By classic E-Com, I mean commerce at the junction of online and offline. There is always a site or mobile application where an order is made, automation systems, for example, a warehouse, and there is an offline part-operating activities, in the vast majority of cases it is Last Mile delivery. In addition, there are monitoring systems to control the effectiveness of all online and offline processes.
We have introduced the Gist Freimvork for quick and transparent decision -making, and the Inner Source approach for autonomy in development, and now all the elements of the mosaic are collected. It's time to round up, it was an interesting race, thank you for your participation! Let's summarize.
conclusions
The experiment is a transparent and effective tool to achieve the goal.
Conducting it in a real environment, we study our audience, which allows each time to make changes more and more useful, and also understand why not all features should remain in the Prode.
But so that chaos does not happen, you need a clear process, distribution of roles and the appointment of a responsible one.
In the active phase of the experiment and when analyzing, it is worth monitoring several key metrics at once and do not forget to draw a conclusion (in our case, press one of the three buttons).
The adoption of quick decisions and autonomy in development helps to achieve results and maintain the motivation of the team.
In some companies, the R&D team is engaged in fundamental research of new technologies. The purpose of such research can be both the development of current processes and completely new verticals in business. For example, a few years ago, blockchain was such a technology.
We will immediately indicate that we are talking about something else. R&D in Delivery Club is a solution to applied problems. Our business is growing rapidly, the number of orders is growing, and most of our internal processes are based on data and algorithms. With an increase in the volume of input data, some algorithms cease to be efficient, and those components that we were completely satisfied with yesterday, today we often find inoperable.
As you may have guessed, such problems are often weakly deterministic, and, as a result, are accompanied by many uncertainties from all sides. Let's highlight the key ones:
Let's take a closer look at what we are faced with, and whether it is really difficult.
How to reach the goal
We always clearly understand the business goal - in which direction we plan to move, where we need to go. But it is far from always immediately clear how to approach this goal when it comes to research tasks. What strategy to take: through experiments or immediately a big rollout of features? Building simulated environments or experimenting in combat? The choice is large - the eyes run wide.
Also, quite recently, we started an experiment with a new ML model in which we are trying to improve Pickup time. In addition to the same data as for Last mile, we transmit the average bill of the order, thereby better predicting the cooking time. And this, as we know, is reflected in the entire forecast.
In the future, we have a full transition to forecasts from the ML model for all segments of the route.
In fact, there are still many places to improve forecasts, but now there is a problem that I would like to talk about.
We do not control the route that couriers choose to deliver the order. This may mean that we calculated the time on one way, and the courier chose another. Naturally, the forecast will be inaccurate. Perhaps one of the solutions will be hints to couriers which way to choose to control the time on a particular route.
In addition, snowfalls and frosts greatly affect the speed of movement. There are already plans to use their own weather service to adjust this speed.
