Automate the decision trigger, not the decision
Most automation advice is a list of things software can do unattended. That is the wrong frame for a small business, because the tasks that eat your week are rarely the ones a machine can finish end to end. They are the ones where a human has to notice something first — that stock is running low, that an invoice is thirty days overdue, that a batch expires next month — and noticing is the part that actually fails.
So the useful question is not "what can be automated" but "what am I currently required to remember." Anything on that list is a candidate. Everything else can wait.
The order to switch things on
Turning on every alert in week one is the classic mistake and it has a predictable outcome: within a fortnight nobody reads any of them, and you are worse off than before because you now believe you are covered. Switch things on in this order, roughly one per week.
- The nightly backup. Not glamorous, entirely automatic, and the only item on this list whose absence can end the business. Confirm it is running and that you know how to restore, before anything else.
- The daily sales summary. One message each evening with today's takings, transaction count and payment-method split. This is the alert that gets read, because it answers a question the owner already asks every night.
- Fiscal transmission failures. You want to know the same day when a receipt did not reach TRA, not at month-end when the gap is a hundred receipts wide.
- Low stock, on your top twenty items only. Not the whole catalogue. See below.
- Overdue invoices, once your receivables data is actually clean. Automating chases on wrong balances damages customer relationships faster than manual neglect does.
- Everything else, when a specific problem justifies it.
Setting thresholds that people will still trust in month three
A low-stock alert set to "10 units" for every product is noise, because ten units of a fast mover is an emergency and ten units of a slow mover is six months of cover. The threshold should be a function of how fast the item sells and how long it takes to replace.
The workable rule is: reorder point equals average daily sales multiplied by lead time in days, plus a safety buffer sized to how erratic that item's demand is. For an item selling 8 a day with a 5-day supplier lead time, that is 40 plus buffer — nowhere near 10. Our free reorder point calculator does this per item, and the EOQ calculator helps decide how much to order once you know when.
Set thresholds only on the items that matter. In most Tanzanian retail and wholesale businesses a small minority of lines produce the large majority of sales; alerting on those and reviewing the rest on a periodic stock report gives you almost all the benefit with almost none of the noise. Review the thresholds quarterly, because demand moves with the seasons, the school calendar and the rains.
Choosing the delivery channel deliberately
Reports and alerts can reach you by email, SMS or WhatsApp, and the channel changes the behaviour more than the content does. WhatsApp is read, which makes it right for anything requiring action today and wrong for anything routine — a WhatsApp message that does not need a response trains the recipient to ignore the next one. Email is right for scheduled reports you will read when you sit down, and for anything you may need to search for later. SMS is the fallback that works on any handset and any network, which makes it the correct channel for genuinely critical alerts to someone who may not have data.
Route by recipient as well as by urgency. The owner wants the daily summary and exceptions. The store keeper wants stock alerts and nothing else. The accountant wants the monthly pack. Sending everything to everyone is the same failure mode as sending nothing.
What runs without you touching it
- Tax arithmetic on every transaction. VAT at the standard 18% computed on the line rather than added by hand at the end, which is where manual errors concentrate. Cross-check any case you are unsure about with the free VAT calculator.
- Statutory payroll deductions. PAYE, NSSF at 10% employee and 10% employer, NHIF, SDL at 3.5% of gross payroll and WCF at 0.5%, calculated each run rather than maintained in a spreadsheet that one person understands. Model the totals with the full payroll calculator, PAYE calculator and NSSF calculator.
- Stock movement on every sale, receipt and transfer, so the stock figure is a consequence of activity rather than a separate thing someone updates.
- Fiscal receipt transmission to TRA, with retry when the connection was down at the moment of sale.
- Running balances for customers and suppliers, updated as documents are raised and payments applied.
- The nightly backup.
Automation that goes wrong, and how
Three failure patterns are worth naming because they are common and avoidable.
Automating a broken process. If your customer balances are wrong, automated reminders send wrong demands to real customers at scale. Fix the data first; automation multiplies whatever it is pointed at, in both directions. Alert fatigue. Covered above, and the most common of the three by a wide margin — the cure is fewer alerts with better thresholds, never more alerts with louder wording. Silent failure. An automation that stops running usually stops quietly. Once a month, verify that the things you rely on actually fired: check that last night's backup exists, that yesterday's summary arrived, and that the fiscal queue is empty. Ten minutes monthly.
Where the hours actually come back
The time savings people report come from a small number of specific places, and it helps to know which so you can check whether you are getting them. Compiling the daily and monthly reports by hand disappears entirely. Manual stock checking shrinks to spot checks and a periodic full count. Chasing payments becomes reviewing an exception list rather than reconstructing who owes what. Tax computation stops being a monthly event. And receipt filing largely disappears once documents are generated by the system that recorded the transaction.
What does not disappear, and should not: deciding what to stock, deciding who to extend credit to, and deciding what the numbers mean. Automation buys you time to do those properly. That is the actual return, and it is a better one than the hours saved.
Related reading: the mobile app guide for where alerts land, the getting started guide for initial setup, and inventory, finance and HRM for the modules these automations run inside. Plans start at TZS 25,000 per month with onboarding included — see pricing, or ask us what to switch on first for your setup.
