Skip to content
Mayrian

Predictive analytics

Forecasts and risk scores built from your history, with the uncertainty shown, so plans rest on evidence.

Start a Project Ask a question
historyforecast
Week 20 forecast
579
80% range
515–643
Backtest MASE
0.71

MASE below 1 means the model beat a seasonal naive forecast on past periods it hadn't seen.

Figure 1. Weekly demand with an eight-week forecast and its 80% prediction interval: the range expected to hold the actual value 80% of the time. Example data.
Start a project

What we build

Predictive analytics solutions

We build forecasts and risk scores from your history and deliver them, with their ranges, into the dashboards and planning systems your teams already use.

  • Demand forecasting

    Forecast sales, orders or bookings to plan inventory, purchasing and staffing.

  • Churn and retention risk

    Score customers or members by likelihood to leave, so retention efforts go where they matter.

  • Predictive maintenance

    Estimate failure risk from sensor and maintenance history to schedule work before breakdowns.

  • Capacity planning

    Forecast call volumes, appointments or workloads to plan capacity ahead of peaks.

  • Customer lifetime value

    Estimate each customer's future value to focus acquisition and retention spend.

  • Revenue and cash flow forecasting

    Forecast revenue, costs and cash by month for budgeting and planning.

Two analysts discussing a chart on a whiteboard

How to get started

Three steps to a plan for predictive analytics

  1. 1Tell us what you needUse the project form or book a call. A few sentences about the goal is enough to start.
  2. 2Free technical consultationWe go through your goals, users, existing systems and constraints with you.
  3. 3Your planA detailed plan covering the right tech stack, architecture, timeline and budget. Then you decide.
Start a project Ask a question first

How it works

A forecast you can check

Every forecast we deliver is tested on past periods and comes with a range, not just a number.

Monthly demand

  • Actuals
  • Backtest forecast
  • Forecast
  • 80% interval
  • 95% interval
BacktestForecast
Aug 2026BacktestActual 1,644 unitsBacktest forecast 1,684 unitsError 2.4%Move across the chart

Backtest. The model forecasts a past period as if it hadn't happened yet, and is scored against the actuals and a naive baseline.

Prediction intervals. The ranges actuals are expected to fall in, 80% and 95% of the time. They set safety stock and budget buffers.

Horizon. Uncertainty grows the further ahead you look, so the range widens month by month.

From a forecast range to a stock level

Drag the stock line, or choose a service level, to see how much stock covers demand that often and how quickly the cost rises.

Forecast 1,000Stock 1,2475% chance of a stock-out6008001,0001,2001,400
95%

Stock to hold

1,247

Safety stock

247

The forecast says 1,000 units, but demand could land anywhere in the curve. Holding stock at the chosen point covers demand that often. Each extra point of service level costs more stock than the last.

Our approach

How the work runs

  1. Discovery

    Agree what the forecast will change, at what level of detail and how far ahead.

  2. Data assessment

    Assemble the history and the drivers, such as seasonality, promotions, prices, holidays or weather.

  3. Model and backtest

    Build models and test them on past periods against a simple baseline, such as last year's figures.

  4. Deliver into planning

    Forecasts with their ranges, delivered to dashboards or planning systems on a schedule.

  5. Monitor and support

    Track accuracy over time and retrain as patterns change.

What we'll need from you

Having these ready keeps the work moving.

  • History and drivers

    Sales, volumes or events at the needed level of detail, plus calendars of promotions, price changes and closures.

  • Planners' time

    The people who use forecasts today, to explain adjustments they make and review backtests.

  • A decision owner

    Someone who agrees the horizon, level of detail and accuracy measure.

  • Target systems

    Access to the dashboards, ERP or planning tools forecasts should feed.

Who's on the project

Our team, working with planners from yours.

1, 1, 1, 2

1 Mayrian   2 Your organization

Data scientist: Builds and backtests the forecasting models.

Services

Predictive analytics services

  • Predictive analytics consulting

    Use-case selection, a data assessment and the forecast horizon and detail each decision needs.

  • Forecast and risk model development

    Models built on your history and drivers, backtested against a baseline before launch.

  • Integration into planning

    Forecasts written to your ERP, planning tools and dashboards on a schedule.

  • Monitoring and support

    Accuracy tracking, retraining as patterns change and support after launch.

Deliverables

What you receive

Appendix A. What you receive

All code, data and documentation are handed over in your accounts and repositories.

In your hands

What the documentation looks like

Every project ends with documents your team can run with. Here is an excerpt of one of them.

Backtest report

Weekly demand forecast, by SKU and store

Backtest over the last 26 weeks, against a seasonal naive baseline

Error by forecast horizon (WAPE)

HorizonBaselineModelForecast value added
1 week24.1%15.8%+8.3 pts
4 weeks29.7%19.4%+10.3 pts
8 weeks33.2%24.9%+8.3 pts

Bias

+1.2%: slight over-forecast, within the ±3% tolerance.

80% interval coverage

81% of actuals fell inside the range.

Measuring success

How success is measured

What we report on in predictive analytics projects. Which measures apply, and their targets, are agreed with you at the start.

Table 2. What we report, and when. Targets are agreed with you at the start.

MeasureReported
MASE, MAPE and WAPEBacktest, then every cycle
RMSEBacktest, then every cycle
BiasEvery cycle
Interval coverageEvery cycle
Forecast value addedBacktest, then quarterly
AUC for risk scoresBacktest, then monthly

Measure

MASE, MAPE and WAPE

Forecast error against a naive forecast and in percentage terms, reported by product, site or segment.

Reported

Backtest, then every cycle

How much a risk score is worth

A cumulative gains chart: drag along the curve to see how many likely leavers you reach by contacting the highest-risk customers first.

0%0%25%25%50%50%75%75%100%100%Random contactBy risk scoreCustomers contacted
20%

Contacted

10,000

Leavers found

51%

Lift

2.6×

Contacting the riskiest 20% reaches 2,041 of 4,000 likely leavers; at random, the same effort would reach 800. The curve shows where the retention budget is best spent.

Readiness check

Are you ready for
predictive analytics?

Five questions, about a minute. You'll see what to settle first and a sensible starting point.

Datasheet

Readiness for predictive analytics

Questions to answer about your data and organization before building

  1. Q1. Do you know which plan or decision the forecast will change, and how far ahead it's needed?

  2. Q2. Do you have at least two years of history at that level of detail?

  3. Q3. Are known drivers recorded, such as promotions, price changes, closures or stock-outs?

  4. Q4. Is there a current forecast or rule of thumb to compare against?

  5. Q5. Can forecasts be fed into the tools where planning happens?

Findings

0 of 5 answered

Answer every question to see the findings.

Start a project

How to start

From first call to production

Start where you are. Each step ends with a decision, so you commit to the next one only when it makes sense.

Protocol

How an engagement runs

Each step ends with a decision on whether to continue

Free technical consultation

One or two sessions

Talk through the goal, the data you have and the systems involved.

Outputs

  • (a) A shortlist of use cases, ranked by value and feasibility
  • (b) A recommended next step
  • (c) A plan covering stack, architecture, timeline and budget
Book the consultation

Estimate the value

What it could be worth to you

Enter your own figures. The formula is shown, and the estimate can go with your enquiry.

Estimate

Inventory carrying cost saved

Predictive analytics · from your own figures

Carrying cost saved = Inventory value × reduction × carrying cost rate(1)

Result

$160,000

a year

Add to my enquiry

Build or buy

When you don't need
a custom build

Part of the free technical consultation: when an existing product covers the need, we recommend it instead of a custom build. These are the options we weigh, alongside the tools you already have.

Related work

Existing products that may be enough

  1. [1]Forecasting in Excel (Forecast Sheet), Power BI or TableauWhen a quick trend line for a few series, where drivers such as promotions don't matter.
  2. [2]Demand planning modules in your ERP or planning suiteWhen your planners already use one and its statistical methods suit your patterns.

Our approach

When a custom build is worth it

  • Thousands of items or locations need forecasting at once
  • Promotions, prices, weather or events drive demand
  • You need prediction intervals and honest backtests
  • Forecasts must feed several systems on a schedule

In your industry

Where it applies

What this work delivers in the industries we serve.

Technologies and standards

Chosen for your project

Built on the cloud you already use. Choose yours to see the services involved; we recommend the full stack in the free technical consultation.

Table 2. Managed services for each layer, by cloud. The highlighted column is the one you chose.

LayerAWSAzureGoogle Cloud
History and driversAmazon RedshiftMicrosoft FabricBigQuery
Forecasting modelsAmazon SageMaker AIAzure ML automated forecastingBigQuery ML (ARIMA_PLUS), Gemini Enterprise Agent Platform
Scheduled runsAWS Step FunctionsAzure Data FactoryCloud Composer
Delivery to plannersAmazon QuickSightPower BILooker

Also runs on any of the three: Snowflake, Databricks, Prophet, statsmodels, LightGBM, APIs into ERP and planning systems.

Forecasting

  • statsmodels (ARIMA, ETS)
  • Prophet
  • scikit-learn
  • LightGBM
  • XGBoost

Data platforms

  • Databricks
  • Snowflake
  • Google BigQuery

Delivery

  • Power BI
  • Tableau
  • APIs into ERP and planning systems

MLOps

  • MLflow
  • Amazon SageMaker AI
  • Azure Machine Learning
  • Gemini Enterprise Agent Platform

Next section

Machine learning

Models that classify, score and recommend, documented and monitored.

Previous: Visualization & analytics

Data Analytics & AI

4 Machine learning

What should we do about each case?

Questions

Common questions
about predictive analytics

How accurate will the forecast be?

You'll know before relying on it. We backtest every model on past periods and compare it with a simple baseline, and the evaluation report shows the expected error for each product, site or segment.

How much history do we need?

We check it in the data assessment. For yearly seasonality we look for at least two full years of history; shorter horizons need less.

Can forecasts feed our ERP or planning tools?

Yes. We write forecasts to your planning and ERP systems through their APIs or imports on a schedule, and show them in dashboards.

How is a project priced?

Well-defined scopes are delivered as fixed-price engagements; when requirements are still evolving, we provide a dedicated team instead. Either way, the free technical consultation ends with a plan covering tech stack, architecture, timeline and budget, so you know the cost before work starts.

More in Data Analytics & AI

How it connects

All of Data Analytics & AI
Also in Data Analytics & AI:Generative AIComputer vision

Planning predictive analytics work?

Start with a free technical consultation: a detailed plan covering the right tech stack, architecture, timeline and budget for your project.

Start a project
A laptop showing a sales dashboard with bar charts