If a Dev Shop Gives You a Timeline on the First Call, It's a Guess

On a discovery call recently, someone on our side said "eight to ten weeks-ish." It was said the way you say something that feels obviously true. The founder had described the product clearly, we understood it, and eight to ten weeks was about what that kind of product feels like. The next day we sat down and actually counted the work. Thirteen separate workstreams, each one sized in team-weeks, added up. The total was not eight to ten weeks. It was closer to twenty-eight.

Nobody lied on that call. That is the uncomfortable part. The number was said in good faith and it was still wrong, because it came out of a completely different process than the estimate did. On the call we were sizing the idea. The next day we were counting the work. Those two numbers are almost never the same, and the first one is the one everybody remembers.

Why the first number is always low

When a founder describes a product, your brain reaches for the last thing that looked like it. A scheduling app. A dashboard with a couple of roles. A tool that turns a form into a document. The estimate that comes back is a comparison, and comparisons feel like knowledge. They are not. A timeline comes from the number of distinct things somebody has to build and the number of decisions that have to be made along the way, and neither of those is visible in the description of the idea.

The parts that make a project long are almost never the parts the founder talks about. They are the account and permission rules once more than one kind of person logs in. The admin console somebody has to run the thing from. The notification path, which sounds like one line and is actually a decision about email versus text, a sending domain, templates, and what happens when a message does not arrive. The data model that has to survive next year without a rebuild. Whatever data already exists somewhere else and has to get moved in. The testing pass. The deploy path. And the stretch at the end where a real user finds every gap a demo never showed you. On the call, all of that lives inside one word: "and login, obviously."

What counting actually looks like

It is not complicated and it is not fast. You write down the workstreams. Not features, workstreams: the chunks of work that have their own design decisions, their own testing, and their own ways to go wrong. You put a number of team-weeks next to each one. You multiply by a rate you actually charge, not a rate you wish you charged. Then you look at the total before you look at the client. It takes an afternoon, it is mechanical, and it disagrees with your gut nearly every time. That disagreement is the entire value of doing it.

One detail inside that math matters more than it looks. Name the unit. A team-week and a dev-week are not the same thing, and if you mix them halfway down a spreadsheet you have built a two-times error into your own proposal without noticing. We caught exactly that while checking our own numbers. "Twelve weeks" from a shop putting three people on it and "twelve weeks" from a shop putting one person on it are two different claims wearing the same word. If a proposal says weeks and does not say whose weeks, ask.

The parts nobody has built before

Here is something we will say in public: we have never priced an AI feature. Every scope of work we have written so far explicitly excludes them. So when this scope came back with two AI workstreams sitting in the middle of it, the honest answer was that we did not have a number, we had a feeling. A confident price on those two lines would have been fiction, and it would have been our fiction to eat later.

The move there is a small paid spike. Fixed price, fixed duration, a written output at the end: what the feature costs to run at real volume, where it fails, what it takes to make it good enough to ship, and whether it is worth building at all. Then you price the remaining work against something you have actually touched. Founders sometimes hear "spike" as a shop stalling for money. It is closer to the opposite. It is a shop refusing to hand you a number it cannot stand behind, and choosing to be honest at the moment that is most expensive to be honest.

When the two numbers are far apart

You do not split the difference. Splitting the difference produces a price nobody believes and a date nobody hits, and it moves the pain from the proposal into month four, where it costs a relationship instead of a conversation. You go back to the scope and cut until the number is real.

That is what we did. We rebuilt the phases around the smallest first release that actually does the thing the founder needs it to do, priced that, and wrote everything else down as later phases with their own prices attached. The founder ends up with a date we can hit and a written list of what is not included. That exclusion list is the most valuable page in the document, and it is the page most proposals do not have. A phase without an explicit "not in this" section is not a scope, it is a mood.

There is a version of this where the shop protects its first number instead. It quietly absorbs the extra work, the date slips twice, the quality drops in the last three weeks, and the founder is told at the end that the scope grew. Every one of those outcomes is paid for out of your runway. The gap between the guess and the count does not disappear because nobody looked at it. It just gets paid later, by you, at a worse exchange rate.

What to do with this on your next call

When a number comes out of somebody's mouth on a first call, write it down as what it is. A shape. Not a commitment, not a quote, not a date you can put in front of an investor. Then ask for the counted version: the workstream list with time next to each line. Ask which of those workstreams the shop has built before, and what happens to the estimate on the ones they have not. Ask what "weeks" means and how many people are inside one. Ask what is explicitly excluded from phase one.

And if the counted number comes back different from what you heard on the call, that is not a red flag. That is the process working in front of you. The shop you should worry about is the one whose first guess and final proposal match perfectly, because it means only one of those two things ever happened.

The risk was never that somebody said a number out loud. The risk is nobody going back to check it.