Why thirty days
Most automation in a Zimbabwean SME fails for reasons that have nothing to do with software: the process was never written down, nobody owned it, the internet dropped for two days, the person who understood it went on leave. A thirty-day pilot is long enough to hit all of those and short enough that hitting them costs little. It is also long enough to produce a number the owner can believe.
The census context matters here. ZimStat’s economic census found 87.9% of Zimbabwe’s 204,798 establishments are micro-enterprises and 76.1% are informal. In a business that size there is no project team; there is the owner and one keen person. The method has to fit that.
The five rules
- One process. Not “customer service”. “Confirming WhatsApp orders.” Not “finance”. “Sending statements to account customers on the 1st.”
- One metric, counted the same way before and after. Unanswered chats at 17:00. Hours from quote request to quote sent. Days from invoice to payment. Unfilled shifts.
- One owner with two hours a week blocked in the diary. If that person does not exist, the readiness score will have told you; fix that first.
- The simplest version that could work. Free features and a spreadsheet before any subscription; a subscription before any custom build.
- A decision on day 30, written down, with the numbers.
The plan
Week 1 — Define and baseline. Write the process on one page using the SOP template. Count the metric every day for five working days, by hand, and write it on the wall. Note the exceptions you meet; they are the spec.
Week 2 — Build the simplest version. For a WhatsApp process: greeting, away, quick replies, labels. For quotes: the template and the pipeline sheet. For invoicing: the quote-to-invoice flow with a US$1 test. For rosters: the rules tab and the red-cell checks. Nothing that needs a developer; nothing that needs a card, unless a free tier does not exist.
Weeks 3–4 — Run and count. The owner’s two hours are for reviewing the count, not for building. When something breaks — it will — the champion fixes the smallest thing that makes it work again and writes down what broke. Internet and power are the usual culprits: with 84.55% internet penetration but mostly on mobile, and household internet spend around US$10 a month, a pilot that assumes a permanent connection will fail on a Tuesday. Design for the outage: the sheet works offline; the away message covers the gap.
Day 30 — Decide. Compare the two-week count with the week-1 baseline. Then apply the rules below.
The decision rules
| Result | Decision |
|---|---|
| Metric moved by the amount the ROI calculator needed, and the process survived an outage | Go. Make it the standard; put the subscription on annual billing; pick the next opportunity from the finder. |
| Metric moved, but only when the champion was watching | Adjust. The process is right; the ownership is not. Assign it, retrain, run 30 more days. |
| Metric did not move | Stop, and say why in one paragraph. The paragraph usually names a different process that should have been piloted. |
| The pilot was never really run | Stop. The readiness score was right. Do the foundation work and come back. |
A worked pilot (illustrative)
A hardware shop in Chitungwiza with six staff picks “confirming WhatsApp orders”. The metric is chats with no reply at 17:00, counted by the owner’s daughter from the business phone. Week 1 baseline: 7, 5, 9, 6, 8 — an average of seven a day, on about sixty messages. The exceptions list has three entries: customers asking for credit, customers sending a photo of a part with no name, and voice notes in Shona that the new assistant cannot follow.
Week 2 builds nothing more than the free app: a greeting message, an away message scheduled for 17:30 to 07:30 and switched on by hand during load shedding, ten quick replies from the reply playbook, and a label called “Follow up” that the assistant clears before leaving. The credit exception gets a rule: “reply with /account and tag the owner”. The photo exception gets “/photo”. Voice notes go to the owner’s nephew, who speaks Shona.
Weeks 3 and 4: 3, 2, 4, 1, 2 and then 2, 1, 1, 3, 0. The average falls from seven to about two. On day 30 the owner puts 5 recovered enquiries a day into the ROI calculator at her own conversion rate and margin, sees a payback measured in days, and writes the paragraph. The next process is the order sheet, because the confirmed orders are now arriving faster than the book can hold them. Total cash spent: nothing. Total time: about thirty hours, most of it counting.
What it costs
Almost nothing in cash — the point of the simplest version — and about 40 hours of people’s time over the month. Put those hours into the calculator as setup cost at the owner’s rate; if the pilot cannot pay that back inside a year on its own metric, it was the wrong pilot.
What to write down at the end
One page: the process, the metric before and after, what broke and how it was fixed, the decision, and the next process. Keep it. Five of these pages are a better automation strategy than any consultant’s deck, and they are what an assessment will ask for first.
The checklist
Use the 30-day pilot checklist; it has a tick box for every item above and a place for the daily count.
Sources
- ZimStat 2023 Economic Census results, reported by The Herald (accessed 2026-09-14)
- POTRAZ Q4 2025 Abridged Sector Performance Report, reported by TechnoMag (accessed 2026-09-14)
- MISA Zimbabwe — Internet affordability and access in Zimbabwe (accessed 2026-09-14)