Project Cost Calculator

What this is for

Enter the shape of the project and get a budget split into the three buckets that matter. It is meant for the moment before you commit: you know roughly what you want automated, you need a number you can take to whoever signs it off, and you would rather not discover the real figure in month four.

The calculator does the arithmetic. The harder question is whether the number you fed it was honest, and that is what the rest of this page is about.

Reading the output: what your estimate is missing

Here is the part that separates a budget from a wish. Suppose you scope a straightforward build: a website form captures a lead, the lead lands in your CRM with the right fields mapped, and a three-message follow-up sequence fires over five days. You estimate 60 hours of work. That estimate feels generous. It is not.

The UK Treasury has published guidance on exactly this problem since 2003, because government departments were getting it wrong at scale. The Supplementary Green Book Guidance on optimism bias is built on a study of large public procurements, and it puts software squarely in a named category: Equipment and Development, defined as projects "concerned with the provision of equipment and/or development of software and systems" including "Information and Communication Technology (ICT) development projects" (HM Treasury, Supplementary Green Book Guidance: Optimism Bias).

The recommended adjustment for that category is the largest in the whole table:

For comparison, a standard building project sits at 24% upper bound and a new motorway at 44%. Software is estimated worse than motorways, by a factor of four and a half.

The guidance is not telling you to triple every number and go home. It gives a four-step method, and step two is blunt: always start with the upper bound, then reduce it in proportion to how far you have actually mitigated the things that cause the overrun. Apply that to the 60 hours.

Step one: start high. 60 hours plus 200% is 180 hours.

Step two: reduce it by what you have genuinely dealt with. The guidance breaks the bias down into contributory factors, each carrying a percentage of the total. Three of the largest for software capital expenditure are inadequacy of the business case (18%), degree of innovation (17%) and technology (18%). Take them one at a time and be honest about the mitigation factor, which runs from 0.0 for not addressed at all to 1.0 for fully addressed:

• Business case, mitigation 1.0. You have written the spec down. Every field is mapped to a destination field by name, the follow-up copy is signed off, and you have agreed what "done" means. That removes the full 18 points.
• Degree of innovation, mitigation 1.0. You are using an established automation platform and a CRM that tens of thousands of businesses already run. Nothing here has never been built before. That removes 17.
• Technology, mitigation 0.5. You cannot stop a vendor changing an API mid-build or deprecating the endpoint you were relying on. Half credit, so 9 points.

That is 44 points of the bias mitigated, so the resultant adjustment is (100% − 44%) × 200% = 112%.

Step three: apply it. 60 hours × 2.12 = 127 hours.

Read that carefully, because it is the whole point of this page. Your honest budget is roughly double your estimate, and that is after you have done the three things that most reduce the risk. Skip the written spec and you are back up near 180 hours. Multiply whichever figure applies by the hourly rate you are actually paying, and you have a build number you can defend.

The same method applies to the calendar. A four-week estimate carries a 54% upper bound on duration, so it becomes six weeks before you have mitigated anything at all. If you have promised a client a launch date, that is the number to promise.

The method: three buckets, and only one of them is a one-off

One-time build. Scoping, integration work, testing, documentation and handover. This is the bucket the uplift above applies to. It is also the bucket people try hardest to compress, which is why it overruns.

Platform subscriptions, monthly, forever. The automation platform, the CRM, the AI API, and any specialist service such as a voice provider. Small numbers that compound: £60 a month is £720 a year and £2,160 over the three years the system will plausibly run.

Maintenance, monthly, forever. Vendors change APIs. A field gets renamed in your CRM and a workflow starts writing into nothing. Someone wants a fourth follow-up message. Budget for it explicitly, because the alternative is that the automation quietly stops working and nobody notices until a month of leads has gone missing.

Over a three-year horizon the recurring buckets routinely exceed the build. Any comparison of two options that only weighs the build cost is comparing the wrong number.

What the platforms actually cost

Subscription tiers move, so treat these as a starting point and check the vendor page. Zapier's own pricing lists a free tier at $0 a month with 100 tasks, Professional from $19.99 a month, and Team from $69 a month, with a currency selector for GBP and others (zapier.com/pricing). The number on the plan is rarely what surprises people, though. Four details from that same page are:

• Tasks are consumed per successful action, and the rate varies. Every step in a Zap and every external connector call uses tasks, and the AI model tier, the code runtime and the connector type each change how many. A workflow with an AI step in it does not cost the same as one without.
• Triggers and polling are free. Checking for new data never costs a task. Only completed actions do. This is the opposite of what most people assume when they budget.
• A useful set of built-in steps is free too: Tables, Forms, Filter, Formatter, Paths, Delay, Looping, Sub-Zap, Digest and Storage. Restructuring a workflow to do its branching and formatting in built-ins rather than in billable steps is a real lever on the monthly bill.
• Overage is the trap. Cross your task limit and you are switched to pay-per-task billing, charged at a higher rate than your subscription tasks. Turn that off and your workflows pause instead until the next billing period. Either outcome is worse than buying the right tier, and neither shows up in a first-month budget.

Self-hosting changes the shape rather than removing the cost. n8n's documentation is clear that the Community edition is free indefinitely and carries "almost the complete feature set", with a Registered Community tier that is also free and adds folders, debug in editor and custom execution data (docs.n8n.io). What it does not include is the list that forces the upgrade: SSO, environments, projects, external secrets, external binary storage, log streaming and Git-based version control all require a paid plan. So the licence is £0 and the honest cost is the server, the backups, the upgrades and the hour you spend when it stops. Whether that trade is worth it is what the n8n hosting cost calculator is for.

For what it is worth on the build side: a simple lead capture and CRM sync automation typically runs £500 to £1,500 to set up, and a full AI voice agent with appointment booking is typically £2,000 to £5,000.

Common mistakes

Budgeting the build and not the year. The most common and the most expensive. Add twelve months of subscriptions and maintenance before you compare anything to anything.

Treating the estimate as the budget. An estimate is the cost if nothing surprising happens. A budget is the cost including the surprises. The Treasury method above is one way to get from one to the other, and its central instruction is to start pessimistic and earn your way back down rather than starting optimistic and hoping.

Scoping in a conversation instead of a document. Inadequacy of the business case is the single largest named contributor to software cost overruns in the Treasury tables, and it is also the cheapest to fix. An afternoon writing down field mappings and acceptance criteria is the highest-return hour in the project.

Automating everything at once. Build the one workflow that touches the most money, run it for a month, then decide what is next with real data. A five-workflow first phase multiplies the integration surface and every integration is a place the estimate breaks.

Forgetting whose time the manual process is currently costing. A build only makes sense against a baseline. If you have not measured what the current process costs in hours, the automation ROI calculator is the thing to run before this one, and the VA versus automation calculator if the realistic alternative is hiring a person.

Assuming the cheapest platform is cheapest. Task-based pricing means a badly structured workflow on a cheap tier can cost more than a well structured one a tier up. Make.com versus Zapier for small business works through where the crossover actually falls.

Frequently Asked Questions

Why is the uplift for software so much higher than for construction?

Because the requirements move. The Treasury tables put Equipment and Development at a 200% upper bound on capital expenditure against 24% for a standard building, and the named contributory factors explain why: inadequacy of the business case, degree of innovation and technology account for the largest shares. A building's specification is largely fixed once ground is broken. Software specifications are discovered while you build, which is a polite way of saying they change.

Should I really budget double my estimate?

Double is what the method produced for a well-mitigated example with a written spec, proven platforms and no novel technology. It is not a scare figure, it is the good case. What you should take from it is the shape of the answer rather than the exact multiplier: if your budget has no contingency at all, it is not a budget. If you want a smaller multiplier, the way to earn it is to remove the causes, which means writing the spec down and choosing boring technology.

Is a fixed price better than hourly?

A fixed price transfers the estimation risk to whoever quotes it, which is worth paying for, and it only works if the scope is written down tightly enough to be priced. A vague brief and a fixed price produces either a change-request argument or a padded quote. The order that works is scope first, then price.

What should I put aside for maintenance?

Enough that a platform changing an API is an inconvenience rather than a crisis. The honest input is how much revenue flows through the automation: something that books appointments needs monitoring and a person who responds, and something that tidies a spreadsheet does not. Budget by consequence of failure, not by size of build.

Do the platform costs scale with my volume or my complexity?

Both, and complexity is the one people miss. Task-based pricing charges per successful action, so a workflow with eight steps costs four times as much per run as one with two steps at the same volume. Doubling your leads doubles the bill; adding four steps to the workflow also doubles it.

Can I use this for a client quote?

You can use it to build the internal number. Do not hand a client the raw calculator output, because it contains your contingency and inviting a negotiation over your contingency defeats the purpose of having one. The agency retainer calculator is the better tool for the client-facing side of the same question.

Getting to a number you can commit to

A calculator can uplift an estimate. It cannot tell you whether the estimate it started from described the right project, and that is where the money actually goes: the field that turned out to exist in three systems under three names, the CRM that cannot trigger on the event you assumed it could, the approval step nobody mentioned. None of that is in the arithmetic, and all of it is in the invoice.

Which is the argument for spending an hour on scope before spending anything on build. If you want that hour with someone who has wired these platforms together before, book a free scoping call and bring whatever you have written down so far, however rough. If you would rather engage through a platform with the paperwork already handled, the same work is available on Upwork and Fiverr.

If you are still at the stage of working out whether the project is worth doing at all, AI chatbot pricing for small business covers the same cost question for one specific build, with the numbers broken out.