Kesey · Build versus buy
Should you build your own CRM? Build the tools, never the record.
Whether you are about to build one or already living with the one you built, the decision splits in the same place. The tools around your pipeline are cheap to build and getting cheaper. The record itself is the one thing you cannot afford to have made yourself.
Build the tools, never the record. Outreach, reporting, enrichment and the scripts that move data between systems are cheap to build and safe to throw away. The record, who your customers are and what was said and agreed, has to hold up for people who did not build it. A homemade one rarely does, and you find that out on the day a new hire, an auditor or a buyer needs to rely on it.
The build is the number people compare, and it is the wrong one. On the assumptions printed below, keeping a homemade record trustworthy runs about 256 hours a year, roughly six and a half working weeks, and none of those hours are the build.
Why I know this
I built one in 2006, and it lasted exactly until someone had to trust it
I helped build a homegrown CRM in 2006, because that is what you did then. It worked. It held the accounts, it held the notes, it did the job for the people who built it. It was migrated the day someone outside the team had to trust it, and not a day before.
Twenty years on, the tooling costs a fraction of what it did and trust costs no less than it ever did. That gap is the build versus buy question, and almost nobody prices it going in. You measure the build against a licence fee, win easily, and never put a number on the day your own system has to be believed by somebody who did not build it.
The line
Two halves, and only one of them should be homemade
The tools
Outreach and sequences. Reporting and dashboards. Enrichment. The scripts that move data from one place to another. These get cheaper to build every year, they are yours to shape exactly how you work, and you can throw any of them away next quarter without losing a thing that matters. Building them is not the mistake.
The record
Who your customers are. What was said on the call. What was promised, by whom, and when. It has to outlive whoever set it up and hold up for people who did not build it. A record earns that by staying correct when nobody is deciding to keep it correct, and a homemade one stays correct only while its author is still paying attention.
The mistake is not building. The mistake is building the wrong half. Teams pour their cleverness into the record, the part that should have been boring and bought, and hand the tools, the part worth building, to an off-the-shelf app that does not fit the way they work.
What changed
Three things that were not true in 2006, and you can check all of them
Each claim below is the vendor's own, read September 2026.
-
The record can now build itself from work that already happened
Day.ai creates companies, contacts and opportunities out of connected email and calendar, including history from before you signed up, so the workspace is populated before anybody types anything. In 2006 the only way to get data into a record was to make a person enter it, which is why every homegrown CRM eventually became an argument about compliance with your own form.
-
A bought record can tell you who changed a number, and when
HubSpot keeps a history on every field and will export it, so a number carries the person or automation that set it and the date. Being able to answer where a figure came from is what makes an outsider willing to believe it, and homegrown systems almost never can, because nobody builds a history for a database they already trust.
-
The licence maths that justified building has weakened
The strongest 2006 argument for building was per seat pricing: every person you added cost you again, so a one time build looked cheap against a bill that grew with headcount. That shape is no longer the only one on offer. Day.ai charges per agent rather than per seat, and Clarify prices against work rather than heads, so colleagues who only need to see the record do not each carry a licence.
Vendors move, and their own pages sometimes disagree with each other, so treat every line above as dated rather than permanent and check it against their documentation before you decide anything on it. The four CRMs compared carries the fuller version, each claim sourced the same way.
The number nobody quotes
What the record costs you every year after the build is finished
A model with every input printed. Not a survey of anyone's build.
Build versus buy gets argued on the build, because the build is the only number anybody puts on a quote. The build is a one-off. Keeping a homemade record trustworthy is not, and it is almost entirely people. Here is the arithmetic, with the assumption sitting under each line so you can overrule it.
- Keeping the schema current as the business changes 96 hours One engineer-day a month. Assumed because a record that stops matching how you sell stops getting used, so these changes do not queue politely, they get made.
- Reconciling the record against the systems that hold the money 104 hours Two hours a week. Assumed because billing, the bank and the pipeline disagree by default, and somebody catches it at month end rather than continuously.
- Answering where a number came from when somebody outside asks 32 hours One day a quarter, and assumed low on purpose. With no field history this is reconstruction from memory and commit logs, and the first time it genuinely matters it runs well past a day.
- Walking each new person through a system whose documentation is the people who built it 24 hours Four hours a hire, six hires. Assumed at a small team's rate of hiring. Double the headcount and double the line.
About six and a half working weeks of somebody's year, before anyone builds anything new. At a loaded $125 an hour it is $32,000 a year, every year. Use your own rate; the hours are the part worth arguing with. Not one of those four lines is the build.
These are our assumptions rather than a measurement of your team, so put your own numbers in. If yours come out materially lower, the arithmetic says keep the record you have, and that is a real answer rather than a polite one.
Two of those rows in full
- Where a number came from · the 32 hour row, the one we assumed low on purpose. What it costs the year somebody asks properly.
- When the person who built it leaves · the 24 hour row at its worst. A notice period buys you about 20 hours.
How you know
Four signs you built the record, not the tools
Answer these about your own setup. Nothing is sent anywhere and nothing is stored; the answer appears underneath.
-
Your reps route around it
Somebody wants to move a deal, so they tell an assistant to move it, and the required fields you built never get filled. A system people bypass in order to do their actual job has already lost. You cannot require your way to a record people trust; you can only build one they do not have to fight.
-
The numbers are fiction
Someone asks who the economic buyer is and the answer gets invented, because everybody knows nobody will ever listen back to the call and check. The pipeline now reads clean and means nothing, for two quarters, until a forecast misses and the room acts surprised.
-
It only works while one person feeds it
Every deal hygiene push works for about a week. Then management loses interest, the reps were never bought in, and the data quietly rots. If your record depends on discipline, it depends on the one week a quarter that discipline actually lasts.
-
You could not hand it to an outsider
The real test is not whether it works for you. It is whether you could hand it to a new hire, an acquirer, or a partner tomorrow and have them trust what it says without a tour. If the honest answer is no, you built a record only its author can read, and those are the ones that get replaced.
Four questions, and the answer is underneath
Answer them honestly, including the ones where the answer is no. A page that can only reach one conclusion is a sales pitch, so this one can tell you to leave your build alone, and for some of you it will.
If you have not built it yet
The version of this advice that saves you the migration
Most people reading this are earlier than the four signs above: the build is a plan, not a regret. The plan is usually right about the problem and wrong about the target. What is driving it is almost always one of three things, and none of them needs you to own the record.
Your object model does not fit. The business has a thing that is not a contact, a company or a deal, and the CRM makes that thing expensive or awkward. It is the best reason on this list, and it still does not mean owning the record. Either pick a record whose structure fits the business, or keep the unusual thing in your own system and have it point at the record instead of replacing it.
The tools are terrible. The sequencing is clumsy, the reporting cannot answer your question, the enrichment is wrong. Build all of it. It is your half, and it gets cheaper every year.
The cost per head is brutal. Check the shape again before you build against it, because per seat is no longer the only option and the section above links two that are not.
The rule that comes out of it: build anything that can be thrown away without an argument. If throwing it away would mean losing the history of your customer relationships, you are not building a tool any more, and the cost of that is not on the quote.
The fix
Keep the tools you built. Move the record onto something that maintains itself.
When the record is a byproduct of the work instead of a tax on it, the reps stop routing around it, the numbers stop being invented, and it survives the week discipline runs out. An outsider will believe a record like that, because nobody had to choose to keep it honest.
Day.ai, Clarify and Attio are three takes on that, and they differ in ways that matter for your situation. Which one you pick matters less than getting the record out of your team's hands. You keep every tool you built, and you stop maintaining the one thing you were never going to be able to prove.
Fit
Who this is for, and who it is not
- You built something and it is now load-bearing and fragile. It runs the business and it worries you, in that order.
- Or the build is still a plan. Cheapest possible moment to have this argument is before anybody writes the schema.
- You have a real process worth capturing. Stages that mean something and a way you actually sell, rather than a pipeline nobody agrees on.
- Your team sells by conversation. The record can build itself from calls and email, because that is where the truth already is.
- Your build works and you like it. Then leave it alone. This is for the people it stopped working for, not the people it still serves.
- You run on Microsoft Outlook. Some of the self-building records lean on Google Workspace and do not support Outlook yet. We will tell you in the first ten minutes rather than the first week.
- You have no process and no conversations to capture. Then a record is not your problem, and no platform will invent one for you.
- You need the odd object more than you need the record. Sometimes the right answer really is your own system alongside a bought record, and we will say so.
Where Kesey comes in
We draw the line with you, then move the record
We are a revenue operations practice and a Day.ai Solutions Partner. We help you draw the line between the tools worth keeping and the record worth moving, migrate the record onto a platform that maintains itself, and stay on to operate it after go-live, which is when the real questions arrive and when most implementations are already gone. The tools you built stay yours. We are not here to sell you a rebuild of the half you got right.
What this page does not claim. There is no case study here and no percentage, because the honest version of this argument does not need one and the dishonest version always has three. What is on this page is a structural argument plus three dated facts you can check yourself against the vendors' own pages.
If somebody shows you a migration that lifted win rates by a specific number, ask what else changed that quarter. Usually quite a lot did.
Questions
Frequently asked
Should you build your own CRM?
Build the tools, never the record. Outreach, reporting and enrichment are worth building and safe to throw away. The record, the trusted account of who your customers are and what was agreed, has to hold up for people who did not build it, and a homemade one usually does not. The working rule is to build anything you could throw away without an argument.
Is it cheaper to build or buy a CRM?
The build looks cheaper because you compare it to a licence fee and win that easily. The cost you did not price is the day somebody outside the team has to trust it: a new hire, an acquirer, a partner. A homemade record gets migrated at that point, and you pay for the whole thing at once. The per seat maths that used to justify building has also weakened, because per agent and per work pricing now exist.
What does it cost to maintain a homegrown CRM?
On the assumptions set out on this page, about 256 hours a year, roughly six and a half working weeks. It goes on keeping the schema current as the business changes, reconciling the record against the systems that hold the money, answering where a number came from, and walking each new hire through a system with no documentation but the people who built it. At a loaded $125 an hour that is $32,000 a year, every year, and none of it is the build. It is a model with its inputs printed rather than a survey, so substitute your own; if yours come out lower, keeping what you have is the right answer.
When should you stop building your own CRM?
When your reps route around it to do their job, when fields carry numbers nobody checks, when it only works for the one week a quarter someone has the discipline to feed it, and when you could not hand it to an outsider without a tour. Those are not tuning problems. They are the sign you built the half that should have been bought.
What is the difference between a CRM tool and the CRM record?
A tool does a job and can be replaced: a sequence sender, a dashboard, a script. The record is the durable, trusted account of your customers and conversations. Tools are worth building because they are cheap and disposable. The record is worth buying because it has to be trusted, and trust comes from it being maintained without anyone choosing to.
What if our data model does not fit a normal CRM?
It is the best reason to build and it still does not mean owning the record. Either choose a record whose object model actually fits, or keep the unusual object in your own system and have it reference the record rather than replace it. Owning one odd table is very different from owning the customer history.
Which CRMs build their own record?
Day.ai, Clarify and Attio each build the record from connected email, calendar and calls to different degrees, so the data is a byproduct of the work rather than a data entry tax. Day.ai populates companies, contacts and opportunities from connected email and calendar history, including history from before signup. They differ in ways that matter for your situation.
Does Kesey do the migration?
Yes. We are a revenue operations practice and a Day.ai Solutions Partner. We help you separate the tools worth keeping from the record worth moving, migrate the record onto a platform that maintains itself, and operate it after go-live. You keep the tools you built.
Stop maintaining the part you cannot prove
Keep the tools you got right. Move the record onto something that stays honest without a person feeding it. A call is the fastest way to find out which half you built.
Book a callProduct capabilities and pricing shapes described here reflect Day.ai, Clarify, Attio and HubSpot as their own published pages read in September 2026. Vendors change both, and their pages sometimes disagree with each other, so check their documentation for anything your decision turns on.
Kesey is an independent revenue operations practice and a Day.ai Solutions Partner. It is not Day AI, Inc., and it is not affiliated with Clarify, Attio or HubSpot.