ACTIONABLE raises €8.5 million to turn customer data into operational decisions

Founded by Nicolas Rieul and Nans Thomas, Actionable has announced an €8.5 million funding round led by Hi inov, with participation from Axeleo Capital. The startup aims to help large companies anticipate customer churn, complaints and purchases by connecting commercial data with operational events. Its deployments at Carrefour, OUIGO and Engie illustrate that ambition. Scaling the business will depend on its ability to replicate these use cases and measure the benefits of the decisions they inform.

Six minutes and twelve seconds. According to Actionable, that is the waiting-time threshold at Carrefour’s grocery collection points beyond which the Net Promoter Score, a measure of customers’ willingness to recommend a business, drops in the data analysed. The retailer reportedly turned that finding into a benchmark monitored nationally, regionally and store by store.

The precision draws attention; its operational implications matter more. An analytical finding can prompt a review of order preparation, staffing availability or the organisation of collections. Customer insight then becomes part of the decisions that determine service quality and its cost.

That is the role Actionable is seeking to establish. On 9 September 2026, the company announced a $10 million (€8.5 million) funding round led by Hi inov, with participation from existing investor Axeleo Capital. Founded in April 2024, it combines the experience of Nicolas Rieul, a former Criteo executive, and Nans Thomas, who previously held a product leadership role at Innovorder and founded Wino. Rieul brings an understanding of the commercial decisions companies seek to improve; Thomas brings experience in building software for specific business needs.

From averages to the customers who need a call

The problem the founders describe begins with a dashboard. A decline in satisfaction appears, teams look for its cause, and then have to determine whom to contact and what response to offer. Several decisions still separate the overall finding from an intervention for an individual customer.

“The measurement was there; the granularity needed to make decisions was not. That is the layer we have built,” explains Nicolas Rieul.

Actionable says it can assign each customer scores for satisfaction, churn risk, the risk of a serious complaint and the likelihood of a repeat purchase, alongside the factors that explain them. The company reports sixteen enterprise clients, including Carrefour, SNCF, Edenred and Engie, and says it analyses the journeys of 117 million consumers.

Business context comes before prediction

To produce these results, Actionable starts by bringing together the available information. An order, a wait, a return and a complaint may be recorded in different software systems. The challenge is to identify the same customer across them, put events in the right order and understand what each field means.

The platform reconstructs that journey within a “Common Customer Data Model”, a structure tailored to the relevant sector. Order preparation time, a train delay or the time taken to process a case become variables that can be interpreted in context.

“The difficult work is turning hundreds of tables and in-house definitions into a customer model that a machine can use without making mistakes,” explains Nans Thomas. The company says it spent two years building this capability and can reduce work that might take months of engineering to a few days.

That promise has direct implications for the economics of the product. Each new contract must benefit from connectors, definitions and processing routines already developed. The proportion of work that can be reused determines the resources required for deployment. Actionable does not provide details of its observed median deployment time or the amount of human intervention required. Those two measures would help assess its ability to multiply deployments.

Adjusting campaigns and service operations

The use cases presented by Actionable show how this preparation can inform different decisions. At Carrefour, the company says individual predictions help determine which customers receive which CRM campaigns. It reports an incremental return on investment of seven times the initial outlay.

At OUIGO, the model is used to identify passengers likely to be dissatisfied and offer them a goodwill gesture before they potentially complain. Actionable says the additional revenue generated by an initial campaign covered the cost of the platform within a few weeks. At Engie, customer service teams reportedly use signals of complaints likely to worsen to intervene before they escalate.

These situations involve the same allocation of resources: which customers should receive a call, who should be offered compensation, and which incidents should take priority? They also raise a choice between repairing a poor experience and investing in operations to make it less frequent. Repeated delays may justify a process review; their commercial consequences may call for immediate action.

Is the customer likely to leave also the one you can retain?

This question introduces a further difficulty. Imagine two customers with a similar risk of leaving. Compensation might persuade the first to return. The second may have moved away or chosen an offer the company cannot match. An identical risk score does not necessarily justify the same spending.

Actionable says it uses “causal AI”. This approach seeks to clarify cause-and-effect relationships and the consequences of an intervention. It requires a distinction between factors associated with a behaviour and those the company can act on with an expected outcome.

The distinction is methodological. Established causal inference tools such as DoWhy rely on explicit assumptions about relationships between variables and an examination of their robustness. An observed correlation between waiting and dissatisfaction is not enough to quantify what reducing the wait would change once other relevant factors have been accounted for.

Actionable does not detail the methods it uses or how it validates the effects attributed to actions. In marketing campaigns, comparable control groups help estimate what would have happened without an intervention. For operational changes, the evaluation must account for other factors that could influence the results.

This rigour determines the value of deployment at scale. A good prediction can improve understanding of the customer base. A well-chosen intervention must deliver an additional benefit that exceeds its cost.

Establishing a position against incumbent suppliers

The combination of experience data, operational data and commercial action already features in competing products. Qualtrics, for example, describes capabilities that combine this information to predict individual behaviour, identify customers at risk and organise follow-up.

Actionable will therefore have to establish its advantage in practical use: deployment time, the effort required from teams, prediction quality, explanations of results and integration with existing tools. Large companies with data teams can also compare buying a platform with developing their own models.

Its place in the customer’s budget matters just as much. A solution used across several business functions and embedded in their regular decisions can become harder to replace. Establishing that position means demonstrating its value to marketing, customer service and operations, with results each can understand.

The 117 million consumers analysed do not, however, constitute a pooled proprietary database. Actionable says its client environments are isolated and that data is not shared between them. Its cumulative advantage could therefore develop through the sector-specific models, integrations and deployment methods it can reuse.

Agents broaden the ambition; international expansion tests the economics

The company now intends to extend the use of this data model to AI agents. It already offers Actionable Intelligence, an analyst agent capable of producing analyses from structured information.

This extension could increase how frequently the product is used and the number of business functions involved. It also adds requirements: tracing calculations, checking recommendations and determining who approves decisions.

The funding will support hiring in product, engineering and sales, as well as international expansion, particularly in the US, through a network of reseller partners. It follows an initial €2 million round announced in September 2024, which combined equity and debt.