Supply-Chain Case Study: 95% of Calls, 11 Years Running

The work runs on knowledge that takes years to build. It walks out with anyone who leaves.

Consider what one of these calls actually involves. A retailer says it sent a purchase order. A supplier says it never arrived, or arrived in a format its system rejected. Both are looking at their own records and both are correct about what they can see. Resolving it means understanding how each of those two companies has configured its side of the relationship, what it has done historically, and which of a dozen small formatting conventions has changed.

An agent who has handled the same retailer’s problems for six years recognizes the pattern on sight and closes the call in one contact. A new starter, however capable, escalates it. Both agents may be equally intelligent and equally well trained. The difference is time on the account, and time on the account isn’t something a provider can manufacture on request.

That shape of work set the requirement before any provider was considered. The company wanted a contact center inside the United States, and specifically one close enough to drive to and walk around. Which removes the single largest lever in outsourcing before the search has opened.

33%

a year, typically

A third of your customer knowledge, gone every year

Standard contact-center turnover runs at roughly a third annually. On work that runs on accumulated knowledge about specific trading relationships, that’s the hidden line item that never shows up in the quote.

Take the cheaper-geography option off the table, and most procurement instincts stop being useful

You can’t buy your way to a better answer on rate, because every provider in a domestic field carries broadly the same cost structure. The spread between the best and worst price you’ll be quoted is small, and it correlates with almost nothing that matters here.

So the whole decision moves onto fit, and fit is precisely what a scored procurement exercise measures worst. You can score a provider on certifications, on headcount, on references, on price. You can’t score whether its people will still be there in year six, which on this work is the only question that matters.

Worse, the cost of getting it wrong never appears in the pricing comparison. Every agent who leaves takes years of accumulated knowledge with them, and their replacement escalates the same problem the leaver used to resolve in a single call. A provider with a lower hourly rate and ordinary contact-center turnover isn’t cheaper on this work. It’s cheaper per hour while being materially worse per customer.

Six people on trial, at a contact center the company could drive to and walk into

Four things decided this search and price wasn’t one of them: delivery inside the United States, close enough to visit, a provider of a size that would treat a thirty-person program as a serious account rather than a rounding error, and a real, staffed building the company could walk around. In the company’s own words at the time, what stood out about the eventual winner was that it had a live, tangible contact center they could go and visit. They already had.

This search was also, deliberately, not blinded. OC runs blinded searches by default because anonymity stops providers pricing against a brand name. Here the binding criterion was relational and physical, so a process that hid the parties would have defeated the requirement. And the program didn’t start as a cutover. It started as a trial: six agent seats, a defined scope, and the expansion decision deliberately deferred until the trial had produced data. That trial graduated twice, to eleven people, then to more than twenty, and eventually to around thirty.

Step 1

Filter for domestic delivery, driveable proximity, and mid-market provider size

Step 2

Walk the floor. This search wasn’t blinded, because the binding criterion was physical.

Step 3

Start with six people on trial, and defer expansion until data holds

Step 4

OC’s VMO benchmarks across three contract cycles and every expansion

Eleven years, three expansions, and better scores than the in-house team

Six people on trial became roughly thirty across three successive contracts. The outsourced team now handles 95% of all fulfillment phone calls, and outperforms the company’s own in-house contact center on the company’s own scorecard. At the point it was measured, the program was answering 93% of calls inside half a minute against a 90% target. And fewer than one person in twenty leaves each year, a rate that’s the whole service mechanism, on work where accumulated knowledge is the product.

Team size, trial to today

The trial

~30

Today

Six-person trial graduated twice, across three contract cycles, to about thirty. The team absorbed the growth rather than tracking it.

Annual turnover, per 30 agents

This team loses a fifth as many as an industry-standard center.

Industry standard
10 leave

This team
2 leave

Two people left in the last twelve months, out of about thirty. That’s the whole retention mechanism this story turns on.

95%

of all calls

Support didn’t scale with the business. It absorbed it.

In most support organizations, headcount grows in step with customers. Here the team went from 6 to 30 across eleven years while taking on 95% of all fulfillment phone calls, and beat the in-house team on the company’s own scorecard.

Four ways to read this outcome

Different leaders read this story against different numbers. All four readings are correct.

If support headcount is growing with your customer count

In most support organizations every new cohort of customers adds agents rather than efficiency, so the cost of serving customers stays a fixed share of revenue instead of falling. That ratio is what investors read as operational maturity. This team went from six people to roughly thirty across eleven years and now handles 95% of all fulfillment phone calls. Cost per hour is what procurement compares. What matters here is that the headcount line stopped tracking the growth line.

If you own the renewal number

On work that runs on accumulated knowledge, the person who picks up either already understands the customer’s situation or doesn’t, and customers notice the difference immediately. Most who leave a provider do so only after repeated poor experiences, and replacing them costs more every year. Fewer than one person in twenty leaves this team annually, which means the answer is usually given by somebody who has seen the problem before. That’s a retention mechanism, not a staffing statistic.

If your quality floor is your job

The outsourced team outperformed the company’s own in-house contact center on the company’s own scorecard, and at the point it was measured was answering 93% of calls inside half a minute against a 90% target. It also took on work that was never in the original agreement. None of that was sold in; it accumulated because the program kept earning it.

If you have to justify the choice you made

There was no cheaper country to move to, because the requirement was domestic from the start. So the decision had to be made on fit, which is what a scored procurement exercise measures worst. What made it defensible was a shortlist where every option had already cleared the bar, a provider the company could physically walk into, and a six-person trial rather than a handover. Twice more this relationship produced searches that ended with no placement, once because the company was better off keeping the incumbent and once because a project didn’t merit the spend.

Things to take from this story

On work that runs on accumulated knowledge, turnover is the hidden line item. A cheaper seat that turns over twice a year was never actually cheaper. It was just cheaper per hour.

Go and walk the floor. A company buying support that depends on retention is buying an employer rather than a vendor, and a building tells you things no reference call will.

Start with six people. A trial small enough to be wrong about is what lets you commit to a provider you’ve known for months rather than years.

95% of the calls, handled by a team that doesn't leave.