Follow the sun and follow the language solve different problems. One buys you coverage at 3am, the other buys you comprehension at any hour. Teams routinely spend on the first while suffering from the second, because a timezone gap is visible on a rota and a language gap is not.
TL;DR
- Work out which gap you have before costing a model. They look identical in a ticket backlog and need opposite solutions.
- Timezone coverage fixes waiting. Language coverage fixes understanding. Neither substitutes for the other.
- A region with poor overlap but good English needs a rota change. A region in your timezone working another language needs content.
- Most distributed companies need one staggered shift and two translated document sets, not a round-the-clock desk.
- Per-agent pricing punishes whichever model you pick, because both add people rather than volume.
- Measure time to first useful answer, not time to first response. The gap between them is where the real problem hides.
Two gaps that look the same in the data
Both failures produce the same surface symptom: tickets from one region take longer to resolve than tickets from head office. If you stop there you will reach for a rota, because a rota is the obvious answer to slowness.
The distinguishing test is where the time goes. A timezone gap shows up as a long wait before the first reply and a fast resolution afterwards. A language gap shows up as a fast first reply followed by several clarification exchanges before anybody understands the problem. Same total duration, completely different cause.
Pull twenty tickets from the slow region and look at the timestamps rather than the totals. The shape tells you which model to buy, and it takes twenty minutes.
What follow the sun actually costs
Genuine round-the-clock coverage means three shifts, which means a minimum viable team considerably larger than most companies expect, because each shift needs enough people to absorb leave and sickness. Below roughly fifteen support staff it is not a realistic structure.
The cheaper version is partial overlap, and it is underrated. One person starting three hours earlier on a rota can convert a region from zero overlap to a usable window, which changes the experience far more than the headcount cost suggests. Most teams should exhaust this before considering anything larger.
There is also a hidden cost nobody prices: handover. Two people in different timezones touching one queue will occasionally answer the same ticket twice, differently, which is worse for the requester than a slow single answer.
What language coverage actually costs
Mostly content rather than people, which is why it is cheaper than it looks. Your highest-volume ticket topics are access, provisioning, setup and connectivity, and those answers are identical for everybody. Translating twelve articles absorbs most of the demand in a new language.
What costs money is the residual: the tickets that need a conversation. For those you need somebody who reads the language, and the honest options are a hire, a reciprocal arrangement with a local team, or an external provider. The mistake is sizing that hire before doing the content work, because the content work usually removes two thirds of the need for it.
A worked example
A 320-person company had two slow regions and assumed one problem. Region one, 40 people eleven hours ahead, averaged 19 hours to resolution. Region two, 55 people in the same timezone as the desk, averaged 17 hours.
The timestamps separated them immediately. Region one waited 16 hours for a first reply and then resolved in under an hour. Region two got a reply in 20 minutes and then took four exchanges to establish what was wrong. One was a rota problem and one was a language problem, and a single round-the-clock desk would have fixed only the first at several times the cost.
The actual fix was one staggered early shift and a translated document set with a document-trained assistant in front of it. Region one came down to six hours, region two to five, and the total cost was one shift allowance plus a fortnight of content work.
Choosing between them
What to do:
- Pull twenty tickets from each slow region and record time to first reply separately from time to resolution.
- Count clarification exchanges per ticket. More than two is a comprehension signal rather than a coverage one.
- Check whether the slow region actually has poor overlap, using the requester’s local hours rather than yours.
- If it is a language gap, list your twelve highest-volume topics and translate those before anything else.
- If it is a coverage gap, try one staggered shift before costing a full rota.
- Publish a response promise expressed in the requester’s working day, whichever gap you have.
Why pricing basis decides the budget
Both models add people rather than ticket volume, which is precisely the case where per-agent licensing works against you. Every figure below was read from the vendor’s own pricing page on 5 and 7 October 2026.
| Zendesk | $55 per agent per month plus $50 for the Copilot add-on | Two licence lines per new shift member |
| Freshservice | $19, $49 or $99 per agent per month, plus $29 for Freddy AI | Two licence lines per new shift member |
| Crisp | Free, $45, $95 or $295 per month per workspace | No change when the team grows |
| Matram | $29, $69 or $199 per month flat, seats unlimited | No change when the team grows |
| SiteGPT | $468 or $948 per year, billed yearly | No change within plan limits |
Disclosure: Matram is owned by the same people who publish PeopleOpsHQ. It has no free tier, it answers rather than actioning tickets, and it is not an ITSM platform, so a staffed rota still needs a service desk behind it. The full set sits in our multilingual IT support comparison.
Final thoughts
The reason this decision goes wrong is that a rota is a thing you can put on a slide and content is not. Coverage feels like a proper answer to slow support, and for a region eleven hours away it is. For a region sitting in your own timezone working in another language it is an expensive irrelevance.
Twenty tickets and two columns of timestamps will tell you which you have. Do that before costing anything, try one staggered shift before a full rota, and translate the twelve articles that cover most of your volume before sizing any hire.
Frequently asked questions
How do I tell a timezone problem from a language problem?
Separate time to first reply from time to resolution on about twenty tickets from the slow region. A timezone gap produces a long wait before anybody responds and then a quick fix, while a language gap produces a fast first response followed by several exchanges before the problem is understood. Counting clarification messages per ticket is the single clearest signal, since more than two usually means comprehension rather than coverage.
How small can a follow-the-sun team be?
Genuine round-the-clock coverage needs three shifts with enough people on each to absorb leave and sickness, which in practice means a support function of roughly fifteen or more before it is a realistic structure. Below that you are building something that collapses the first time two people are away in the same week. Partial overlap achieved with one staggered shift is almost always the better starting point and delivers most of the perceived improvement.
What is the cheapest improvement for a region with no overlap?
Publishing a response promise expressed in the requester’s working day, with a named route for anything genuinely time-critical. Most of the frustration in a no-overlap region comes from not knowing when to expect a reply rather than from the length of the wait itself, so an honest next-working-day commitment that is reliably kept beats an optimistic target that is missed. After that, one staggered early or late shift is the highest-return staffing change.
Does double-answering really happen with split shifts?
Often enough to plan for, and it is more damaging than a slow reply because the requester receives two answers that do not quite agree. The fix is a convention rather than a tool: assignment has to happen as a deliberate first action before anybody replies, not as a side effect of replying. A three-line handover note at the end of each shift covering what is open, what is time-critical and what was escalated removes most of the rest.
Should we translate the knowledge base or hire someone who speaks the language?
Translate first and size the hire afterwards, because the highest-volume ticket topics are access, provisioning, setup and connectivity, and those answers are the same for everybody regardless of language. Translating your twelve busiest articles typically absorbs the majority of demand in a new language for a fraction of the cost of a role. Measure again a quarter later and the residual need is usually small enough to meet with a part-time arrangement or a named local contact.
Why does the pricing model matter for either approach?
Because both follow the sun and language coverage work by adding people rather than by handling more tickets, and most support platforms bill per agent. That means the structural change you are making to improve service arrives as a licence line on top of each salary, and on products where the AI capability is a second per-agent charge the per-person figure roughly doubles. Flat and per-workspace pricing break that link, which is worth as much attention as any feature comparison.
