Hiring multilingual support, month after month

Template — replace with a real engagement. The structure below is what converts; substitute your own client, numbers and timeline before publishing.

A SaaS company serving customers across Western Europe had already built a small English-language support hub in Sofia and wanted to consolidate its German, French and Dutch queues there too. The requirement was a steady intake — TBC agents per month over TBC months — rather than one large batch, because they onboarded in cohorts and the training team could only absorb so many at a time.

Volume hiring in Bulgaria is not a headcount problem. It is a language problem, and the three languages are not equally available. C1 German in Sofia is a competitive but real market. French is thinner. Dutch, at any usable level, is close to a rounding error and is mostly people who studied or lived in the Netherlands and came back. Treating the three as one requisition would have guaranteed missing the Dutch target while over-delivering on German.

The second problem was attrition. The client’s existing hub had the churn that the sector expects, and every departure meant a replacement on top of the growth plan. A pipeline that only meets the growth number is a pipeline that shrinks the team. And support hiring has a specific failure mode: filling seats with people who will leave in three months looks like success for one quarter and like a disaster for the next four.

We split the mandate into three separate searches with three different strategies, three different salary bands and three different timelines, and reported on them separately. Bundling them into one number would have hidden exactly the thing the client needed to see.

How the pipeline was built

• German — steady advertising plus direct approach into the existing outsourcing and shared-service population in Sofia and Plovdiv, where many agents are already doing similar work for less interesting products.
• French — the same, extended to Bulgarians returning from France and Belgium, and to francophone graduates of the local French-language schools.
• Dutch — treated as a niche search from the start, run continuously rather than in bursts, with the client agreeing a premium band and a remote-from-anywhere-in-Bulgaria option. We told them upfront the Dutch number would be met last, if at all, at the volume requested.
• A screening bar that did not move with the pressure. Language level assessed in a live conversation with a native or near-native speaker, never from a CV claim.

On the retention side we changed what we screened for. Instead of filtering for prior support experience, we filtered for people who wanted the shift pattern the client actually ran, and were explicit about pay, progression and evening coverage in the first call. Candidates who dropped out at that point were the ones who would have dropped out at month three.

Over TBC months we placed TBC agents in total: TBC German, TBC French and TBC Dutch. The German and French cohorts tracked the plan. The Dutch target was met at TBC percent of the original number, which is what we had said would happen; the client covered the shortfall by routing part of the Dutch queue through English, a decision they made early because they had the honest forecast early.

Attrition in the placed cohort at TBC months was TBC percent, against a hub baseline of TBC percent. Cost per hire came out at TBC, and the training team never received a cohort larger than it had asked for.

The client also kept the split reporting after the engagement ended. Seeing German, French and Dutch as three separate lines rather than one monthly total is what let them make the routing decision TBC months before it would have become urgent.

Before publishing: replace this paragraph with what actually went wrong on the engagement and how you handled it. A case study in which nothing went wrong reads as marketing and buyers discount it.