A repeatable data-science playbook across a private equity portfolio

QuantSpark built a repeatable data-science playbook for a leading European software-focused private equity investor and ran it across more than seven portfolio companies over an 18-month-plus engagement. At one company, processing more than 25 million call-centre logs to optimise call handling drove more than £2m of incremental EBITDA. At a second, a churn early-warning model is estimated to have saved a further £1m-plus of EBITDA through proactive retention. At a third, an interactive dashboard reshaped growth strategy and investor conversations. Every project in the portfolio delivered, or was on track to deliver, EBITDA growth in the multiple millions of pounds.

The underlying problem was common to the whole portfolio rather than specific to any one business. The investor's portfolio companies were data rich but data siloed: product usage, service-call, billing and marketing information sat in disconnected legacy systems, with no route to an end-to-end view of the customer. That mattered because private equity works to a clock. Inside a typical three-to-five-year hold, management teams need to unlock value quickly: reducing churn, sharpening renewals and building the evidence base that will convince the next buyer the growth story is real. For mid-sized businesses, building that kind of data capability from scratch had previously been technically unrealistic within the time and budget a hold period allows.

QuantSpark was engaged at firm level, not company by company, working directly with the sponsor's value-creation team. Blended teams of consultants, data scientists and developers worked in daily iterations and weekly sprints, applying the same two-step method at each portfolio company. The first step was infrastructure: build a cloud data platform that cleaned and connected the disparate sources into a single dataset, with automated KPI dashboards sitting on top so management could see the business clearly, often for the first time. The second step was intelligence: layer machine learning onto that clean foundation through two or three targeted projects per company, drawn from a menu that included churn early-warning models, lead-prioritisation scoring and renewal call-timing optimisation. Rather than leave each build as a one-off, QuantSpark productised the reusable pieces, including a churn-analytics dashboard and a revenue-metrics tool, and handed them to the sponsor's own analysts to run and extend after the engagement moved on.

That two-step method scaled into a five-stage workflow repeated across the portfolio: the sponsor's value-creation team commissioned the engagement at fund level; QuantSpark built the unified data platform and KPI dashboards at each company; targeted machine learning use cases were layered on top; the resulting tools were packaged and handed to the sponsor's analysts; and the whole playbook was then replicated at the next portfolio company, refined by what had been learned each time. It was this repeatability, more than any single model, that let a blended team cover seven-plus companies in eighteen months.

The systems involved were categorical rather than exotic: a cloud data platform to unify the sources, automated KPI and BI dashboards for management reporting, and a small set of machine learning models targeted at specific commercial levers, churn, lead conversion and renewal timing, rather than one do-everything model. The value these systems generated is best read as three separate cases rather than a single blended average. The £2m-plus figure at the call-centre optimisation project is presented as delivered, tied directly to processing more than 25 million call logs. The £1m-plus figure at the churn-model project is presented as an estimate, reflecting that retention benefits accrue over a longer and less certain horizon than an operational efficiency gain. The third project, the BI dashboard, delivered strategic rather than monetised value, shaping how management and the investor discussed growth. Each portfolio company will have translated its own operational uplift, whether in call-handling time, retention rate or renewal conversion, into an EBITDA figure using its own cost and margin structure; that is the standard, generic method for attributing this kind of data-science intervention to the bottom line, though the source does not set out the detailed calculation, and it explains why some figures in the portfolio are stated as delivered and others as estimated or still in progress.