In this post, we will share with you the updates that took place in Mindbox for March and April.
Changes in client data platform:
- New menu and interface structure.
- A page where we show how much you spend on Mindbox.
- Improved product and order importing.
- Protection from fake email addresses and numbers.
- Export of clients fromsegment to file.
New in segmentations:
- We learned how to give out virtual loyalty cards.
- Combined random and fixed triggers.
- Recommendation widget constructor.
- Improved recommendation algorithm.
- New recommendation page.
- Visual recommendation viewer upon set up.
Client data platform
Menu structure change
We renamed consumers to clients. Added a new section «Data», where info on clients, their actions, segments, and products falls:
The «Campaign» section now keeps info on all campaigns, loyalty rewards and special offers, algorithms. Use the «Loyalty rewards, and special offers» to search for promo codes and discount cards, and the «Algorithm» section for item recommendations.
Integrations received their own section, «Integrations». It contains settings that are needed for the integration but rarely used once it is complete. The menu contains a dashboard for integrations, links to additional fields, and product lists. This section also contains action templates, channels, and contact points:
We updated the «Points» section in the unified client profile. This menu can be used to check your balance for each account and point accruals.
We received feedback that the changes were unexpected and too large. We apologize and promise to give a heads up about such changes.
We show how much you spend on Mindbox on a separate page
Occasionally our clients want to know how much they pay and for what. In order to receive this information previously, they would have to write to one of our managers, who would then submit a request to accounting.
We excluded human involvement from this process and created an «Account details» page. On this page, information regarding your project is published: you can get a quote for a new rate, ask for new services to be connected, view previous bills.
Sped up the import of products and orders
Previously in certain scenarios imports could get in the way of one another, while during new product imports, the speed of their upload to the system would fall. Now, imports will bother each other less, and the upload of new products to Mindbox will be higher.
Added protection from fake email addresses and phone numbers
Clients experienced several problems due to fake contacts:
- Contacts like +7(111)1111111 and email@example.com generate a merger of clients in the system. Client data is ruined: profiles that have no relationship to one another become merged. Client profiles with normal and fake contacts suffered quite a bit.
- Client importing didn’t work well if there was a lot of repeat data. Contact firstname.lastname@example.org, important 5,000 times resulted in 5,000 accounts being merged, with the important process freezing occasionally. The import required to be stopped, the problematic contact had to be found, and the file cleaned. This process would occur each time.
- «Garbage contacts» useless in the base as is.
To solve the issue, we made a blacklist of incorrect contacts. Contacts from the list cannot be added to the base via import and operations. The most popular fakes were already added to all projects of Mindbox clients. If the import froze due to the repeated merging of contacts or fakes constantly being sent through the API. Add problematic contacts to the list, and in the future, they will be automatically rejected. This will result in fewer issues with importing garbage contacts into the database.
Instructions on how to blacklist a number.
Simplified the export of clients from a segment to a file
We simplified the exchange of data about a client to the display advertising systems. For this, we added a new step in v3-operations: «Export – load clients into a file». We wrote about this in more detail in our developer portal.
To download through our API, you must first send a POST request, that will leave a task to formulate the file. Followed by a GET request to receive a link to the file. There will be a CSV file with the MindboxId, email, mobile number, gender, and date of birth of a client in the download.
We started to limit heavy segmentations
When segments refresh because of a heavy event or there are too many events (example, during an order import), segments use more and more resources and start to suck it away from other mechanics – leading the productivity of other mechanics to fall significantly.
For this to not happen, we monitored and created an automatic slowdown of periodic and problematic segmentations. The problematic segmentation is limited in its access to resources, it works slower but does not get in the way of other mechanics. If after a small period of time everything is okay with it, we return the resource and regular status.
Added a segmentation view
We added separate settings for viewing and editing segmentations. Now, depending on the clearance given to a manager or marketer, he can both view and edit.
On the screenshot, we present the interface for viewing segmentations when a manager does not have edit clearance. A manager can view it, but the button to save changes is inactive.
Allowed the deletion of manual mailings
Previously, mailings were forever stored after being sent, and nothing could be done with them. Now, the sender of manual mailings gets a «Delete» button. Prior to deleting a mailing, we will check if it is being used in other entities: filters, segments, and triggers. If it is being used, then we won’t let it be deleted in order to prevent messing up the work of other mechanics.
Together with the mailing, all actions and reports tied to it are also deleted. This means that after it is deleted, you will not be able to see how much money the mailing yielded to the business, what it’s open and click rates were, and which clients it was sent to.
Added a summary report in mailings
We added separations for data from online and offline channels to the summary. Additionally, information on the percent of unsubscribers from each mailing is also presented there, as well as the average income per recipient. In the report, we provided a link, which can be used to move to any mailing from the report.
We figured out how to give out virtual loyalty cards
We integrated with OSMI Cards and can now issue a virtual card upon a buyer registering for the loyalty program. The virtual card can be issued in addition to a plastic card or without it, in order to cut back on card emission costs. The virtual card is issued via SMS, a QR-code, email, or on-site.
The virtual card can be used to show and refresh the actual balance of a client in real-time, as well as personal data and any other information.
We can send push notifications to the card and inform the buyer about deals, new offers, and send an authorization code at the register.
Virtual OSMI cards can easily be integrated with the Mindbox Loyalty program. For the register to recognize a client’s card through a wallet-card or a mobile application, no additional developments are needed. From the perspective of a register, this is seen as a regular plastic card.
Combined random and fixed triggers
The functionality of random triggers was added to the previous «Fixed triggers». Now there is only one option in the menu for this, known as «Triggers».
For all triggers, there is now the ability to add groups of steps and their probable completion rate, a random trigger can be reconfigured to launch upon a specific event:
(1) Probability with which this group of steps will be completed.
(2) The steps themselves, what the trigger should do with the client.
Creating a recommendation widget in the customizer
Our partners from Popmechanic developed an interface for creating a recommendation widget. The widget is used to show clients recommendations from Mindbox. Within the constructor, one can create a widget, establish parameters, call operations, and launch them. Now, the widget can be made without the help of developers.
Improved the personalized recommendation algorithm
Ran tests of the algorithm. We evaluated how well the new algorithm predicts what a user will buy in the future. According to historical data, we were able to predict which good a client will buy next month with a high degree of success. The new algorithm is two to three times more effective with predictions.
Also, we taught the algorithm to automatically select the best criteria, based on which recommendations will be made, for specific projects. This will allow to automatically adapt a recommendation mechanism for a specific project and put forward the most effective recommendations.
New page with a recommendation list
The new page of item recommendations is located in «Algorithms»:
Recommendations became more transparent on the new page: now you can manually launch their recount after a settings change, see the status and recount time. Currently, there are only personalized recommendations and «Popular in categories» sections on the new page, which is why it will co-exist with the old one. As new algorithms emerge, the old recommendation page will be entirely replaced.
Currently, you can only add one personalization and popularity algorithm to a category, we will lift this limitation by the end of June.
Added recommendation visuals during their set up
We added the possibility to see which goods are recommended to a given client within the interface for all new algorithms. This simplifies the testing of algorithms: now you don’t need to make a widget or a test letter, recommendations can be viewed immediately upon set up and for each individual client. This will also be possible for all algorithms that will be added in the future.
We take into account stock and price across regions when showing recommendations
We could have done this before, but that required to create separate recommendation mechanics for each region. Now, however, one mechanic is enough. We are automatically showing appropriate prices based on region as well as goods that are available there.