Services
One system that tells you the truth.
Most of what follows attaches to the first item. We would rather do one implementation properly than sell five things badly.

ERPNext implementation and migration
The main work. Moving an owner-operated business off spreadsheets, a legacy package or a half-abandoned system, and onto one platform that runs orders, inventory, accounting and fulfilment together.
- Discovery first, in which we model what your business actually does rather than mapping your old tables onto new ones. Table-to-table mapping is the failure mode this exists to replace.
- Migration of the chart of accounts, price tiers, customer records and tax logic, reconciled to zero variance rather than to “close enough”.
- Parallel running, so the new system has to prove itself against the old one before anybody depends on it.
- Custom applications where the standard product genuinely does not fit, for example sell-by-the-pound pricing where the true weight is not known until fulfilment.
Operations and process
Finding the constraint that is actually binding. On the last engagement, a modelled requirement of 51 units a month fell to 12 once the evidence showed the limit was a scheduling decision on the floor rather than a supply problem upstream. Nobody had to buy anything.
Delivered as written procedures, onboarding material and in-person training, because a process that exists only in somebody’s head is not a process.
Analytics and econometric studies
A study over your own transaction history, aimed at one question you cannot currently answer: what a customer is worth, where margin is leaking, which price moves demand and which does not.
The last of these ran to a full written report, across years of transactions and a large family of model specifications. Every figure in it carries the query that produced it.
Scientific and research computing
Engineering work on simulation and research code: making it run, measuring what it costs, and checking that its numbers survive contact with their own data. We are engineers, not your scientists. The biology or the chemistry stays yours.
- Throughput and cost measurement. What a job actually costs per run and per core-hour, measured under the load it really runs at rather than on an idle machine. On our own published study the obvious backend was 2.6 times too slow and the machine was only 71 per cent busy under a naive queue.
- Standing up published code from a cold start, through the version drift and the undocumented environment assumptions, and reporting honestly what broke.
- Validation and failure detection across large batches: finding the runs that completed, wrote a well formed result file and are silently wrong. Two of ours did exactly that.
- Statistical review of a result before it is published, including whether the test being used is one the data qualifies for. We have retracted two of our own findings on that basis.
- Sizing a job before it is committed: how many runs are needed before the answer means anything, and what that will cost.
What that looks like in practice
The worked example is public and every figure in it carries its source: 500 protein folding simulations, what they cost, and the three times our own numbers turned out to be wrong.
If you have a simulation, a pipeline or an analysis you want measured, reproduced or checked, write to highlandwillowllc@gmail.com and describe the job. We will tell you if it is not something we should take on.
Websites and ongoing care
Small sites, usually five to seven pages, built to qualify enquiries before they reach you rather than to win design awards. Self-contained, fast, and handed over as files you own.
Optional monthly care covers updates, backups with a restore actually tested on a schedule, and monitoring. A backup nobody has restored is not a backup.
How an engagement runs
- First
- A call. If your existing setup is adequate, we will say so.
- Then
- A paid discovery, producing a written model of your operation and a scope with the unknowns named rather than buried.
- Then
- Implementation in reversible stages, each one dry-run first, with you able to stop at the end of any stage.
- Finally
- Handover: procedures, training, and where you want it, a support contractor we help you hire and develop.
Recurring care and the work we have done often enough to price in advance have published starting rates: see services and pricing. An implementation is quoted against a written scope instead, because the honest range is wide and depends on what discovery finds. Either way you will have a number before you commit to anything, and the cost of leaving will be in it.
Not sure it fits?
Say in one sentence what your system cannot tell you. If this is not the right work for you, that is a quick answer and a free one.