A switching-cost moat exists when the customer's total cost of moving to a rival (money, time, project risk, career risk) exceeds the benefit the rival offers. Then the seller can raise prices a little every year, keep the customer, and earn excess returns on a stock of locked-in relationships. Notice what this implies: the moat belongs to the installed base, and the product matters less than you'd expect. It can even be mildly inferior, and the customer rationally stays. Switching costs come in three flavors, in rising order of subtlety:
- Financial: explicit exit costs, meaning termination fees, unamortized implementation spend, retraining budgets, data-migration costs. These are the easiest to see and the easiest to buy out. A rival can literally write a cheque.
- Procedural: this is the big one. The product is embedded in the customer's workflows (processes, integrations, data schemas, muscle memory, certifications), so replacing it means re-engineering the organization and risking operational failure. Core-banking software is the archetype. A bank's core system posts every account, every day, so replacing it is open-heart surgery on a running patient. It's a multi-year, nine-figure program with career-ending failure modes, and several famous bank migrations have gone publicly, expensively wrong. As a result, core-banking vendors (Temenos, Infosys Finacle, TCS BaNCS, Oracle FSS globally, plus Fiserv, FIS, Jack Henry among US community banks) enjoy retention rates around 95 to 99% and decade-plus relationships. They can take small annual price escalations forever. The customer isn't delighted. They're committed, and for an analyst, committed is better.
- Relational: trust and habit built up in one human relationship, such as the family CA, the private banker, the specialist distributor. It's real but it's the weakest form, because it walks out the door with the people. Grade it down unless it's institutionalized, as with audit relationships that persist across partner rotation.
Advertisement
What do switching costs look like at TCS?
Worked example (India): TCS, embedded workflows at civilizational scale. TCS's claim goes like this. Once TCS runs your policy-administration system, your core modernization program, your payroll and settlements plumbing, replacing it is costly, risky, and slow, so relationships compound. To check the mechanism, look at what TCS has historically reported. It says ~95%+ of annual revenue comes from existing clients (repeat business). Client relationships span decades, and in the client pyramid the $100m+ relationships keep growing. Wallet-share expansion inside locked-in accounts is the actual growth engine.
Again the figures are illustrative, modeled on ~FY25 scale, so pull the actual annual report. Revenue is ~₹2,55,000 crore (~$30bn). EBIT margin has held at ~24 to 25% for over a decade, while peers with the same labor pool run 15 to 21%, and that premium is the embedded-workflow rent. A services firm's invested capital is little more than working capital, so ROCE runs north of 50% on total book equity, and effectively ~100% excluding the cash pile. Worked Example 2 computes it step by step.
Ten years of that, through COVID and demand cycles, is a confirmed procedural-switching-cost moat.
There's an open question on direction, which you should flag now and settle when you study moat dynamics. Does generative-AI-driven productivity deflate the pricing unit, the person-hour, faster than TCS can re-price outcomes? A moat in the installed base survives only as long as the base's workflows themselves survive. Be evenhanded here. This is contested, and the 10-year trail so far still says "moat."
What do switching costs look like at Jack Henry?
Worked example (US): Jack Henry & Associates. This is the pure-play version of the same physics: core processing and payments for ~thousands of US community banks and credit unions. Contracts run 5 to 8 years, and retention is famously ~99%. A community bank's core conversion is so feared that de-conversion fees are a disclosed revenue line. On illustrative ~FY24 figures (pull the 10-K), revenue is ~$2.2bn with operating margins of ~22%+, and ROIC runs around the high teens to ~20% against a ~8 to 9% USD hurdle, with remarkable stability year to year. None of that is spectacular, and all of it is durable. A switching-cost moat often looks like this: modest growth, metronomic margins, decades of spread.
Morningstar has long rated it wide-moat on these grounds. Treat that as a cross-check, and never as your evidence.
How do you test for switching costs?
Look for switching costs in the order that settles the question fastest. Gross retention and net revenue retention come first. In SaaS, NRR ≥110% means locked-in customers are expanding. Then contract length, deferred revenue and RPO, and any disclosed termination or de-conversion fee, because a fee is a switching cost with a price on it. Revenue from existing clients confirms what those say. TCS and Infosys disclose repeat business directly, and average client tenure with the client-size pyramid ($1m/$10m/$100m+ buckets in IT-services disclosures) adds detail. Price escalators in the MD&A are the weakest evidence here. Last comes the negative check. If sales and marketing spend keeps rising just to hold revenue flat, customers aren't locked in, whatever the deck says.