Home › Roadmap
What to actually do
A staged plan, and what it costs
Nothing here requires doing everything at once, and the first stage requires no software at all. The order matters more than the speed: each stage makes the next one cheaper and safer.
Every price below is a placeholder until the team verifies it. Check each one against the vendor's own published pricing, record the date you checked, and cite it on the sources page. Software pricing changes constantly and a client will check. A figure you cannot source is worse than no figure.
Stage 0 — Before spending anything
Two weeks. Cost: the owner's time.
Map where every fact enters the business
The seven-row table on the data page, filled in for this shop. You cannot integrate what you have not listed.
Decide the single source of truth for each fact
Stock, price, what a customer owes. One authoritative place each, written down.
Count the stock properly, once
Any system will inherit whatever figure it is given. Starting on a June guess means paying for a system that is wrong on day one.
Clean the customer list
Merge the duplicates, fill in the missing purposes, delete what there is no reason to keep. This is also the first real POPIA step.
Write the two lists
What the shop will change to suit software, and what software must accommodate. See Making it fit. Do this before any demonstration.
A business that skips it buys software to manage a mess it has not described, migrates the mess into it, and then pays a monthly subscription for the privilege.
Stage 1 — Point of sale and stock
Months 1–2. The stage that returns the most.
What changes
- Every sale is captured once, at the counter, linked to a customer and a product
- Stock moves automatically when a sale is rung up
- Prices live in one place
- Tax invoices print with the required particulars every time
What it is worth
- The stock figure becomes real, so reordering stops being a guess
- Dead stock becomes visible — R23 798 of it in the demonstration data
- Shrinkage becomes visible — R3 616 over twelve months
- Month-end stops being a weekend
The point of sale comes before accounting as everything else is built on the sale. You cannot do the accounting if the sales records are incorrect. As long as the sale is captured once and correctly linked to a customer and product then the stock, prices and VAT will follow. The first two weeks of integration will be slower so ringing up a sale on a screen will take longer than a book. Implementing in a quieter week rather than a busier week will give time for the staff to be trained efficiently without slowing down operations. For those first two weeks parallel running with the old book and the new system is recommended for when efficiency is needed whilst learning the new system and to double check or cross reference any errors that might have taken place that will be more easily corrected.
Local vendors are the most suitable option such as Yoco where their core plan costs R0 a month and has a point-of-sale software included that charges 2.3% for in person local card payments. On the Core plan, up to R50 000 of card turnover a month excluding VAT with lower rates above that.
Yoco card machines cost R1 400 once off as of 12 October 2026. Loyverse offers a free point of sale that records transactions offline. The true costs are the card fees so model it on the shops card turnover and testing the products against how Lerato sells, by the metre and by the bag before signing.
www.yoco.com. (n.d.). Yoco fees, costs & transaction rates. [online] Available at: https://www.yoco.com/za/pricing/ [Accessed 12 Oct. 2026].
Yoco.com. (2026). Card machine prices & transaction fees | yoco. [online] Available at: https://www.yoco.com/za/card-machines/pricing/ [Accessed 12 Oct. 2026].
loyverse.com. (n.d.). Free POS software. Point of sale system. Loyverse POS. [online] Available at: https://loyverse.com [Accessed 12 Oct. 2026].
Stage 2 — Customers, credit and POPIA
Months 2–4.
What changes
- One record per customer, with the duplicates merged
- Credit limits enforced at the counter, not remembered
- Statements produced instead of written out
- Consent, purpose and retention recorded on every record
- The Information Officer registered and a counter notice up
What it is worth
- R99 962 currently owed across ten accounts becomes visible and chaseable
- Overdue accounts stop being discovered by accident
- A POPIA access request becomes answerable in one click
- The R12 000 extended to an unverifiable record stops being possible
Stage 3 — Accounting, properly joined
Months 4–6.
Connect the point of sale to whatever the bookkeeper uses, so sales, VAT and debtors flow across without being retyped. Prefer two products that already integrate over two "better" products joined by a developer — see the four ways to integrate.
Stage 4 — Analysis, and only then
Month 6 onwards, once there is trustworthy history.
Margin by product and by customer, seasonal patterns, reorder suggestions from actual sales rate. Everything on the Analytics screen. Attempting this before Stages 1 and 2 produces confident charts drawn from numbers nobody checked — which is worse than no charts, because somebody will act on them.
What it costs
| Item | Once-off | Monthly | Notes for the team |
|---|---|---|---|
| Point-of-sale software (per till) | R ____ | R ____ | Check per-user pricing at four users, not one |
| Till hardware, scanner, printer | R ____ | — | Price a scanner — it is what makes capture fast enough to stick |
| Card machine | R ____ | % per transaction | The transaction fee is the real cost; model it on actual card turnover |
| Accounting package | — | R ____ | Ask the bookkeeper what they already use before choosing |
| Data migration and setup | R ____ | — | Often underestimated. Most of it is cleaning, not loading |
| Staff training | R ____ | — | Count the hours the shop is short-staffed, not just the trainer's fee |
| Backup and connectivity | R ____ | R ____ | Include a UPS or battery for the till. Load shedding is not optional |
| POPIA registration | R ____ | — | Verify whether a fee applies; cite the Information Regulator |
| Indicative total, year one | R ____ | R ____ |
The total cost of ownership includes the once off fees including the monthly subscriptions which would be a total of R13 130 to start up and R2 200 to keep it running. The cost of switching vendors later would include exporting the data, retaining staff and running two systems for weeks. The slower first two weeks is not priced here as it depends on how many staff hours are lost. For an honest payback argument, we can evaluate the current system against the shops leakage with the dead stock being able to have better visibility of the inventory which would allow the store to clear unpopular stock, it is a once off event. Annual shrinkage can be checked with tighter stock tracking at the tills reducing stock loss that is unrecorded. For outstanding debtors automated customer accounts and clear sales logs shorten payment cycles and recover tied up working capital.
Adding the three together equates to R127 376 overstates the payback. Dead stock is a one off, and it would be cleared at less than cost, Shrinkage is a yearly loss that no system removes entirely. Debtors' money is already earned, so collecting it improves cash, not profit. The honest argument is that one off cash from stock and debtors plus the month end weekend saved can cover year one, but shrinkage will not. Note that a system does not make an unprofitable store profitable, the only benefit of this is to have more accurate numbers of sales and manage stock more efficiently and to be able to make more informed decisions to reduce leakages.
What could go wrong
| Risk | How likely | What it costs | What reduces it |
|---|---|---|---|
| Staff go back to the book on busy days | High | The whole investment. Half-captured data is worse than none. | Involve staff in choosing it; start with what makes their day easier; watch for it openly in month one |
| The old mess is migrated into the new system | High | Paying a subscription for the same problem | Stage 0. Clean first, migrate second, and mark migrated records as unverified |
| Vendor lock-in discovered late | Medium | Trapped, or an expensive extraction | The export test before signing. Export once a quarter and keep the file |
| Load shedding stops trade | High | Lost sales at the worst moment, and a reason to abandon the system | Offline capture, battery-backed till, a written manual fallback everybody has practised |
| Only one person knows how it works | Medium | The shop stops when they are on leave | Two people trained on every task; short written procedures |
| Scope creep — trying to do all four stages at once | Medium | Nothing finishes; confidence is lost | One stage at a time, with a decision point before the next |
| A POPIA complaint before the basics are done | Low | Regulatory exposure and reputational damage locally | Stage 2's compliance steps are cheap; do them early |
Risks that matter the most is the staff drifting back to the book on busy days. Having a record half captured results in the investment of the new system a waste. To mitigate this more involvement can be put towards using and learning the new system to bring it up to speed with keeping book records.
Checking two to four weeks after the migration period if the system is really being used during busy hours is necessary.
Migrating the old mess onto the new system is the second high risk where this could result in paying more in subscriptions to resolve it. The first step should be cleaning before migrating and marking everything as unverified until someone has checked it.
The third is load shedding which would stop sales at any moment. This would result in abandoning the system and missing out on sales and to fix this we would need to get a system that has prioritised offline capture and have backup power when there is loadshedding so that operations are not halted.
Our recommendation is starting stage 0 with the cleaning before stage 1 as it would be free but only cost two weeks of time at most then move onto the migration. Stage 1 is the point-of-sale system with stock run in parallel with the book for the first two weeks. The next is registering an information officer which is free and brings the R99 962 of debtors into view.
This is stage 2 on the roadmap. Stage 3 would be treated as a decision until the bookkeepers work shrinks and everything else has been managed accordingly. Stage 3 is accounting and stage 4 is analysis which only comes once there are 6 months of verified data.
What could be left out deliberately would be spending money on AI forecasting tools or a full ERP system. Having a tool forecast a years' worth of unverified numbers will lead the business into the wrong decisions, and an ERP system is not needed yet for a four-person shop. It would only be worth visiting once there is at least 6 months of verified trading history.
Decision points
Each stage ends with a question, not an invoice. If the answer is no, stopping there is a legitimate outcome — and a cheaper one than continuing out of momentum.
- After Stage 0: is the data clean enough to migrate? If not, stay here. It costs nothing but time.
- After Stage 1: are staff actually using it, three months in, on a Saturday? If not, fix that before buying anything else.
- After Stage 2: can you answer a POPIA request and produce an accurate debtors list? If not, Stage 3 will not help.
- After Stage 3: does the bookkeeper's work take meaningfully less time? If not, the integration is not doing what it was bought for.