ERP software development

ERP Software Built For Your Operation, Owned By You

We build the system your company runs on, from scratch. You own the code and the documentation. One question decides this purchase: what are you left holding when we are gone?

A packaged product is bought and then configured. A system built for your operation is a different purchase. The risk in it is rarely that the build goes badly. It is that you cannot leave the firm that did it. So our answer is written down, against a published standard. You can buy that standard and hold every firm bidding for this work to it, us included.

Who this is for

  • Operations

    You hold the whole operation in your head, and you cannot be ill at close.

  • Finance

    You reconcile three systems by hand and still cannot say what a job cost.

  • Owner or director

    You have already said a number out loud to a board or a bank.

Jobs live in one workbook, stock in another, invoicing in a third. The numbers are joined up by one person who is very good at it. That is not a failure of capability. The workbook is correct. It is unauditable, and it runs through a single pair of hands. Nobody else can check it, and that person cannot take a week off in the wrong week. If the number is wrong, it is wrong in their handwriting, and everybody knows whose handwriting it is.

There is usually a second thing, and it is rarely said in the meeting. Somebody in the room championed the last system. They sat through the demos, they signed the case, and it half-landed. Asking for another one means asking the same people to back the same judgement twice. A second demo moves nothing. What survives that conversation is something written down that the person who did not sit through it can check.

Where a month-end close lands: APQC, 2,300 organisations, calendar days

Put your own close against these three marks before you talk to anybody. The first and third are quartile thresholds rather than averages, so they mark the edges of the middle half.

Show data table
Calendar days from trial balance to completed consolidated financial statements.
Item Calendar days
Fastest quartile 4.8
Median 6.4
Slowest quartile 10
Calendar days to complete a month-end close, at the fastest quartile, the median and the slowest quartile. Calendar days from trial balance to completed consolidated financial statements. APQC Open Standards Benchmarking, General Accounting survey, reported 2018, 2,300 responding organisations.

Those three marks are APQC figures, from its Open Standards Benchmarking General Accounting survey reported in 2018. It covered 2,300 responding organisations. A close that sits past is usually slow because the numbers live in several places. Somebody then has to make them agree by hand.

What a wrong system does to a closeroughly ten times
baseline

4 days

Alleged normal close

alleged

43 days

Alleged first close after go-live

Alleged in one complaint, at one organisation, for one close cycle. It is not a rate for the category and we do not offer it as one.

What a wrong system does to a close (working days)
Optionworking days
Alleged normal close4 days
Alleged first close after go-live43 days

In its 2017 complaint the company alleges that after going live on a new ERP in November 2012, its first monthly close took 43 working days against a normal 4. United States District Court for the Eastern District of New York, No. 1:17-cv-06997, complaint filed 30 November 2017, paragraph 76.

That is one organisation and one close cycle, from a complaint that has not been proven. The close is where a wrong system of record shows first. It is the one moment when every part of the operation has to agree with every other part. What the filing gives you is a size you can check yourself, from a court record.

Panorama's 2018 survey of 237 organisations: how much of a packaged ERP product they modified

Self-reported share of the application's code modified, in the publisher's own bands.

Show data table
Self-reported share of a packaged ERP product's application code modified. Share of respondents in each band.
Item Self-reported share of respondents (%)
No customisation 11%
1 to 10% 10%
11 to 25% 33%
26 to 50% 37%
Over 50% 8%
Completely custom, in-house or best-of-breed 1%
Self-reported share of a packaged ERP product's application code modified. Share of respondents in each band. Self-reported share of a packaged ERP product's application code modified. Share of respondents in each band. Panorama Consulting Solutions, 2018 ERP Report, n=237 valid responses.

Of 237 organisations Panorama surveyed in 2018, the largest group reported modifying 26 to 50 percent of the application's code. Somebody maintains that difference for as long as the system runs.

The other half of the picture belongs beside it. In Panorama's 2023 report, from 183 organisations, more than a quarter implemented their software without any customization. The publisher's stated reason: "This lack of customization is possible because vendors' pre-built industry best practices are becoming much more tailored to the unique needs of certain industries. Also, vendors are making configuration easier, lessening the need for customization." The two reports cover very different populations. They do not form one trend line.

A packaged ERP product can hold your data. What comes with it is a way of working. Either the system changes to fit your process, or your process changes to fit the system. Reshaping the business around the software is the cost nobody prices at the start.

Our own reading of that, and it is ours rather than the publisher's, is this. Where an industry has a vendor template that already matches how the work is done, the gap closes. Where the way you run the operation is what you compete on, you pay for the gap twice. Once to buy it, and once to keep the difference alive.

If the market's ERP products fit your process, use them. We build for the companies they don't fit.

What happens to each of your processes

For every process you run today, ask one question. Is it a deliberate advantage, an accident of history, or something not worth putting into a system at all?

  • Encode as-is

    Choose this when

    It is a deliberate advantage, or a regulator or a contract fixes how it has to work.

    It costs you

    You carry that complexity forward, and every later change to the system has to respect it.

  • Change first

    Choose this when

    It is a workaround for a limitation that is about to disappear, or an accident nobody chose.

    It costs you

    Somebody's job changes before the system is built, and that conversation is harder than the software.

  • Leave it out

    Choose this when

    The volume is genuinely low, and it is cheaper to keep doing it the way you do it.

    It costs you

    It stays outside the reporting as well, and somebody has to remember that at close.

A system of record holds project cost, capital expenditure, procurement, inventory, job costing and compliance reporting. What it does not hold is the customer relationship. Pipeline, bookings and who spoke to whom are a different system. That is a system for the customer relationship, the pipeline and bookings rather than the operation behind them.

We ask about the awkward ones first. A variation order, retention, a part-received purchase order, and a cost code cross-charged between two jobs. The size of the work turns on two counts. How many of your processes get encoded as they are, and how many you change first.

Every change after that becomes a change note. Time, effort and cost are laid out before approval. It enters the plan only once you sign it off, confirmed in a meeting and in writing. Nothing in the build moves on a re-scope you have not explicitly accepted.

What a changeover has to answer

The questions in front of any business that has to change its system of record without stopping trading.

  1. The close today

    How the close really happens today, including the parts nobody ever wrote down, has to be on paper before anything is designed.

  2. The first slice

    Which slice of the operation is worth having first, and whose week it takes work off, is a decision you are in.

  3. Both systems live

    While both systems hold a version of the truth, somebody has to reconcile them, and against numbers you already trust.

  4. The day it changes

    There is a day when the old system stops being the truth. Who calls it, and on what evidence, gets written down.

  5. History that will not tie

    History that will not tie out has to be found and explained, then decided on rather than quietly rounded.

How much history has to come across, and how clean it is, is the other thing that moves the size of the work most. Years of jobs in a workbook with three spellings of the same supplier is a different piece of work from a clean database.

We will not put a number of weeks on any of this before we have seen how your close actually works. What gets agreed up front is which slice lands first, because that is the part you can feel.

The stack we build on is published, so your own people, or another firm entirely, can read it before you commit to us.

  • The source code, in your own repositories
  • Deployments running in your own cloud accounts
  • Accounts provisioned and the keys handed over if you have none
  • Architecture, deployment and CI/CD pipeline documentation
  • BRD and SRS at the depth the foundation needs
  • A change note for every change you accepted
  • Documentation kept current through delivery

On ownership, the commitment is this. "Every project starts with documentation: architecture, deployment, CI/CD pipelines, BRD, SRS, and change notes. Progress is tracked and measured throughout. If you choose to move the work in-house or to another partner, the handoff is a process, not a hostage situation. The freedom is yours from Day 1."

Code lives in your repositories and deployments in your cloud accounts. If you have your own Git hosting, cloud accounts and storage, we work inside them. If you do not, we provision them, set them up, and hand the keys over.

You do not have to take any of that on trust. ISO/IEC/IEEE 14764:2022 is a published international standard for software maintenance, and this handover is built to its Clause 9 list. The standard names clauses called Scope of maintenance, Designation of maintenance organization, Processes, Organization and maintenance activities, Training, and Maintenance records and reports. Its own definition of maintainability is the "degree of effectiveness and efficiency with which a product or system can be modified", which makes it a property of the product. Buy the standard and hold every firm bidding for this work to the same list, us included.

The limits, so you can weigh the rest. A maintenance plan does not make us your maintenance organisation unless that is separately agreed. Owning the source is not by itself the ability to change it tomorrow: somebody still has to understand your operation, and documentation does not supply that.

We have built ERP for construction and capital-expenditure businesses.

Where compliance reporting is part of what the system carries, how we scope and verify compliance work against a named standard before it ships is a separate piece of work, and what a vendor security review actually checks before an enterprise build is approved is worth reading before you sign anything.

Whether to build this at all

  • The way you run the operation is what you compete on, and fitting it into a packaged product would mean reshaping the business around the software.

  • You know which of your processes are deliberate and which are accidents, or you are ready to sit down and decide.

Do not build it

Saying this out loud costs us work, and it is the reason the rest of it is worth reading. If one of those is you, the honest answer sits somewhere else, and it is cheaper for you to find that out here than three meetings in.

Before you start an ERP build

The questions that come up on every one of these calls, answered here rather than held back for the call.

What makes an ERP build bigger or smaller?
Six things, and you can size your own build against them before speaking to anybody. How many of your processes get encoded as they are, against how many you change first. How many years of history have to come across, and how clean they are. How many other systems have to keep talking to it once it is live. How many people transact in it every day, against how many only read from it. Whether compliance reporting is one set of rules or several. And how much of a half-built system already exists for us to inherit, if any. On changes after the plan is agreed: every change request becomes a change note, with time, effort and cost laid out before approval. It enters the plan only once you sign it off, confirmed in a meeting and in writing. Nothing in the build moves on a re-scope you have not explicitly accepted.
What do we own, and could another firm pick it up?
The code lives in your repositories and the deployments in your cloud accounts. If you have your own Git hosting, cloud accounts and storage, we work inside them. If you do not, we provision them, set them up and hand the keys over. Every project starts with documentation: architecture, deployment, CI/CD pipelines, BRD, SRS, and change notes. If you choose to move the work in-house or to another partner, the handoff is a process, not a hostage situation. The freedom is yours from Day 1. Clause 9 of ISO/IEC/IEEE 14764:2022 is the published list a handover can be checked against, and you can buy the standard and hold any firm to it, us included.
Should we fix the system we already have instead of building a new one?
Sometimes yes, and we will say so. If the system already models your operation correctly and the pain is reporting, an integration or one broken process, repairing it is cheaper and less disruptive than replacing it. Replacement earns its cost when the model underneath is wrong, because everything built on a wrong model has to be worked around by hand for as long as it runs.
How do you know what our operation does? Nobody who sold us the last one did.
We sit through a close and write down how it really happens, then ask about the awkward parts first: a variation order, retention, a part-received purchase order, a cost code cross-charged between two jobs. Each process then gets one of three answers, encode it as it is, change it and then encode the new one, or leave it out of the system, and you are in the room when each of those is decided.
When do we see something working, rather than a milestone?
The first slice goes live in the operation, chosen so that something comes off somebody's week. We will not put a number of weeks on it before we have watched your close, because a figure given before that is a guess dressed as a commitment. What we will agree before anything is built is which slice lands first and what it takes off whose desk.

Tell us what your operation does

Describe how the work actually runs today, and we will tell you whether this is worth building or whether you would be better off buying a product.

Two outcomes here are good ones. In the first we agree this is worth building. In the second we tell you to buy a product and spend the money on using it properly. That happens, and it costs one conversation to find out which of the two you are.

Tirth Bodawala, Chief Technology Officer of Atyantik Technologies
Tirth BodawalaChief Technology Officer, Atyantik Technologies
  • A conversation about your operation

Where the numbers live, and who joins them up, is enough to start.

We reply within one business day. No newsletter and no sales sequence.

How we handle what you send is set out in our privacy notice.