Customer experience
Make suitable products, content or services easier to find. Reduce choice overload.
Recommendation Systems
For each customer.
A relevant offer.
We develop recommendation systems for products, content and services, using interests and interaction history to make offers more relevant.
Signals → Selection → Validation
A relevant customer experienceMeasure impact after launch
From interest to actionHelp people choose.
Help people choose.
Measure the outcome.
Recommendations help users choose. Business impact is assessed through an agreed experiment.
Average transaction
Suggest a useful addition to a purchase. Test its impact on basket composition.
Sales
Improve offer relevance and assess results using agreed metrics.
Personal selection logicMany options.
Many options.
Relevant ones closer.
Recommendations can select products, content or services. Switch scenarios to see which signals inform selection.
Personal selectionProcessing diagram
Products that fit
1
Accessory
Complements the selected product
2
New option
Similar to viewed items
3
Selection
Reflects interests
CatalogueViewsPurchases
Selection considers data and business rules: availability, restrictions and display context.
Personal selectionProcessing diagram
Content of interest
1
Video
Based on your interests
2
Article
Related to your reading
3
Playlist
Based on listening history
TopicsViewing historyFeedback
Selection considers data and business rules: availability, restrictions and display context.
Personal selectionProcessing diagram
The service you need
1
Offer
Based on order context
2
Specialist
Based on defined criteria
3
Service
Based on interest history
ContextInterest profileAvailability
Selection considers data and business rules: availability, restrictions and display context.
Where it worksOne idea.
One idea.
Different touchpoints.
We design logic for your audience's behaviour and your product.
Stores and marketplaces
Product, similar item and basket addition recommendations.
Food and delivery
Dish and offer selection using interests and order context.
Streaming
Personalised music, video and playlist selections.
HR and expert search
Match vacancies and specialists using agreed suitability criteria.
News and advertising
Relevant topics, articles and promotional offers.
Social platforms
Content and connection suggestions using interests and interaction history.
What informs selectionInterest data.
Interest data.
Your business rules.
Determine which signals are available and useful. Each feature's contribution is tested against the task.
Events
Views, purchases, ratings and other agreed interactions. Assess historical completeness.
Catalogue
Product, topic or service characteristics. Agree consistent descriptions and updates.
Context and rules
Availability, restrictions, display goals and product conditions are incorporated into selection.
Impact validationA hypothesis in the model.
A hypothesis in the model.
An answer in the experiment.
Compare approaches across agreed user groups. Define A/B test conditions before launch.
The current selection method.
Standard selection, existing logic or another agreed baseline provides the comparison.
The new recommendation.
Test new logic under comparable conditions. Assess useful effects and possible side effects.
Target action and conversion
Basket and average transaction
Repeat interactions
Metrics, duration, sample size and deployment decision criteria are agreed for the project. Performance gains are assessed through testing.
How we workA clear process.
A clear process.
A verifiable result.
From problem definition to integration and business impact testing.
-
01
Task
Define the goal, recommendation placement and success metric.
-
02
Project assessment
Consider data, quality and speed requirements. Calculate scope and cost.
-
03
Data analysis
Study the catalogue and interaction history. Prepare the model foundation.
-
04
Development and testing
Develop, integrate and launch A/B tests.
Questions answeredDiscuss the essentials.
Discuss the essentials.
Before launch.
What about a new user?+
For initial interactions, a baseline selection can use popular options, context or answers to short questions. The approach is selected for the product and tested separately.
Do we need extensive purchase history?+
History volume and type are assessed at the start. Some workflows can use catalogue characteristics and context. Personalisation feasibility and quality are tested on available data.
Where can recommendations appear?+
In catalogues, product pages, baskets, applications, feeds or other interfaces. Integration and response times are agreed for your system.
How do we know recommendations are useful?+
First select metrics, then assess model quality and A/B test impact. Compare results with the existing approach or a baseline.
Start with your challengeCloser to their interests.
Closer to their interests.
Closer to your customer.
Tell us about your product, available data and where recommendations should help users.
- Catalogue, content or services
- Interaction history and restrictions
- Recommendation goal and placement
Solution scope, timelines and terms are agreed for your project.
Loading the contact form…