What the CAEC process looks like on a typical project
How the CAEC process takes a connected device from rough idea to tested prototypes: what happens in each phase and what you receive.

Most founders who contact us have heard the names of the four phases of our CAEC process — Concept, Abstract, Exact, Conclusion — but not what actually happens inside each one, or what lands on their desk at the end of it. That gap matters, because a phase name tells you nothing about when you have to make decisions, when money gets committed and when you can still change your mind cheaply.
This post walks through a typical connected-device project, the kind with a sensor board, a battery, a radio, an enclosure and a companion app, and shows what each phase produces and what we need from you along the way. It is written for founders and product managers who are about to commission hardware for the first time.
Why the CAEC process has four phases, not two
Many hardware projects are run as “design, then build”. The trouble is that the expensive mistakes are made at the boundary between the two: an architecture choice that looked fine on paper becomes a respin, a tooling change or a certification surprise. Splitting the work into four phases puts deliberate checkpoints where those choices are cheap to revisit.
The rough shape is:
- Concept — what the product should be, deliberately ignoring constraints.
- Abstract — the same product brought back to manufacturing reality, with an architecture and a costed plan.
- Exact — detailed design, a small prototype batch, testing and bench reports with proposed fixes.
- Conclusion — fixes applied, devices reproduced, documentation and handover.
Each phase ends with something you can hold, read or review, and each one narrows the space of options before the next one starts spending money on copper and plastic.
Concept: describing the product without apologising for it
On a typical project, Concept starts after the scoping call. We write down what the device should do for the person using it, how it should feel in the hand, where it lives, how long it should run between charges and what the app should show. We produce an outline architecture, candidate materials and conceptual 3D renders.
The important part is what we leave out. We do not yet let the bill of materials, the battery chemistry or the tooling budget veto ideas. We have written about why the Concept phase deliberately ignores what is possible; the short version is that if constraints arrive first, nobody ever finds out which of them were worth fighting.
What we need from you here is honesty about the user and the use case, not technical answers. “It sits on a warehouse shelf and nobody touches it for a year” is far more useful than a preferred microcontroller.
Abstract: where the plan meets the factory
Abstract is where most of the conversation happens, and where most of the real decisions get made. We take the concept and test every part of it against cost, power, radio, certification and manufacturing. A typical set of questions on a battery-powered sensor product looks like this:
- Radio. Does the data volume and range point to BLE, Wi-Fi, LoRaWAN or cellular? That choice drives the power budget, the certification route and the backend.
- Power. What is the average current in each operating state, and does the chosen cell actually deliver the promised life once self-discharge and cold temperatures are taken into account?
- Processor. An ESP32-class part if Wi-Fi is central, an STM32L4 or nRF52840 if sleep current and BLE matter more. The radio decision usually makes this one for us.
- Enclosure. 3D printing or SLS for early units, CNC or injection moulding for volume, and what ingress rating (IP54, IP67) the environment really requires.
- Software. What the device sends, what the cloud stores, and what the app or dashboard must show on day one.
Some concept ideas survive untouched. Some get adjusted — a thinner case becomes slightly thicker to fit a cell that meets the runtime target. A few get dropped, and we say why. The output is an architecture document and a costed plan with a range per phase, so you know what Exact and Conclusion will cost before you commit to them.
This is also the cheapest moment to change direction. Moving from LoRaWAN to BLE in Abstract means editing a document. Doing it in Exact means a new board.
Exact: drawings, boards and the first honest measurements
Exact is where the design becomes physical. Schematics and PCB layout, firmware, enclosure CAD, backend and app all proceed in parallel against the architecture agreed in Abstract. Because one team owns all of it, interface questions — which pin wakes the processor, what the payload format is, how the app learns about a firmware version — are settled between people who sit next to each other rather than across three contracts.
We then build a small prototype batch. Not one unit: a single board hides assembly variation, and a handful lets us see which problems are design and which are build. Those units go through bench testing, and where the product needs it, environmental runs such as temperature cycling, along with pre-certification checks for emissions and radio performance.
Every problem we find is written up with the measurement, the likely cause and a proposed fix. The format matters; we covered it in a bench report a founder can actually read. Typical findings on a first batch include:
- sleep current a few times higher than budgeted, traced to a pull-up resistor or a peripheral left powered;
- radio range shorter than expected because the antenna sits too close to the battery or a ground pour;
- a snap fit that cracks after repeated opening, or a gasket that does not compress evenly;
- firmware edge cases around reconnection or an interrupted OTA update.
None of these is unusual. The point of Exact is to find them on a handful of units, not in the field.
Conclusion: prototypes you can put in front of people
In Conclusion we apply the fixes from the bench reports, re-verify the issues that were found and reproduce the devices. You receive fully operating, mostly production-ready prototypes: units good enough for pilot users, investor demonstrations or a contract manufacturer’s quotation.
Alongside the hardware comes the documentation: schematics, PCB files, BOM, firmware source, enclosure CAD, backend and app code, and the test reports. You own all of it. If you later move manufacturing elsewhere or bring engineering in-house, nothing is locked behind us.
One thing Conclusion does not include is formal certification. CE RED or FCC Part 15 testing happens at an accredited lab, usually on units built from the Conclusion design. Our pre-certification work in Exact is there so that lab visit is a confirmation rather than a discovery.
What changes for you in each phase
The phases also change what we ask of you. A rough guide:
| Phase | Your main input | What you receive |
|---|---|---|
| Concept | Users, use case, ambition | Outline architecture, materials, concept renders |
| Abstract | Trade-off decisions, target cost | Architecture document, costed plan per phase |
| Exact | Feedback on prototypes and reports | Prototype batch, test results, bench reports with fixes |
| Conclusion | Sign-off | Verified prototypes, full documentation, IP handover |
Throughout, progress is visible in our project portal, with a direct line to the engineers doing the work. On a typical connected device the whole sequence takes three to six months from concept to tested prototype batch; software-only projects move faster.
What we would do before starting any CAEC process
Whether you work with us or with anyone else, a few things make the first two phases much faster:
- Write one paragraph describing the person using the product and where it lives.
- List the three things it must do and the three things it must never do.
- Decide your target unit cost range and expected first-year volume, even roughly.
- Note any market you must certify for, since that shapes the radio and the timeline.
- Collect anything that already exists: sketches, earlier prototypes, code, competitor teardowns.
Bring those to the first conversation and Concept can start from the product rather than from a questionnaire.