• Build vs. Buy - Calculator
What a DIY value tool actually costs
Everyone who would use it: value, sales, marketing, customer success.
New value cases your team builds each month.
Pick the one closest to your plan. The three differ by how much of it your team has to keep running.
Both of the last two are an application to whoever uses it. The difference is who carries it. Configure on a platform you already pay for and the vendor keeps the backend alive. Write your own and your team does, for as long as the tool exists.
What does it cost over three years?
Where does the money go?
Open every line. Red numbers show the costs listed inside.What does that money buy, and what does it not?
A DIY value artifact does
- Produce a strong first version Narrative and numbers, from your own drivers, ready to be refined with your champion.
- Hold your value model in one place Your value leader sets the narrative and the numbers once, and the whole team works from the same model.
- Read live CRM and call data in So the seller starts from a point of view instead of a blank page.
- Give each deal a durable case The seller reopens it and keeps working on it through the cycle.
It does not
- Let your champion co-own it A case they shape and validate, rather than a document you email them.
- Hold one value story across the cycle PoV to business case to realised value, unbroken.
- Show adoption and quality you can track Who validated the numbers, and when.
- Make deals teach the next deal A governed store so each case sharpens the next.
A better AI model does not fix those four. They need a real backend, a governed store of data, and a surface your buyer and your sellers can both work in.
Build it if you need a one-time artifact that does not have to live past the interaction. Buy when the case has to keep living.
Almost all of this cost is people, and that is the number to put next to any platform quote you are looking at. Cuvama is one of those platforms.
What is behind the number.
Treat this as an estimate. It assumes a B2B SaaS company past $20M in revenue with a value leader in place. One person costs $180,000 a year fully loaded, counted as 2,000 working hours at $90 an hour. AI usage runs at $1.50 per value case, and you can set that to zero if it is inside a plan you already pay for. The horizon is three years, because that is how a CFO reads a software commitment. It leaves out procurement, legal, the work your value team stops doing, and any deal lost to a wrong number. All four would make the number bigger.How this number is built.
Three answers drive it: how many people would use the tool, how many value cases your team creates each month, and which of the three build paths you're on.The build cost is one-off, and most of it is your value team's time rather than an engineer's. Around sixty hours goes into the value model itself, meaning three to five drivers your team believes in and the math behind each one. You do that work whether you build or buy, so it sits outside the comparison. What's specific to building is writing the agent and locking the drivers so it refuses any number it can't trace back to your inputs, testing it against a closed-won and a closed-lost deal until the same inputs give the same answer twice, and showing the team what it does and where it stops.
The run cost repeats, and it's where most of the three-year figure comes from. Someone looks after the tool, which is nobody's job on paper and lands on whoever built it. Someone helps people use it, because a few sellers will pick it up on their own and everyone else needs a hand. And your value team reads the cases before they reach a customer, because nothing is locked and a wrong number costs more than the checking does. That last line scales with the number of value cases you create, so it grows as the tool succeeds.
Infrastructure only shows up on the third path. Configure on a platform you already pay for and your vendor keeps the backend alive. Write your own and you're paying for servers, a database, a deploy pipeline and a security review, for as long as the tool exists.
What the money buys is real. A DIY tool will produce a strong first version of the business case, hold your value model in one place, read live CRM and call data so the seller starts from a point of view, and give each deal a case they can reopen. What it doesn't buy is the four things in the column above, and those don't arrive with a better model. They need a real backend, a governed store of data, and a surface your buyer and your sellers can both work in.
COMMON QUESTIONS
Isn't building basically free if we already pay for the AI platform?
The license is the smallest line on the page. AI usage runs around $1.50 per value case, and you can set that to zero if it’s inside a plan you already pay for. The cost is people: the hours to build the tool and lock the drivers, the hours to test it before it touches a real deal, and then the standing time of whoever keeps it alive, helps sellers use it, and checks the numbers before they reach a customer. That last group repeats every month for as long as the tool exists, and it’s the part that never makes the original estimate.
What's the difference between building an agent, configuring an app, and writing your own software?
It comes down to who carries it afterwards. An agent is a prompt holding your drivers and your math, and someone runs it and copies the output into a document or a slide, so what you keep running is a prompt. Configuring an app on a platform you already pay for adds a screen and a CRM connection, and your vendor keeps the model, the servers, the database and the upgrades alive, so what you keep running is your configuration. Writing your own software means a repository, a deploy pipeline, a security review, and people whose job it is to maintain all of it, so what you keep running is the software. The last two look identical to whoever uses them, and the difference only shows up on your side of it.
What does a DIY value tool cost per year?
It depends on the path and the size of the team, which is what the calculator works out. The shape holds across all three: a one-off build cost in the first year, then a run cost that repeats. The run cost grows with how many people use the tool and how many value cases they create, because more cases means more checking and more people means more helping. The model prices one person’s time at $180,000 fully loaded, counted as 2,000 working hours at $90 an hour. The horizon is three years, because that’s how a CFO reads a software commitment.
When is building the right call?
When the case only has to survive the interaction it was made for. If a strong seller needs a first version of the business case for one deal, narrative and numbers together, an AI platform will produce a good one and you should let them. Building stops being the right call when the case has to keep living, which is when your champion needs to work inside it rather than receive it finished, when one value story has to carry from the first call through to the customer success handoff, when you need to see who genuinely validated the numbers, and when what you learned on this deal should reach the next one. Those four need a real backend and a governed store of data, and no model supplies either.