Machine Learning & Predictive Analytics
Predictive models built in Python around your own historical data, scoped after a review of what that data can support.
Custom quote after a short consultation · Timeline agreed in your quote
Overview
Machine learning turns the records you already hold into models that estimate what is likely to happen next: which customers may leave, how much stock a period may need, or which cases deserve a closer look. RF ETS builds these predictive models in Python, with SQL used to pull and shape the data where it lives in a database.
Because results depend so heavily on data quality, volume and the accuracy you need, we do not sell fixed packages for this service. Every project starts with a data review. We check what the data contains, whether the target you want to predict is realistic, and which modelling approaches are worth testing, then send a written scope and quote.
Once the scope is agreed, we prepare the data, engineer features, train and compare candidate models, and evaluate them on data held back from training. You receive the trained model, documented code, and an evaluation report that explains how the model performs, where it is weaker, and how to retrain it as new data arrives.
Who it’s for
- Businesses with several years of sales, customer or operational records that want forecasts or risk scores
- Operations teams that want to prioritise cases, orders or leads using patterns in past outcomes
- Product teams that need a prototype model before committing to a larger internal build
- Researchers and analysts who have a dataset and a prediction question but no time to build the model
Problems it solves
- Know which customers, orders or cases are most likely to need attention
- Replace rule-of-thumb estimates with forecasts based on your own history
- Find out early whether your data can support a useful prediction
- Get a model your own team can understand, rerun and retrain
What RF ETS delivers
- Data review summary with a recommended modelling approach
- Cleaned, feature-engineered training dataset
- Trained model with the comparison of candidate models
- Evaluation report covering performance on held-out data and known limitations
- Documented Python code or notebook with retraining instructions
How we work
Data review
You share a sample or full extract and the question you want answered. We assess quality, volume and whether the target is predictable.
Scope and quote
We propose the modelling approach, evaluation method and deliverables in a written quote before any build work starts.
Build and compare models
We prepare features, train several candidate models and compare them on data the models have not seen.
Report and handover
You receive the chosen model, code and an evaluation report, plus a walkthrough of how to use and retrain it.
What we need from you
- Historical data with the outcome you want to predict, or access to the database that holds it
- A clear description of the decision the model should support
- Notes on what each field means and how the data was collected
- A contact who can answer questions about edge cases during the build
How pricing works
This work varies too much for fixed packages. We scope it with you, then send a written quote with price, timeline and deliverables before any work starts.
Tell us what you need
Use the quote form or book a consultation and describe your goal, constraints and deadline.
We scope it with you
We review your material and agree deliverables, assumptions and what is out of scope.
Written quote
You receive a fixed price or milestone plan, timeline and revision terms before any work starts.
Not included
- Third-party costs: domains, hosting, paid plugins and themes, software licences, API or AI-model usage fees, data-provider credits
- Work outside the written scope agreed before work starts (handled as an add-on or a custom quote)
- Ongoing support after the delivery and launch-support window unless a support package is bought
Need something different?
Tell us what you need, and we’ll prepare a solution and pricing based on your requirements.
Machine Learning & Predictive Analytics: common questions
How is a machine learning project priced?
This service is quote-only. After a data review we send a custom quote based on data volume, preparation effort, the number of approaches to test and the deliverables you need. Nothing is charged for build work until you accept that quote.
Can you promise a certain level of accuracy?
No. Model performance depends on the signal in your data, its volume and how stable the patterns are. We report measured performance on held-out data honestly, explain the limitations and tell you early if the data looks unlikely to support a useful model.
What happens if the scope changes after we start?
Revisions within the agreed scope are included as set out in your quote. If you want to add a new target, extra data sources or a different deployment approach, we price the change as an add-on or an updated quote before doing the work.
Do you deploy the model into our systems?
The core deliverable is a trained model with documented Python code your team can run. Wrapping it in an API or connecting it to an application can be scoped separately, usually alongside our API development service, and confirmed in your quote.
How much data do we need?
There is no single number. It depends on how many outcomes you have, how often the event you are predicting occurs and how many useful fields exist. The data review exists to answer exactly this question before you commit to a build.
Often combined with
Data Analysis with Python & SQL
Data analysis with Python and SQL for business questions: descriptive analysis, segmentation, regression or forecasting, with charts and a report.
Custom quoteView service →Data & AnalyticsData Cleaning & Preparation
Data cleaning and preparation for spreadsheets and exports: de-duplication, standardising, merging files and a data-quality report.
Custom quoteView service →Research & Technical ServicesStatistical & Research Data Analysis
Statistical analysis for researchers and evidence-based reports: descriptive statistics, inferential tests, regression and a methods write-up.
Custom quoteView service →