Visa Consulting and Analytics ran its Retail Spend Monitor over the seven weeks from 1 November 2025. Online retail spending rose 49.9% year on year. E-commerce accounted for 11.5% of total retail spend across the period, with in-store taking the rest.

Total retail growth over the same festive season was modest, in the low single digits.

That gap is the planning problem. Revenue grew a little and the work content grew a lot, because the two are not the same thing and most peak models only measure the first one.

Rands and picks are different units

A store replenishment order is cases and pallets. One pick moves 24 units. The carton is the supplier's carton, the label is a pallet label, and the destination is a receiving dock that expects you.

An online order is one to three eaches. Each one needs its own pick, its own box selection, its own pack, its own label, its own manifest line, and a delivery to a residential address that may not be there.

Same rand value, very different work.

So when online revenue grows 49.9% at a roughly stable basket value, order count grows about the same amount, and every one of those orders is a full cycle. Model peak on turnover and you will size labour, packing stations and carton stock for a volume that does not match what arrives.

Size peak in picks, orders and cartons. Then convert to rands at the end if finance needs it that way.

What to decide in September rather than November

Three decisions, in the order they take time to implement.

1. Pick strategy for eaches. Discrete order picking is the default in most South African DCs because it started as a store replenishment operation and the online channel was added on top. At low online volume that is fine. At 11.5% of spend and growing at 50% it stops being fine, because walking distance per unit picked is several times higher than it is on a case pick.

Batch and cluster picking change that, and they are configuration and process changes rather than capital ones. Optioryx publishes 20% higher picking productivity and 40% shorter walking distances for its Pulse Pick module, which handles routing and clustering. Treat that as direction and measure your own walking distance per order line before and after.

The reason this is a September decision is that you cannot test a new pick strategy in the last week of November. It needs to be live, measured and corrected while volume is still normal.

2. Carton range and box selection. An eaches operation ships far more cartons than a case operation, and each one is chosen by a packer under time pressure from whatever range you gave them. Too few sizes and you ship air. Too many and selection slows down and error rates rise.

Optioryx publishes 15% lower shipping costs, 30% less carton used and 35% less air shipped for its 3D Cartonization module, and pick-to-box removes the repack step entirely by assigning the destination carton at order release. With diesel up R4.32 a litre since 5 August, the air in those cartons costs more this peak than it did last peak.

3. Labour, and when you commit to it. South Africa is short more than 59,000 skilled logistics workers, and temporary staff for peak are contracted, screened and inducted well before they are needed. A pick strategy decided in November arrives after the people have already been trained on the old one.

The January half

The returns wave lands on the same eaches, and it is the part of peak that gets planned last.

We covered the shape of the extended October to February peak in an earlier article, including category returns rates and the capacity returns consume. The point worth adding here is about touches rather than volume.

An each that took one pick to go out takes receipt, inspection, disposition, refurbishment or repack, and put-away to come back. That is four to six touches against one, on a unit that generated no new revenue. Returns processing competes for the same floor space, the same scanning hardware and the same people as January replenishment, and January replenishment cannot wait.

A DC that has not decided where returns will be processed by the end of September will process them in a receiving lane in January, which is how the lane gets blocked.

How this fits alongside your existing WMS

None of the three decisions above requires a new warehouse management system.

Pulse sits on top of your existing WMS through an API. Optioryx states plainly that it is not a WMS and does not replace one. Integration runs three ways: manual file import and export, flat files over CSV or SFTP, or a direct API into your WMS, TMS or ERP. Optioryx reports 65 or more warehouses live across 28 countries and holds ISO 27001 certification.

That is the practical route at this point in the year. There is no version of a WMS replacement that lands before 27 November, and there is no April capital cycle between now and then either.

Three things worth doing this month

1. Restate your peak forecast in picks and cartons. Take last November and December actuals, split store replenishment from online, and grow each at its own rate rather than at a blended revenue number. The online line is the one that compounds.

2. Measure walking distance per order line. Most WMS systems will give you this, or it can be derived from pick confirmations and the location master. It is the single number that tells you whether your pick strategy still fits your order profile.

3. Name the returns area and the returns owner now. Not the process, just the square metres and the person. Both are harder to find in January than in September.

The timing

Black Friday is 27 November. That is 10 weeks from today.

Ten weeks is enough to change a pick strategy, rationalise a carton range and clear a returns area. It is not enough to do all three in November alongside the volume they were meant to absorb.

Start with the forecast restated in picks. If the online line is growing at anything close to 50%, the other two decisions follow from it.

Sources

The Visa figures cover a seven-week period from 1 November 2025 and use Visa payments network data plus survey estimates for other payment methods. They are a spend sample, not a national census.