist, Sthenos included. Each comes from NIST or from a question our own pages already publish.

1. Which of the three roles are you staffing: data engineer, analytics engineer or machine learning engineer?
2. What will the model be validated against, and how is bias measured? Our AI and machine learning page names a labelled set rather than a demo.
3. Who monitors model performance and data drift after launch, and what happens when a model degrades?
4. Where are the engineers? Data residency, contract terms and your own compliance review all depend on the answer.
5. What is explicitly out of scope? The excluded half is the half that generates change requests.
6. Who owns the code, the repository and the cloud accounts when we are done?

Questions buyers ask about data science

What is data science?

NIST SP 1500-1r2 defines it as the methodology for the synthesis of useful knowledge directly from data, through a process of discovery or of hypothesis formulation and hypothesis testing.

What does a data scientist do?

In NIST's definition, a data scientist has sufficient knowledge of business needs, domain knowledge, analytical skills, and software and systems engineering to manage the end-to-end data processes in the analytics life cycle.

What are examples of data science work a business can start with?

The services on this page give five: predictive analytics to forecast trends and identify risks, time series forecasting for demand, sales, inventory and pricing, operational analytics to detect bottlenecks, business intelligence dashboards, and data wrangling that makes datasets analysis-ready.

Do you offer data strategy consulting?

Yes. Data Science Consultation, the first service on this page, defines the right data strategy by aligning business objectives with analytics, machine learning and scalable data architectures, focused on practical use cases and long-term data maturity.

How much do data science services cost?

Sthenos rates are $150 to $250 per hour, depending on the seniority and mix of the team, and a typical project runs $50,000 to $200,000. Where the scope is clear we quote a fixed price for a defined outcome, with a costed roadmap before any build starts.

Who owns the code you write?

You do, entirely, from the first commit. The repository and the cloud accounts are in your name, with no licence back to us and no dependency on us continuing to exist.

More about data and AI work at Sthenos

Have a decision that takes too long? Tell us the data behind it and the question it has to answer, and you will get a straight answer on whether a model, a dashboard or neither is the right instrument. Talk to our engineers.