Corinna Covini, Copenhagen Fintech: What has Predicti been working on?

Luke Tricker, Head of Marketing : We have built a housing prediction model that looks at who is most likely to buy a new home within the next 12 months in Denmark. If you look at the local housing market, there are approximately 1.4 million owned houses orapartments. With our prediction model, if you were to contact the top 20percent, around 280,000 out of that 1.4 million, we would be able to identifynearly 60 percent of the total number of buyers for the next 12 months.

Corinna: Who is thisproduct relevant for?

Luke: It isrelevant for anything connected to the homebuying process: estate agents, bankslooking at mortgage products, home insurance providers, moving insuranceproviders, and customer relationship management teams.

We want toenable insurers and banks to be proactive in customer engagement rather thanreactive, and to reach out to customers sooner in the process.

The model ranksaddresses by likelihood, helping financial institutions understand whichhouseholds to contact earlier in the homebuying journey.

*do you have data about the effect thata proactive customer engagement generates? Which impact it has on the customers(banks, insurers ecc.)*[LT1] 

Corinna: What is thesecret sauce of this prediction model?

Amalie: I would say the waythe data is set up. The model has been built from Danish registry and publiclyavailable data sources, including historical data on people who have actuallymoved from 2020 onwards, such as CPR and BBR-related data[LT2] .We now have software that combines these different data sources into logicaltimelines.

We use theaddress as a proxy. We are not saying that a specific named person will buy ahome. We are ranking addresses and households by likelihood, and can saywhether one address is more likely to buy than another address ranked below it.

Corinna: What has beenone of the hardest parts of building this?

Amalie: Makingsure the output was correct. The initial building process was relatively fast,and the first results we got were surprisingly good. But we still spent a lotof time testing and making sure that we had no feature leakage or data leakagethat would alter the predictions.

The way wetested it was by training the model on historic data up to 2024. That gave usthe full year of 2025 to compare the model against publicly available data andask: did the predicted households actually buy something in the next 12 months?

Corinna: Where willPredicti go from here?

Luke: This housingprediction model is the first outward-facing model we are going to market with as astandalone product[LT3] .The second step would be to develop more models. We have many in the pipelinearound life stages, such as being able to predict which life stage someone isabout to enter, or which kind of customer segment they are moving into.

If an insurer or a bank cameon board with us tomorrow, we could give them access to the platform,[LT4] and their data science teams could build their own models based on theircustomer data. You could build advanced churn scores, [LT5] churnmodels, product recommendations and much more. There are a lot of datascientists out there who would have many ideas.

Amalie: We alsomeet smaller or medium-sized customers that do not have a large data scienceteam, or may not have one at all. In those cases, we can build models for themin-house.

Corinna: Which sectorproblem can Predicti help solve with more customized predictive models?

Luke: Insurersand banks often have legacy systems that are fragmented and hard to extractdata from.

For theseteams, it would take a long time to build something like what Amalie has builtfor Predicti, because they would first have to collect the datasets, connectthe historical information and manage all the different moving parts.

Predictialready has a validated setup. We could take their data, put it into oursystem, and it would be almost ready to plug and play.

Corinna: How has beingNordic-based impacted this development journey?

Luke: BeingNordic-based has worked to our advantage. Across the region, digital adoptionis high compared with much of Europe, particularly in insurance and banking.The Nordic region is one of the most digitally mature regions in Europe.

If you thinkabout life in Denmark, everything is connected to your CPR number. Everythingis online, you can log into services easily, and there is a lot of digitalself-service in everyday life. People expect that same level of personalizationand ease from their insurer and their bank.

What we do isenable the personalization people already expect in other areas of life, butwithin their insurance and banking interactions.

Platforms likeCopenhagen Fintech's Beyond Nordic Scaleup Program also support knowledgesharing among scaling peers. Getting to meet other people in the industry, seewhat they are doing, and connect at events such as Nordic Fintech Week isvaluable. The collaborative aspect of the ecosystem is a strong asset.

Corinna: Where do yousee the sector going?

Luke: I believehyper-personalization products like the ones Predicti is building are thedirection of travel.

If you think about socialmedia, IKEA or other digital services, they treat you much more as anindividual than many financial services do today. I think more localizedmessaging and more personalized offerings are where the sector is heading.

[LT6] 

 [LT1]With a recentcustomer, their outreach became at least 9 times more efficient in identifyingcustomers with actual intent to buy a house, when using the model, compared totraditional outreach.

 [LT2]Remove thespecific reference to CPR and BBR

 [LT3]Please reword toread "This housing prediction model is the first of our new ML models thathave been built and trained on top of our collected data platform. We have manymore in the pipeline, such as being able..."

 [LT4]"...wewould give them access to the platform where we collect all the data sourcestogether, and then their data science..."

 [LT5]Delete"churn scores"

 [LT6]Predicti is anAI powered intelligence solution enabling enhanced engagement throughout thecustomer journey. We integrate multiple data sources into one powerfulplatform, to produce highly personalized, real-time customer insights,recommendations, and predictions through advanced Machine Learning models.

‍