NP‑hard decision problems

The hardest decisions in industry are the same handful of problems.

What to load on which truck. Which jobs run on which machine. Where workloads go in a data centre. Who works which shift. Which projects get funded. Underneath the industry language it is the same hard mathematics — and it is where computers struggle most. heurics builds one engine for that mathematics, and applies it alongside the people who make these decisions.

What “NP‑hard” means in practice

Too many possibilities to ever check, and no shortcut to the best one.

Some problems get harder in a way that no faster computer can rescue. Add one more item and the number of possible answers doesn’t grow — it multiplies. Choosing 30 things out of 100, a small problem by industrial standards, already allows more combinations than there are grains of sand on Earth.

So nobody checks them all. In practice a business gets one of two things: an exact solver that can prove its answer is the best one, but only up to a certain size, or a rule of thumb that answers instantly and leaves value on the table that nobody can measure. Most real decisions are made in the gap between the two.

30 of 100a modest selection problem — a shortlist, a shift, a load plan
3 × 1025ways to make that one choice
~900,000 yrsto try them all at a trillion per second

Problems we’re exploring

Ten places a hard decision shows up every day.

These are the problems we are focused on today. Because the engine is general, the list grows with every conversation — if the hard decision in your work is not here, bring it to us.

Cloud & data centres

GPU cluster defragmentation

Capacity sits stranded in half-full machines, so new jobs wait while expensive GPUs stay idle.

Warehousing

Slotting and pick batching

Pickers spend much of their shift walking because fast-moving items sit in the wrong places.

Mortgage finance

Loan pooling

Closed loans are assigned to pools by hand or with simple tools, and small misallocations add up across every month.

Education

Course and exam timetabling

Thousands of sessions, rooms and students must fit together, and one change ripples through everything.

Sales

Territory alignment

Accounts have to be split fairly between reps, and every re-draw upsets someone’s patch.

Fixed income

Bond portfolios for small accounts

Lot sizes and minimums make it hard to build a sensible portfolio when the account is small.

Manufacturing

Facility and plant layout

Where each machine and station sits decides how far people and materials travel, every day.

Strategy & finance

Capital project selection

There are more worthwhile projects than budget each year, and the shortlist is usually built from rankings and spreadsheets.

Aviation

Airport gate and stand allocation

Flights must be matched to gates while delays keep rewriting the plan, hour by hour.

Telecoms

Tracking-area and cell-ID planning

Cells have to be grouped so that paging load stays manageable and handovers stay smooth.

The engine

One engine for the whole family, fast enough to use inside the working day.

heurics is a general-purpose engine for these problems. Rather than a bespoke model for each, the same engine takes the decision, the objective and the rules that cannot be broken, and searches for the best answer it can find in the time available — on ordinary hardware, not a cluster.

Answers you can act on

Every answer respects every hard rule. Stop the clock whenever you like and what is on the table is usable.

Scale past where solvers stop

The interesting sizes are the ones where an exact solver stops proving anything. That is the range we build for.

Speed that changes the workflow

An overnight run gets done once. A run that takes seconds gets done ten different ways before the decision is made.

Every product we build is this engine, with one industry’s language in front of it.

Benchmarks

Tested on the field’s own problems.

We test the engine on public benchmark libraries and on real published traces, across several problem families: cloud placement, selection, packing and scheduling. Each result is published in the research notes with its method and limits, once the run behind it has been validated end to end.

Read the research notes

Public benchmarks. Each comparison uses a test set, a protocol and a reference result published by others, so it can be repeated independently.

Let’s connect

Got a hard decision you make again and again?

If you make a hard decision again and again, in any industry, tell us how it is done today, the last time it went wrong and what that cost, and what you would do if a good answer arrived in seconds.

A reply by email is just as welcome as a call.

connect@heurics.com