LeukosScience, data and education PTENES Contact us
Home / Data analysis
Data analysis consulting for companies

Find out what your numbers already know.

Sales, losses, costs, crop yields, soil tests: most companies keep data that has never been analyzed. We turn those spreadsheets into clear answers and better decisions, with the rigor of a PhD in computational research and the speed of artificial intelligence.

Entry point

The 80/20 diagnosis in 7 days.

In most businesses, a few causes explain most of the result: 20% of products bring 80% of revenue, 20% of failures cause 80% of losses. This is the Pareto principle.

You send the spreadsheets you already have. Within a week, we show you which few items weigh the most in your company, how much they represent and three concrete actions to start with, in a short report and a 30-minute call.

  • Fixed price and deadline agreed before we start.
  • No new software: we work with the spreadsheets and systems you already use.
  • Plain language: clear charts and conclusions, no jargon.
  • Confidentiality by contract: your data is never shared or reused.
Who it is for

Questions worth money.

Agribusiness and crop production

Why does one field yield less than another? Which fertilization pays off? What does the history of soil and leaf tests show? We combine yield, soil, fertilization, weather and variety data, and deliver the analysis to your agronomist, who makes the recommendation.

Growers, agronomic consultancies and cooperatives

Quality and processes

Why did losses increase? Is the process under control? Is the laboratory method reliable? Statistical process control, root-cause analysis and statistical validation of analytical methods.

Packing houses, food industry, dairies and laboratories

Retail and services

How much will I sell next month? Which customers are leaving? Where is cash tied up in inventory? Demand forecasting, ABC analysis and customer segmentation.

Distributors, retailers and service providers

Academic research

Statistical design and data analysis for experiments, theses and dissertations. We run the analysis and explain every result; the writing and conclusions remain the researcher's.

Graduate students and research groups
Toolbox

The right technique for each question.

Artificial intelligence speeds up the work. Choosing the technique and checking every result stays with a scientist.

Pareto and ABC analysisWhich few items weigh the most?
Customer segmentationWho are the best customers, and who is leaving?
Correlation, PCA and heat mapsWhat moves together with the result?
Decision treesWhich simple rules separate good from bad?
ANOVA and statistical testsIs the difference real or chance?
Design of experimentsHow to test with fewer plots and more certainty?
Statistical process controlDid the process go out of control? When?
Time seriesHow much will I sell, harvest or spend?
Predictive modelsMachine learning to estimate the outcome of each case.
Soil mapsHow does soil vary within the farm?
OptimizationWhat is the lowest-cost combination that meets the rules?
Risk simulationWhat is the chance of loss as prices and weather vary?
AI for textWhat do complaints and open survey answers say?
AI for imagesClassify or count from photos, when volume justifies it.
Sample deliverable

Losses at a vegetable packing house.

An illustrative example with fictitious data, showing the format of the 80/20 diagnosis report.

Pareto chart of losses by cause Bars show weekly losses by cause in descending order; the line shows the cumulative percentage. The first three causes add up to 78% of losses. 0100200300400 0%50%100% kg lost per week 80% 420260180110604030 TransportWashingSizeRotGreeningCrackingOther
Bars: kg lost per week for each cause. Line: cumulative percentage of losses. Total: 1,100 kg per week (fictitious data).

What the data shows

  • 3 of 7 causes add up to 78% of losses: bruising in transport (38%), breakage during washing (24%) and off-size product (16%).
  • Transport alone loses more than the last four causes combined.
  • Transport losses concentrate in loads stacked above 12 crates.

Three actions to start with

  1. Limit stack height on the truck and measure again in 30 days.
  2. Adjust water pressure in the washer and record breakage per batch.
  3. Sell off-size product through a separate channel instead of discarding it.
Formats

Start small and grow with results.

1. 80/20 diagnosis

7 days, fixed price. The few items that weigh the most and three concrete actions.

2. In-depth study

3 to 4 weeks. Causes, forecasts and scenarios, with the techniques the problem requires.

3. Monthly follow-up

A dashboard that updates itself and a monthly call about what changed and what to do.

We also offer team training so your staff can analyze their own data, in spreadsheets or Python.

Confidentiality and AI

Your data is your business's secret.

  • Non-disclosure agreement before we receive any spreadsheet.
  • Clear rules for AI: we only use services that do not train their models on your data, and the contract lists the tools involved.
  • Personal data protected: names, IDs and contacts are removed whenever the analysis does not need them, in line with data protection law.
  • Verified and reproducible: every number is checked, and each analysis step is kept so it can be rerun next month.
Who does the analysis

Dr. Washington de Almeida Pereira

PhD in Chemistry from the University of Brasília, with theoretical and computational research: principal component analysis, heat maps, correlations and predictive models with machine learning and neural networks, published in international journals.

He applied machine learning to predict carrot yield, studied phosphorus chemistry in soils and has programmed in Python since graduate school. He teaches chemistry and biochemistry in higher education and created the BaitaAjuda platform.

Academic CV (Lattes) →

First step

Send us a spreadsheet. We will show you what it hides.

Tell us the question that costs your company the most today and what data you already have. We reply with a proposal for the 80/20 diagnosis. We work remotely with companies anywhere.