Friday, March 20, 2020

Ocado shuts its website. Let's brainstorm!

So today Ocado shuttered the rest of its website. I've long pondered their impressive operation, so let's do some arbitrary brainstorming based on the sparse information I have.

  Neighbour Groups - Improve route efficiency and customer absence

The idea here is pretty simple - that a group of neighbours voluntarily connect their accounts, and permit the other members to know when their delivery will be. That would then trigger the availability of free delivery into the same slot.

Ocado wins here in two ways: firstly, by the obvious route of increasing the clustering of deliveries such that the cost for the additional deliveries is minimised, permitting faster route completion, lower fuel and staff costs per delivery, or in the oversubscribed corona scenario playing out, to increase the deliveries per shift.

The second way they win is that in the event that a customer in such a group is out, the driver has a list of alternative people to deliver to - either by default or by electing to do so, the customer can permit delivery to another member of the group. This should reduce the time and expense wasted by absences.

A key factor right now that would drive take up by customers would be that they could book these slots if any of their neighbours already had that slot. So joining or spurring a neighbour group would be a way to get a precious de delivery slot otherwise unavailable.

Ocado could also offer a lower minimum spend on these slots, as the operational cost is barely higher than someone editing their order.

Segregate customers into alternating groups to cut load

Suppose we divide customers into red and green. The red customers could order deliveries on red days, and the green customers on green days. The website would also only be available for red customers on red days, and green customers on green days. The split can be easily done based on customer ID.

This should be pretty easy for customers to grasp, and once they have tried to log in on the wrong day, they are unlikely to keep trying on the wrong days. They could even be given the Red Login Page and the Green Login Page on static servers which would inform them it was the wrong day without touching the database.

A further benefit could be derived by optimising the division of customers to group them geographically - this would lower average miles per delivery, and thus increase the number of deliveries per shift. People in a given area would all know which colour applied to them.

The division could be done into more groups, though that would make it harder for customers to know which days were theirs - a 2 day pattern fits 2 weeks.

Essentially the entire operation could in theory be divided this way, including stock allocations and databases, as it would become 2 Ocados.

 Predefine approximate delivery schedule and offer the few slots closest to it

At present when I choose a slot I am offered lots of choice. Normally that would be important to Ocado's competitive offering. But this is a different world. By only offering the few slots which are closest to the planned route, Ocado can really heavily optimise the number of deliveries it can make.

An example plan might be to take all the orders over a given few days, remove duplicate addresses, and work out what route would have roughly minimised the total distance or would have minimised the total delivery time. Or alternatively put in a set of dummy orders which anchor the available schedule and leave the system in a permanent state of filling gaps.

Ocado is now in the driving seat, and the customers are far less fussed about particular slot availability.

This approach also gives Ocado a useful lever to pull- they can increase or decrease the number of slots they offer from this ranked list, so that they can respond to further changes in demand or capacity.

It could also permit longer routes with larger vehicles to mix in with the standard set.

Offer dry deliveries

Ocado cannot rapidly ramp up its delivery force because of the need to refrigerate the load. But many people will just need storecupboard items, so Ocado could offer dry shops which are all non fresh, and could be delivered by a more temporary workforce drafted in for capacity. These vehicles could also be larger and on longer routes, perhaps with an additional delivery person to cut unloading times.

This offering could be entirely parallel with the main Ocado site and network, even using a duplicate website which only shares user authentication with the main site. For Ocado technologies it would be a demonstration of their platform and its capacity to adapt.

Warehouseless logistics

What is a lorry but a warehouse with wheels? Given how briefly some products are stored, it is conceivable that for some consignments it would be possible to transfer goods directly fr the original lorry to the delivery trucks without a warehouse in between. Temporary interchange points could be set up in strategic locations, entirely bypassing the main warehouses. This would work well with the dry delivery idea, and could be rapidly expanded - ultimately this could look like emergency humanitarian aid with only a handful of product lines being distributed.

A pasta, bog roll and soap delivery service would be gold in the south east right now.


A very straightforward way to improve unloading times would be to put an unpacker with each driver. This would improve turnaround at each address. Delivery costs might have to go up as a result, but at present I think that would be accepted by the customer base.

Raise delivery charges

 This is a pretty simple step to address multiple facets. The hope would be that the extra cost would incentivise people to voluntarily reduce their delivery frequency, whilst offering a stronger incentive to take the reduced cost delivery slots which I assume are near existing deliveries. If combined with the unpacker role above, there is a straightforward narrative to give the customers when the change is announced.

In sum:  Ocado is in the position to radically increase capacity by turning down some of the niceties traditionally offered

Ocado, and Ocado Technology, have built a system on some underlying assumptions and requirements about delivery slots and customer expectations, and that has been very successful at producing a logistics system. But now and for the foreseeable future, customers are willing to drop their expectations and accept a compromise, which opens the doors to some pretty huge changes to the operating model which could multiply the number of slots and deliveries per van per shift many times over.

I would probably rank these ideas time wise as: raising delivery charges (immediate), predefined routes (hours-days), segregate customers (days), unpackers (days-weeks), then the rest are weeks to months.

No comments: