Bespoke CRM

In three years, when this CRM has to do something different, who changes it?

You are weighing a product you could start on Monday against a system built around how your company works. Price and duration get discussed. Neither is what gets regretted. One branch ends with nobody able to maintain what was built. The other ends with nobody willing to change what was bought. That question already has a written answer on both sides, and you can read both before you sign.

Talk to a technical person

Tell us what your process has to hold. We reply within one business day, with a technical person on the call. That conversation ends one of two ways. We come back with a scope, or we tell you that a product on the market already fits and you buy that instead.

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

The held date

One object in your process decides this, and a one-minute test tells you whether a product can hold it.

Somebody has asked for the fourth of next month. You have said yes for now. Nothing is signed and no deposit has moved. To everybody else, the date is simply gone. It is not a deal, not a contact and not a calendar entry. You can make a product hold it. What that costs is configuration, and somebody owns that configuration every time the process changes again.

Left unconfigured, it lands wherever it fits worst. Then count where the same booking lives: enquiry, hold, proposal, deposit, event, invoice. Somebody at the desk assembles the whole guest by hand, across screens, every time the phone rings.

Here is a one-minute test for any product you are looking at. Name the object your process has that nobody else has. Open the product and try to store it without using a notes field. If the notes field is the only place it fits, the product cannot hold your process. It can hold a description of it, and your team spend their day translating between the two.

The quieter half of this is worth saying once. Whoever signed for the last system is the person who has to account for the next one.

The received wisdom is that a project like this runs long and runs over. A study of 4,677 IT projects says something more useful.

Show data table
Group sizes: 1,612 ERP, 459 HRM, 216 MIS, 684 SCM, 1,706 other. The unit is a ratio to the original estimate rather than a percentage, so 1.0 is on budget. The dataset records no CRM category and no split by company size.
Dimension Median Mean
ERP 0.9 1.3
HRM 1 1.5
MIS 1 1.8
SCM 1 2.2
Other 1 2.2

The middle project of every type lands on or under its estimate. The average is dragged up by a small number of very large overruns.

Actual cost against estimate on 4,677 IT projects, by project type: median beside mean. Flyvbjerg and colleagues, Journal of Management Information Systems, 2022. Group sizes: 1,612 ERP, 459 HRM, 216 MIS, 684 SCM, 1,706 other. The unit is a ratio to the original estimate rather than a percentage, so 1.0 is on budget. The dataset records no CRM category and no split by company size. Flyvbjerg, Budzier, Lee, Keil, Lunn and Bester, Journal of Management Information Systems, vol 39 no 3, 2022, Table 3. Population: 4,677 IT projects completed 2002 to 2014 across 66 countries.

Two things the chart cannot carry, both from the same study. Overruns and underruns are about equally frequent, so an estimate is roughly as likely to be beaten as missed. And terminated projects are absent from the dataset, so these figures are a floor on the risk.

That changes where your attention is worth spending. There is little to win by arguing an estimate down before anything has been written. What is worth having in writing is what happens on the day the estimate turns out to be wrong.

Who is designated to maintain it, and do they stay with it once it ships

Whether a system survives the people who built it is arranged during the build. There is an international standard with a clause titled Early involvement in development, and the title takes a minute to look up.

It is ISO/IEC/IEEE 14764:2022, Software engineering. Software life cycle processes. Maintenance, issued jointly by ISO, IEC and IEEE. It carries a clause numbered 8.7 and titled Early involvement in development. The standard is purchasable rather than free, so the clause titles are what you can check without buying it. It is a maintenance standard rather than a handover standard, which is worth knowing before you go looking.

So you have one question for everybody bidding, and it works on us too. Who is designated to maintain this after launch, and do they stay with it once it ships? A firm that has not thought about it will answer with a document they promise to write later.

These systems outlive everybody. In July 2025 the US Government Accountability Office reviewed 11 of the federal government’s most critical legacy systems. They ran from about 23 to 60 years old, median 41. GAO picked those 11 as worst cases, so read them as the far end rather than the norm.

Our own answer starts with people: the team that built the system supports it after launch. The people who did the work continue with it, rather than handing it to a separate support group who did not. The plan below is where what outlives us gets written down.

The other branch has already published its answer, in a document almost nobody opens before buying.

Whichever product you are considering, find its support lifecycle page before you sign. It sets out the phases the product moves through and what a customer can still ask for in each one. Read the row for the phase you would be buying into.

Two things about what you will find there. The first is that it is a reasonable document. No vendor with very large numbers of users can take design requests from everybody and hold one product together, and security fixes running to the end of the line is the right call.

The second matters more to you. Check what those phase dates are anchored to. Where they run from the product’s own release, the clock has been running since before you bought, and a product in its fourth year has already spent four years of it. That is the year-three question, answered on the other branch, in the vendor’s own words, for free, today.

Tell us what your process has to hold

List the objects your process has that nothing on the market stores properly, and the systems that have to keep working afterwards. We reply within one business day, with a technical person on the call. The conversation ends one of two ways. We come back with a scope, or we tell you that a product on the market already fits and you buy that instead.

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

What actually happens, in order

The order is set by your data and your calendar, not by a stack of documents.

  1. We write down what your process actually has

    We sit with whoever runs the day and write it in plain English, object by object. Nothing gets built until you have read that back and told us it is right.

  2. We draw the line between what is written and what is plugged in

    Email delivery, calendar, telephony, payments and accounting are usually connected rather than built. You approve where that line sits before anybody starts.

  3. Your history moves, and how is part of the scope

    How much comes across, whether it moves in one pass or runs in parallel, and when the old system stops. Those are decisions you make in the scope, not defaults we set.

  4. It goes live against your calendar

    You name the window it has to be working in. The plan is built backwards from that window rather than forwards from a start date.

  5. What outlives us is written down, not remembered

    Architecture decisions go into the technical requirements specification. Every change is reviewed by a senior software engineer and CI gates every merge. That is what lets somebody who never met us change this in year three.

Connecting the systems you already run is its own piece of work, and it has its own page. If what you need is bespoke and is not a CRM, start with custom software development instead.

Accommodation businesses in some markets earn most of the year in a few months, and that shape decides what a delivery date is worth to you.

Show data table
Croatia, accommodation businesses, 2024, not seasonally adjusted. The unit is an index against the 2021 monthly average, not money.
Item Net turnover index, 2021 monthly average = 100
Jan 23.8
Feb 30.8
Mar 38.1
Apr 139.1
May 204.2
Jun 223.3
Jul 473.2
Aug 530.1
Sep 286.2
Oct 107.6
Nov 49.1
Dec 53.4

August carries 24.6 percent of the year and runs 22.3 times the quietest month. Both figures are our arithmetic on the Eurostat points shown.

Croatia, accommodation businesses, 2024: monthly turnover index against the 2021 monthly average. Eurostat. Croatia, accommodation businesses, 2024, not seasonally adjusted. The unit is an index against the 2021 monthly average, not money. Eurostat, monthly services turnover, accommodation (NACE I55), unadjusted, 2021 = 100, series sts_setu_m.

Concentration belongs to your own market rather than to hospitality generally, so that shape is not automatically yours. Take the same Eurostat series in the same year. The peak month carries 24.6 percent of the year in Croatia and 20.6 percent in Greece. In Italy it is 14.4 percent, in Spain 14.0 percent. Italy also peaks in September rather than August. Those four shares are our arithmetic on the published points, not a Eurostat finding.

Which is why we ask for your window instead of offering you ours. A system that arrives after your season has opened has missed the year. Tell us the month you cannot be behind, and the plan gets built against that month.

What is true from day one, in writing

Forward these five to whoever signs. Not one of them depends on us still being here.

For whoever signs

Five things you can forward

  • The code is yours From day one every line of code is your IP. 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.
  • Nothing moves on a change you have not accepted Every change goes through a documented change note, with time, effort and cost laid out before approval. Nothing in the build moves on a re-scope you have not explicitly accepted. That holds during the build and afterwards on an AMC or lean-mode arrangement.
  • Every change is reviewed Code review by a senior software engineer, CI gating every merge, and architecture decisions written down in the technical requirements specification.
  • Leaving is a process If you move the work in-house or to another partner, the handoff is a process and not a hostage situation. That freedom is yours from day one.
  • What we do not do We do not certify against specific compliance frameworks ourselves. We engineer systems that pass your auditors’ requirements. That is a different thing, and you should know which one you need.

The CRM we have built, and what we will not tell you about it

The class of work is on the record. The rest is not, and the reason is worth more to you than the numbers would be.

We build bespoke CRM from scratch. It is designed and written for one company, rather than configured out of somebody else’s product. We have delivered it for hospitality businesses and for event-management businesses. That is where the held date, the guest who comes back and the enquiry-to-invoice chain all sit inside one operation.

Now the limit, and its two halves have separate reasons. Our clients are private by default, so we do not publish their names. Outcome figures are a different question: what we have on the record for this work is the class of it, and any number beside that would be one you have to take on trust. A number you cannot check will not survive the first question in your own board meeting.

That is the rule you can hold us to on everything above. Both charts carry a named publisher, a stated population and a unit. They come from outside and you can go and read them. Anything we say about our own work carries none of that, so treat it as our claim and test it in the conversation.

When this is the wrong call

  • Your process holds something no product on the market stores properly.

  • Several systems you already run have to keep working alongside the new one.

  • Somebody on your side can sit down and describe how the day actually runs.

Do not start here if

  • A product on the market already models the way you work. Buy it. A build earns its keep when your process holds an object no product does, and not before.

    Bespoke software that is not a CRM

  • You need somebody to configure and run a packaged platform you have already bought. We do not staff techno-functional roles for packaged platforms, and that is a different job from this one.

    What we do build

  • Nobody on your side can spend real time describing how the operation runs. The specification is reviewed in plain English with you before any code. That needs a person who knows the day.

    Talk it through first

  • You need certification against a named compliance framework. We do not certify against frameworks ourselves. We engineer systems that pass your auditors’ requirements, which is a different thing.

    Ask what we can and cannot sign

Before you start a conversation

The five that come up every time, answered in full.

Who owns the code, and what happens if we stop working together?
You do. From day one every line of code is your IP. 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. The accounts are yours either way. Move the work in-house or to another partner later, and the handoff is a process. It is not a hostage situation, and that freedom is yours from day one.
Can you configure and run Salesforce, HubSpot, Zoho or a similar platform for us?
What we build is bespoke. It is designed and written for your company, rather than configured out of a product somebody else designed. We do not staff techno-functional roles for packaged platforms. If you need a person to configure and run a product you have already bought, that is a different job. You want a different kind of firm for it.
How do we get a number out of you?
The number follows the scope, so the scope comes first. A form cannot produce either one. Four things move the scope, and you already know all four. Count the objects your process has that no product stores properly. Add the systems you run that have to keep working afterwards. Then the history: how much of it moves, and whether it moves once or runs in parallel for a while. Last, how much is written for you against how much is plugged in. After that, every change goes through a documented change note, with time, effort and cost laid out before approval. Nothing in the build moves on a re-scope you have not explicitly accepted. That holds during the build and afterwards on an AMC or lean-mode arrangement.
What happens if the people who built it leave?
The team that built the system supports it after launch. The people who did the work continue with it, rather than handing it to a separate support group who did not. Ask that of everybody bidding, including us: who is designated to maintain this after launch? There is an international standard, ISO/IEC/IEEE 14764:2022, with a clause titled Early involvement in development. It is a maintenance standard rather than a handover standard, and it is purchasable rather than free, so the clause titles are what you can check yourself. On our side, architecture decisions are recorded in the technical requirements specification, every change is reviewed by a senior software engineer, and CI gates every merge.
When can we reach you if something breaks after it is live?
Business hours are Monday to Friday, 10:00 to 19:00 IST. Within those hours, the response target is four working hours for P1. For P2 it is twenty-four working hours, and seventy-two for everything else. That target is acknowledgement by a named person who engages with the issue, not resolution. How long a fix takes depends on what the issue turns out to be. We do not commit to that in advance. These targets belong to a post-launch support arrangement on AMC or lean mode, rather than to the build.

Tell us what your process has to hold

The conversation this starts ends one of two ways. We come back with a scope, or we tell you that a product on the market already fits and you buy that instead.

Write down the objects your process actually has. A held date. A guest who comes back. A service history. An enquiry that turns into an invoice. Add the systems that have to keep working afterwards, and roughly how much history has to move.

Every requirement document is reviewed in plain English with you before any code is written. A build earns its keep when your process holds something a product cannot. If it does not, that is worth knowing before anybody writes a specification.

One conversation, and you will know which of the two branches you are on.

Tirth Bodawala, Founder and Chief Technology Officer of Atyantik Technologies
Tirth BodawalaFounder and CTO, Atyantik Technologies
  • If you leave, the handoff is a process and not a hostage situation.
  • Every line of code is your IP from day one, in your own accounts.
  • We reply within one business day, with a technical person on the call.

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