Tinyclues: Helping digital marketers deliver better campaigns
About Tinyclues
Tinyclues is a CRM marketing solution that enables marketing teams to build and scale intelligent marketing plans powered by predictive insights from first-party customer data. More than 1,000 users use the Tinyclues platform, including Samsung, FNAC DARTY and Air France, and they have all experienced significant uplifts in campaign revenue and increased customer engagement.
Tell us your challenge. We're here to help.
Contact usTinyclues reduced its operating costs and improved its predictive models by leveraging BigQuery and Vertex AI on Google Cloud, enabling it to deliver faster, smarter insights to its clients.
Google Cloud results
- Reduces infrastructure costs by at least 50%, supporting business expansion into new sectors
- Removes limitations on data scaling, enabling new and improved predictive models and delivering better data insights
- Enables clients to easily connect their BigQuery data warehouses to Tinyclues, reducing implementation times to two to three days
Scoring times down from 20 minutes to two
Many digital marketers are making the switch from data analytics that uses third-party data such as tracking cookies, to first-party data provided directly by consumers. What hasn't changed is their key goal: delivering customer-centric communication. That means finding new and improved ways of delivering targeted campaigns that connect with the right audiences.
French company Tinyclues's mission is to help retailers, travel companies, and the luxury sector by applying advanced deep learning to customer relationship management (CRM) marketing. With Tinyclues, businesses can connect their first-party data and use it to map out demand, predict intent to buy, and manage campaigns across multiple channels through the Tinyclues platform. These businesses’ customers then receive communications that are relevant to their interests, even for things they may not have even searched for yet.
"Tinyclues has run hundreds of thousands of campaigns for our major clients, and mathematics and deep learning is in our DNA," says Michael Aidane, CTO/CPO, Tinyclues. "We offer an industrialized solution that's fast to implement and delivers demonstrable improvements in CRM revenue very quickly, up to 150%."
Tinyclues had previously built its own stack and deployed it on another public cloud. But it wanted to reduce its operating costs and overcome the limitations of its database, so it migrated to Google Cloud.
"We decided to switch to Google Cloud and embrace a managed services philosophy to enable us to focus on our core expertise," says Aidane. "Being the best at what we do is what helps us differentiate ourselves in the market and better serve our customers."
"We decided to switch to Google Cloud and embrace a managed services philosophy to enable us to focus on our core expertise. Being the best at what we do is what helps us differentiate ourselves on the market and better serve our customers."
—Michael Aidane, CTO/CPO, TinycluesChoosing a platform that supports advanced AI
Tinyclues previously had to manage development, staging, and production environments, leading to high infrastructure costs. Its DevOps team was spending a lot of time on infrastructure maintenance (updating Kubernetes versions etc.) and keeping its system updated. In order to streamline these operations, it decided to switch to managed services, and looked for a new cloud provider to fulfill its data and machine learning needs.
"A big factor in our choice of Google Cloud was its data and machine learning services," says Aidane. "Tensorflow and Kubeflow were both incubated by Google, and we use both of them. By switching to Google Cloud, we knew we'd be able to benefit from updates and new releases as soon as they came out."
Tinyclues ran two proof of concept studies in late 2021 to ensure the data-related pain points would be resolved. The company then migrated over its clients one-by-one. It restructured and reconfigured its services as it went, ensuring performance was improved for every deployment, completing the whole process over a period of 15 months.
"Our clients had to continue running their campaigns, so we couldn't take a big bang approach," says Aidane. "We stretched the limits of what's possible with Vertex AI, but we benefited from very dedicated customer engineers who worked closely with us to make the right resources available."
Reducing computation times to deliver insights faster
The biggest data challenge for Tinyclues is dealing with variable data from across sectors and retaining the semantic complexity of that data, while ensuring the solution is adaptable enough to work as a SaaS product.
For a typical client, it might receive four years' worth of data including interaction logs and page views. To handle those volumes, Tinyclues leverages BigQuery, orchestrating the workloads using Cloud Composer. "BigQuery simply works. You don't need to worry about how big the cluster needs to be, and we've been able to run models that previously would have crashed," says Aidane. "From the get-go, the team has been surprised with how magical it is."
To train its models, Tinyclues uses Vertex AI, and deploys them using BigQuery ML. This means scaling up the inference stack is no longer a problem, removing the need to either overprovision or waste precious minutes spinning up an on demand stack. Where previously it took up to 25 minutes for a brand to receive scores for all its users in relation to a particular product, it now takes two or three minutes, even for very large clients. BigQuery ML also enables Tinyclues to provide multi-offer inference, looking at 2,000 topics in less than 20 minutes to work out which ones are the best marketing candidates.
"Without the serverless power of BigQuery ML, delivering multi-offer inference simply wouldn't be cost effective," explains Aidane.
"BigQuery simply works: you don't need to worry about how big the cluster needs to be, and we've been able to run models that previously would have crashed. From the get-go, the team has been surprised with how magic it is."
—Michael Aidane, CTO/CPO, TinycluesOffering CDP capabilities with Google Cloud and Tinyclues
As part of its move to expand in markets abroad, Tinyclues is using Google Cloud Marketplace to reach more clients globally. "We don't have the resources to go to market in every major market, but Tinyclues is now available with Google Cloud Marketplace," explains Aidane. "That helps us connect with people who are searching for customer data platform (CDP) functionality."
By using Tinyclues together with Google Cloud data solutions, companies can cover their CDP needs without needing to purchase a separate customer data platform. For customers who are already using BigQuery, it's also much quicker to set up Tinyclues. Building a new data pipeline could take up to 40 days, whereas working with client data directly within BigQuery reduces the set up time to a few hours.
Delivering new services and reaching new clients
Since switching to Google Cloud, Tinyclues has successfully reduced its operating costs. "Switching to Google Cloud and restructuring our platform has enabled us to reduce costs by 50% or more," says Aidane. "That opens up new markets for us. We're now able to calculate costs very precisely, as well as exploring opening our offering to SMBs."
Speeding up calculation times with Vertex AI & BigQuery ML also enables Tinyclues to offer a much richer user experience, with interactive features that previously would have been too expensive to deliver. It can pre-compute data sets for every potential offer a client might want to promote, across all categories and brands. That means clients can easily drag and drop around topics and products in the platform and work out an optimum campaign schedule in under five minutes. They are able to plan strategically, exploring the best communication opportunities across all potential campaigns.
The next step for Tinyclues is to reap the benefits of the improvements it has made to its platform, which have enabled it to accelerate its roadmap for delivering new features for its clients. "One thing we've been able to develop is predicting purchasing behaviors in relation to products that have only just been created," says Aidane. "If Samsung brings out a new phone, they can work out who they should target with that particular campaign, for example. Working together with Google Marketplace, we're looking forward to reaching a lot of new clients and offering new capabilities."
"Switching to Google Cloud and restructuring our platform has enabled us to reduce costs by 50% or more. That opens up new markets for us, as we're able to calculate costs very precisely and reach SMBs via Google Marketplace."
—Michael Aidane, CTO/CPO, TinycluesTell us your challenge. We're here to help.
Contact us
About Tinyclues
Tinyclues is a CRM marketing solution that enables marketing teams to build and scale intelligent marketing plans powered by predictive insights from first-party customer data. More than 1,000 users use the Tinyclues platform, including Samsung, FNAC DARTY and Air France, and they have all experienced significant uplifts in campaign revenue and increased customer engagement.