I design mobile experiences for businesses that sell real things.
Twelve years designing and building the mobile side of South African brands — luxury furniture, plastic packaging, motorsport, metal fabrication, temperature sensors. The work is interaction design, prototypes that hold up on a real phone over a real connection, and a number I’m willing to be judged on after launch.
Six clients, six problems worth solving.
Real businesses, live in production, from a furniture showroom in Pretoria to a race track in Meyerton. Each case covers the constraint, the interaction decisions, how I prototyped it, and what happened after launch.
Joy Furniturejoyfurniture.co.za
Luxury furniture retail · E-commerce
A R38,900 sofa, bought from a phone
High-ticket furniture with a 10–12 week lead time, sold mostly through WhatsApp. I rebuilt the product page so the price, the instalment and the lead time are honest before anyone taps buy.
+31%add-to-cart rate
×2.4WhatsApp enquiries
Open case study
A R38,900 sofa, bought from a phone
High-ticket furniture with a 10–12 week lead time, sold mostly through WhatsApp. I rebuilt the product page so the price, the instalment and the lead time are honest before anyone taps buy.
A R38,900 sofa, bought from a phone
High-ticket furniture with a 10–12 week lead time, sold mostly through WhatsApp. The product page had to be honest about the price, the instalment and the wait — before anyone taps buy.
- Role
- Design & build (Pedas Design Studio)
- Client
- Joy Furniture, Pretoria
- Timeline
- 8 weeks
- Stack
- WordPress · WooCommerce · Float · WhatsApp
The constraint
Joy sells sofas and dining sets between R22,000 and R62,000, with a 10–12 week lead time on anything not in stock, and the real conversation happens on WhatsApp. Two things were working against that. The lead time lived in a dismissible banner at the top of the page, where it read as an advert rather than a condition of purchase. And the WhatsApp button opened a blank thread, so every enquiry started with a consultant asking which product, which colour, which price — the customer retyping what they had just been looking at.
Analytics said 71% of sessions were mobile, add-to-cart was healthy, and the seven-field “Get a Quote” form was where people stopped.
Interaction decisions
- A sticky buy bar that tells the whole truth. Price, the instalment line (“or R9,725 × 4 with Float”) and the lead time sit together in a bar pinned to the bottom of the screen. The lead time moved out of the banner and into the price block, because it is the second thing a buyer needs to know and the first thing they ask about.
- WhatsApp handoff with context attached. The button opens a pre-filled message containing product name, SKU, price and colour — so the consultant’s first reply is an answer, not a question. The customer types nothing twice.
- One gallery, honest crops. A single 4:3 gallery rather than a wall of thumbnails: swipe with position dots, plus a toggle between “in a room” and “on white” so scale is legible on a 390 px screen.
- Specs above reviews. Dimensions, materials and what’s included sit in an accordion above social proof. For a R40,000 purchase the deciding information is a measurement, not a star rating.
- Combo deals as arithmetic. Bundles show the saving as a number (“save R22,800”) in the card, not a red percentage badge — the discount is the offer, so it should be readable at a glance.
- A shorter quote path. Seven fields became three, and “Interested in” arrives pre-filled from the category the person was browsing. Mobile completion is mostly a question of how many keyboards you make someone open.
How I prototyped it
A clickable Figma prototype mirroring the WooCommerce template, tested on customers’ own phones in the Pretoria showroom over throttled 3G. Eight participants, two tasks: find a dining set for six people under R25,000, and find out whether the TV bed is in stock. The second task was the useful one — everyone found the product, and nobody could tell me when it would arrive until they scrolled back to the banner at the top. That single observation is where the sticky bar came from.
Outcomes
Support calls asking about lead time dropped by roughly half over the same period — the number moved because the information moved, not because we added a cleverer banner.
We were treating instalments as a payment detail. Customers were treating them as the decision.
What I’d do differently
The Float instalment line belongs on the category cards too, not just the product page. Buyers shortlist in the grid, and a R38,900 sofa looks like a different object when the grid says “R9,725 × 4”.
Riverwalk Furnitureriverwalkfurniture.co.za
Designer furniture · E-commerce
Quiet luxury, without the showroom
Leather sofas and chesterfields for architects and designers. The catalogue read like a spec sheet, so I turned it into a lookbook with materials and finishes one thumb away.
+27%scroll to materials
3.1×product shares
Open case study
Quiet luxury, without the showroom
Leather sofas and chesterfields for architects and designers. The catalogue read like a spec sheet, so I turned it into a lookbook with materials and finishes one thumb away.
Quiet luxury, without the showroom
Leather sofas, chesterfields and beds bought by architects and interior designers. The brand is restraint; the catalogue was reading like a price list.
- Role
- Design & build (Pedas Design Studio)
- Client
- Riverwalk Furniture, Pretoria
- Timeline
- 6 weeks
- Stack
- WooCommerce · WhatsApp · showroom visit
The constraint
This is a specifier audience. They are not shopping for themselves: they are choosing a piece for someone else’s room, on a phone, between site visits, and then sending it to a client for approval. The site was organised like a warehouse — three flat category images, then a grid of products that led with price. The thing that actually decides a leather sofa purchase in this market, the hide, the stitching, the fill and the dimensions, was a paragraph at the bottom of the page that most people never reached.
Interaction decisions
- Category entry as a lookbook. One piece per screen, full-bleed, with the name and material in a single line at the bottom edge. Scrolling becomes browsing rather than scanning a grid of thumbnails.
- Materials as a sheet, not a page. Leather grades, stitching, fill and dimensions open in a bottom sheet over the product photograph. The piece never leaves the screen, so the detail is read with the object rather than instead of it.
- Share to client. A single action exports a clean vertical card — photograph, dimensions, price, showroom address — sized for a WhatsApp forward. Specifiers forward images; so the forward became the product.
- Restraint in the interface itself. Two type weights, no gradients, no badges, no countdowns. Sale prices show the original struck through and the new number in the same colour. The restraint is the brand, so it can’t be a separate style guide that the interface ignores.
- A showroom block that ends the loop. Address, hours, one-tap directions, and a WhatsApp thread that opens with the specific piece already named.
How I prototyped it
Two rounds with local interior designers, six participants, on phones. I deliberately prototyped the share card as a static image first and asked them to send it to me in WhatsApp — the first version exported at 1:1 and the dimensions became unreadable in the chat preview. The revised card is 4:5, which survives the crop WhatsApp applies.
Outcomes
The buying decision doesn’t happen on the product page. It happens in a WhatsApp thread between a designer and their client.
What I’d do differently
The share card came late, as a convenience. It should have been the primary action on every product, built first and tested first — because it’s the moment the product actually gets discussed by the people who pay for it.
Novopaknovopak.co.za
Plastic packaging · B2B catalogue
Packaging specs that fit in one hand
Tubes, tubs, bottles and jars sold to product developers. Buyers arrived knowing a volume and a neck size, not a product name — so the catalogue became a filter, not a list.
+42%quote requests
2.8SKUs per enquiry
Open case study
Packaging specs that fit in one hand
Tubes, tubs, bottles and jars sold to product developers. Buyers arrived knowing a volume and a neck size, not a product name — so the catalogue became a filter, not a list.
Packaging specs that fit in one hand
Tubes, tubs, bottles and jars. Customers arrive knowing a volume and a neck finish, not a product name — so the catalogue became a filter instead of a list.
- Role
- Design & build (Pedas Design Studio)
- Client
- Novopak, plastic packaging manufacturer
- Timeline
- 10 weeks
- Stack
- WooCommerce catalogue · PDF datasheets · enquiry shortlist
The constraint
Novopak’s customers are product developers and brand owners launching a cosmetic, food or chemical line. They don’t shop by family — they arrive with a volume, a diameter and a neck finish, and they’re standing on a factory floor or at a trade show with one hand free. The old catalogue listed Tubes, Bottles and Jars as three destinations, each ending at a contact form that carried nothing the visitor had already chosen. A shortlist of four candidate jars became four separate emails, or one email listing them from memory.
Interaction decisions
- Filter before browse. Volume, diameter, neck finish and material as sticky chips at the top of the grid. The page answers a specification question instead of presenting a family of products.
- An enquiry shortlist, not a cart. B2B buyers compare — so the action adds a SKU to a shortlist that turns into one structured request. The button says “Add to enquiry · 3” and the count is always visible.
- Spec tables with unit toggles. ml/oz and mm/in, a dimension diagram with labelled callouts, and a per-SKU PDF datasheet — because the datasheet is what gets forwarded to whoever signs off internally.
- Two intents, two paths. “Request a quote” (volume, target date, artwork status) is separated from “Request a sample”, since the two requests land on different desks. Combining them is why so many B2B forms get half-filled.
- A configurator for the non-standard. The custom tube page takes diameter, length and cap type and draws a live dimension sketch, so a non-standard request arrives with numbers attached rather than an email that says “something like this”.
- Built for one thumb. 48 px targets, a sticky action bar, and no interaction that depends on hover.
How I prototyped it
A clickable prototype with the real SKU data, tested with six buyers — three from cosmetics, three from food. The task was “you need a 250 ml jar with a 63 mm neck for a balm; get three options and a price”. The first build had the filters collapsing on scroll, which lost everyone mid-task; pinning them and keeping the shortlist count in the sticky bar fixed it in the second round.
Outcomes
What I’d do differently
I built the filter labels from the catalogue, not from customer language. Three buyers asked for “neck size” while the interface said “closure diameter” — same measurement, different vocabulary, and the search box was where they went looking. Label in the customer’s words, then map to the spec internally.
Midvaal Racewaymidvaalraceway.co.za
Motorsport venue · Bookings & events
Race day, one-handed, at the gate
Track days, drags and venue hire for two very different audiences. Every event card now answers four questions: when, how much, can I race, how do I get there.
+38%event → entry enquiries
−44%“what time?” calls
Open case study
Race day, one-handed, at the gate
Track days, drags and venue hire for two very different audiences. Every event card now answers four questions: when, how much, can I race, how do I get there.
Race day, one-handed, at the gate
Track days, Friday night drags and venue hire, for two audiences who want opposite things from the same page — and both of them are on a phone with bad signal.
- Role
- Design & build (Pedas Design Studio)
- Client
- Midvaal Raceway, Meyerton
- Timeline
- 7 weeks
- Stack
- WordPress · events · add-to-calendar · cached directions
The constraint
Two audiences, one calendar. A competitor needs the date, the entry fee, the gate time and the paperwork — licence, indemnity, scrutineering, noise limit. A spectator needs the date, the start time, the ticket price and the road in. Previously both arrived at a list of event posts where the times lived in body copy, and the single most-asked question on the phone was “what time does it start?”. The venue is out on the Meyerton road, so a meaningful share of visits happen on a weak connection, sometimes at the gate, sometimes already late.
Interaction decisions
- One race-day card, four answers. Date and time as a badge, event name, competitor and spectator pricing side by side, and a single primary button. If a card can’t answer when, how much, can I race and how do I get there, it isn’t finished.
- An audience toggle that re-sorts the page. “I’m racing” / “I’m watching”, remembered between visits. Racing puts the entry requirements and the entry button first; watching puts directions, gate times and what to bring first.
- Add to calendar, and a map that works offline. A one-tap .ics download, plus a small service-worker cache so the event page and the directions open at the gate with no signal — the moment the page is most needed.
- Entry requirements as a checklist. Licence, indemnity, scrutineering, noise limit — individually checkable, saved locally, with a remaining count in the sticky bar. Paperwork is the reason people turn up and can’t race.
- Built for sunlight. Large type, high contrast, single column, and every action inside the thumb arc — read at 11am standing next to a hot car, not at a desk.
- Sharing designed for how race days spread. A WhatsApp share that sends the card as an image with the date and gate time burnt in, because a race day is shared in a car club group, not by email.
How I prototyped it
I prototyped on a phone, outdoors, in the middle of the day — bright screen, sunglasses, one hand. Eight people: four regular competitors from the track’s Facebook community and four first-time spectators. The decisive finding was not a layout problem: given a page with a calendar on it, every spectator still opened WhatsApp to ask a human what time to arrive. So the design stopped trying to replace the phone call and started making it unnecessary — gate time on the card, in the share image, and in the calendar entry.
Outcomes
What I’d do differently
Competitor entry still ends in a phone call or a WhatsApp message, because the venue takes payment at the gate. A real entry form with a payment link would remove the busiest manual step on race morning — but it’s an operations change before it’s a design one, and I’d want the venue’s staff in that conversation from the first week.
3D Metal Workx3d-metalworkx.co.za
Metal fabrication · Lead generation
Send a drawing, get a quote
Laser cutting, plasma, welding and rolling for Joburg trade. A buyer with a job can’t tell what a shop can make, so capability limits and a drawing upload became the whole page.
×2.6enquiries with drawings
−61%quote turnaround
Open case study
Send a drawing, get a quote
Laser cutting, plasma, welding and rolling for Joburg trade. A buyer with a job can’t tell what a shop can make, so capability limits and a drawing upload became the whole page.
Send a drawing, get a quote
Laser cutting, plasma, welding, rolling and punching for Joburg trade. Buyers arrive with a job in hand and no idea whether this shop can make it — so the page had to answer that, or the enquiry is a phone call.
- Role
- Design & build (Pedas Design Studio)
- Client
- 3D Metal Workx, Johannesburg
- Timeline
- 6 weeks
- Stack
- WordPress · showcase · upload · call-back slots
The constraint
Fabrication is a capability business, and capability is invisible in a photo. A buyer looking at a balustrade gallery can’t tell whether the shop can roll a 6 mm tube, cut 20 mm plate, or handle a 3 m sheet — so every enquiry started as a phone call, and the ones that came through the form arrived without a drawing, which made them impossible to quote without a second conversation. Three exchanges to get to a number is a week, and in trade work a week is a lost job.
Interaction decisions
- A capability matrix, not a services list. Process by material by thickness by maximum size, in a table that reads properly on a phone: two columns that stack, sticky row headings, no horizontal scroll. It answers “can you do this” in one glance, which is the only question that matters at the top of the funnel.
- “Can you make this?” as the primary action. A dashed drop zone that accepts a photo, a drawing or a PDF, opens the camera on mobile, and falls back to WhatsApp with an attachment for people who’d rather do it there.
- Call-back windows instead of a form. Pick a 30-minute slot rather than submitting an email address into a void. For trade buyers the phone is the medium; the job of the page is to guarantee it rings at a known time.
- Showcase filtered by process. Every case states material, thickness and finish, so the gallery becomes evidence instead of decoration.
- A structured enquiry, so quoting is cheap. The form produces a single message containing process, material, size, quantity and the attachment — which is why turnaround collapsed. The design work happened as much in the inbox as on the page.
How I prototyped it
Paper first, then a clickable build, tested with six people who buy fabrication for a living — two shopfitters, two sign companies, two maintenance managers. I watched where they went when they had a real drawing in hand, and every one of them opened the gallery, scrolled, and left to phone. That’s how the matrix became the hero element and the gallery became secondary: the site had been showing finished work when the buyer needed limits.
Outcomes
What I’d do differently
The thickness ranges went onto the site before I’d confirmed them with the workshop floor, and two were optimistic. A pricing and capability page written from a sales conversation rather than from the machine is a promise the business has to keep — that one needed a correction and an apology to a customer.
Thermocouple Productsthermocoupleproducts.co.za
Temperature sensors · B2B catalogue
A catalogue an engineer can use in gloves
Thermocouples, RTDs and heating elements made to order in Edenvale since 1992. Maintenance engineers search by dimension and temperature, not by family — so we let them build the spec instead.
+35%spec enquiries
−29%repeat “spec” calls
Open case study
A catalogue an engineer can use in gloves
Thermocouples, RTDs and heating elements made to order in Edenvale since 1992. Maintenance engineers search by dimension and temperature, not by family — so we let them build the spec instead.
A catalogue an engineer can use in gloves
Thermocouples, Pt100 RTDs, thermowells and heating elements, made to order in Edenvale since 1992. Engineers search by dimension and temperature — so we let them build the specification instead of browsing families.
- Role
- Design & build (Pedas Design Studio)
- Client
- Thermocouple Products (Pty) Ltd, Edenvale
- Timeline
- 9 weeks
- Stack
- WordPress · spec catalogue · datasheets · drawing upload
The constraint
The buyer is a maintenance engineer with a line down. One hand free, gloves on, heat and noise around them, and a phone. They are not browsing a catalogue — they know they need a Type K probe, 6 mm diameter, 300 mm long, 1/4 BSP, good to 400 °C. The site was organised the way the factory is (thermocouples, RTDs, band heaters, immersion heaters), which is the way the business thinks and not the way the problem arrives. Enquiries then came in as free text — “a probe like before, about 300 long” — and it took three emails to get to a quotable specification.
Interaction decisions
- Search by specification. Sensor type (K, J, PT100), probe diameter, length, thread, maximum temperature and sheath material as chips, plus a real search field with synonyms — so “neck size” and “thread” find the same thing.
- Build-your-spec instead of a text box. The configuration becomes a structured message containing all six values, and the summary can be screenshotted and handed to a colleague. Non-standard work keeps its own path: send us a drawing, a sample or a problem, with camera capture built in.
- Spec tables that reflow. No horizontal scrolling on a phone: sticky first column, tap-to-copy part numbers, workable at 360 px wide. A datasheet PDF sits on every product, because that’s what goes into the maintenance file.
- Call an engineer, as a feature. For this buyer a phone call is the fastest correct answer, so a sticky button reaches the workshop with the part context already attached — the site stops competing with the phone and starts feeding it.
- Built for the plant. 52 px targets, 4.5:1 contrast, legible at full brightness, nothing that depends on hover or a precise tap.
How I prototyped it
Two rounds of six, half of them engineers from maintenance teams rather than buyers, tested on their own phones during a break — the closest I could get to the real context without stopping a line. The first build put the product family first and the spec chips second; the moment the chips were on top, every participant went straight to them and skipped the family entirely. The second finding was about completeness: when the enquiry was six fields, quotes came back in one email; when it was one text box, they took three.
Outcomes
Nothing about this market is a shopping experience. It’s a person with a line down, trying to read a part number through a glove.
What I’d do differently
The spec chips came straight from the catalogue columns, which meant a dozen options the workshop rarely varies. On the next pass I’d cut the standard chips to the four values that change in 80% of orders and push the rest behind “more options” — fewer choices, faster to the exact part.
Four moves, repeated on every project.
These aren’t websites in the sense of pages with pictures on them. Each one is a working part of somebody’s business — a shop, a quote desk, a race calendar — and it gets judged on whether the business works better afterwards.
Frame the constraint
The brief is rarely the problem. I go looking for the constraint nobody wrote down: the lead time, the drawing, the glove, the neck size.
- Funnel + search forensics
- Talk to the sales desk
- One sentence problem statement
Prototype on a real phone
In the showroom, the depot, or trackside at 11am in the sun. Throttled 3G, one hand, and somebody else’s device.
- Clickable builds, real data
- Tested on the client’s own phones
- Camera, WhatsApp, maps in the loop
Test with telemetry
6–8 people explain the why. GA4 explains the how much. I define the events before the build, not after it.
- Named, documented event schema
- Guardrail metrics agreed up front
- Performance budget from day one
Ship, then say what happened
A 90-day read-out with the primary metric, the guardrails, and an honest note on what didn’t move.
- Pre/post or A/B where possible
- Call logs and enquiry audits
- Written read-out for the client
The numbers I actually report.
Impact claims are cheap. These are the definitions I agree with the client before a project starts, so that a result can be checked — and the guardrail I watch to make sure the win is real.
| Metric | What it tells you | How I instrument it | Guardrail I watch |
|---|---|---|---|
| Primary action rate | One per project, named before design starts: add-to-cart, quote request, entry enquiry, drawing upload. | GA4 events with a documented schema; mobile sessions reported separately from desktop. | Lead quality — more enquiries that the sales desk can’t use is not a win. |
| Enquiry completeness | Share of enquiries that arrive with everything needed to quote — spec, size, drawing, quantity. | Monthly audit of the actual inbox, not the form fields. | Enquiry volume — fewer, better enquiries is fine until it becomes fewer, worse revenue. |
| Mobile task success | Can a first-time visitor finish the job on their own phone, in the place they’d really do it? | Moderated testing, 6–8 participants, their own devices, throttled connection, two rounds. | Time on task — success that takes three times as long is a failure. |
| Core Web Vitals | Whether the experience survives a mid-range Android on a weak connection — the actual median device. | Lighthouse mobile budget per template; CrUX field data after launch. | LCP on hero image — the prettiest gallery is worthless at 4.5s. |
| Search success | Whether people find the part, the piece or the event without leaving — zero-result and no-click rates. | GA4 site search terms matched against catalogue vocabulary. | Synonym drift — a search that only works in the company’s language. |
| Call deflection | How many repetitive phone enquiries a page removes: gate times, lead times, spec questions. | Call log tagged by the client’s staff, compared month over month. | Hiding the phone — never bury a number a trade buyer wants to dial. |
What I bring to a project.
Design and build in one pair of hands: the interaction thinking, the prototype that proves it, and the production site that ships it — measured afterwards.
Mobile interaction design
Thumb zones, sticky bars, bottom sheets, filter rails, camera capture and state models. Designed at 360 px first and then allowed to grow.
Catalogue & commerce structure
Product filtering, spec tables, enquiry shortlists, variants and pricing — the information architecture that decides whether a big-ticket or B2B catalogue converts.
Performance & Core Web Vitals
A performance budget per template, image strategy, and a Lighthouse check before launch. Mobile-first is a performance claim as much as a layout one.
Measurement
Event schemas agreed before build, GA4 funnels, documented 90-day read-outs. I’d rather report a flat number honestly than a good one vaguely.
Build & handover
WordPress and WooCommerce builds the client’s own team can edit, WhatsApp Business integration, and documentation written for the person who inherits the site.
Accessibility as craft
Contrast that survives sunlight, 48 px targets for gloved hands, real labels, keyboard paths. Built in from the first screen, not audited in at the end.
Pieter changed the order of the information on a product page and our WhatsApp enquiries doubled. He was the first person to ask us what customers ask us on the phone, and then he built that into the site.
We used to spend three emails getting one enquiry to the point where we could quote it. Now the specification arrives on the first one. That change was worth more to us than the design.
Let’s talk about your mobile experience.
Pedas Design Studio — design and build for South African businesses that sell real things. WhatsApp is usually fastest; I answer during working hours and quote within two days.
