The models we design these systems against.
Four metrics and three scenarios, labelled for what they are: illustrative.
We do not yet have results a named client has agreed to publish, and we are not going to pretend otherwise. This page shows the order of magnitude we design for and the kinds of problem we work on. The day a client agrees to show their own numbers, we rewrite it from scratch.
The targets we design against
Four target values we hold our own work to. They are design thresholds rather than averages from deployments: a specification we set ourselves, not a promise of outcome.
Illustrative scenarios and figures — they show the order of magnitude these systems are designed for, not audited results from named clients.
- ROI
- 450%
Return on investment: what comes back for every unit spent on the project, measured after a year.
- ROAS
- 8.5x
Campaign revenue per unit of ad spend. A channel below the threshold gets switched off.
- AI Response
- < 2s
How long an agent takes to answer a customer's first question, including outside office hours.
- LCP
- 0.8s
How long before the main content of a page appears. Google counts it, and so does the visitor.
Three kinds of problem we solve
Each scenario describes the kind of company we work with and the effect we design for. None of them is the story of a specific client.
Illustrative scenarios and figures — they show the order of magnitude these systems are designed for, not audited results from named clients.
Websites and applications (B2B SaaS)
Company profile
B2B SaaS platform
Modelled result
+240% leads in 90 days after a redesign
Marketing and sales (e-commerce)
Company profile
Fashion e-commerce
Modelled result
7.3x ROAS and −38% CAC over Q4
AI agents (HR and logistics)
Company profile
Logistics operator
Modelled result
An agent handles 80% of repetitive cases, saving 120 hours of manual work a month
How we get to numbers like these
We start with a free audit, and the audit can end with "not worth it". If it is worth it, we sort out the process before writing the first line of code, then ship in small pieces and measure each one. The scenarios above show where that method leads when the problem is well defined.
Have a process you want to sort out?
Describe it in a few sentences. We come back with a plan for a free audit and a straight answer on whether this is work for us.