Hosted help desk software is a ticketing system the vendor runs on its own servers: you pay a monthly fee per agent, your team logs in through a browser and nobody on your side patches anything. An installed system is the older model. You buy licences, put the software on a server you own, and your IT team keeps it alive. For a desk of 20 agents looking three years out, hosted is usually cheaper. For a desk of 60 planning to keep the tool for five years, it usually isn’t.
That’s the short answer. The longer one depends on four checks that matter more than the feature list: the real three-year cost, whether you can get your data out, whether the tool talks to your CRM, and what happens on the afternoon the vendor goes down. If you’re still settling on what your help desk is actually for, sort that first, because none of these checks mean much without it.
What changed in 2009?
Two things, and only one is technical.
Browser-based ticketing has grown up. A few years ago a hosted help desk meant a slow web form and a vendor you’d never heard of. Now the hosted products handle email-to-ticket, SLA timers, reporting and a customer portal, which covers most of what a phone and email desk needs day to day.
The second change is money. Most budgets I know of have capital spending frozen this year. Buying a server and a block of licences is a capital request, and it’s the first thing finance strikes off. A subscription is an operating cost that shows up as a monthly line, and that is a far easier signature to get when everyone is protecting the customers they already have and spending as little as possible to do it.
I see this from both sides. The eCommerce software company where I run the service desk sells its own platform as a hosted service, so I know how attractive the monthly-fee pitch sounds and where it glosses over things.
How do the costs compare over three years?
Build the comparison yourself. Vendor quotes are designed to make their model look best, and the costs that tip the decision are usually the ones nobody quotes: your own admin time.
The numbers below are illustrative. Swap in your own quotes.
- Hosted: $50 per agent per month.
- Installed: $900 per agent licence, 20% annual maintenance, $6,000 for the server and backup, and about $5,500 a year of an administrator’s time for upgrades, patches and restores.
| Agents | Years | Hosted | Installed (licences only) | Installed (with admin time) |
|---|---|---|---|---|
| 20 | 3 | $36,000 | $34,800 | $51,300 |
| 20 | 5 | $60,000 | $42,000 | $69,500 |
| 60 | 3 | $108,000 | $92,400 | $108,900 |
| 60 | 5 | $180,000 | $114,000 | $141,500 |
Look at the 20-agent row. Licence against subscription, installed looks slightly cheaper. Add the admin time and it’s $15,300 more expensive. That one column is why so many small desks are moving to hosted this year.
Now look at 60 agents over five years. The subscription keeps charging for every seat, every month, forever. The installed system’s big costs are behind you by year two, and it comes out $38,500 cheaper.
To build it in Excel, put agents in B1, hosted price per agent per month in B2, years in B3, licence per agent in B4, maintenance rate in B5 (0.20), server cost in B6 and annual admin cost in B7.
- Hosted total:
=B1*B2*12*B3 - Installed total:
=B1*B4+B1*B4*B5*B3+B6+B7*B3 - Break-even in years:
=(B1*B4+B6)/(B1*B2*12-B1*B4*B5-B7)
With 20 agents the break-even comes out at 8.3 years, which is longer than most help desk tools survive. With 60 agents it’s 3.0 years. If the break-even formula returns a negative number, the installed system never catches up at that headcount.
Can you get your data out?
This is the question to ask before price. With an installed system the database sits on your server, so the data is yours even if the vendor’s format is awkward. With hosted software your ticket history lives on someone else’s machine, and leaving depends on what they let you take.
Test it during the trial, not after you sign. Ask for a full export of ten closed tickets and check that you get:
- every comment and reply, in order, with who wrote it
- the original created, first response and closed timestamps
- your categories and custom fields
- attachments, or at least a list of them
If the export flattens the timestamps or drops your categories, your reporting history won’t survive the move. Most of what you’d want to report on depends on a ticket category scheme that holds up over time, and you’ll lose that the day you switch if the export can’t carry it.
Will it talk to your CRM?
If sales or account management keeps customer records in a CRM, support tickets need to point at the same customer. Otherwise the account manager walks into a renewal meeting not knowing the customer logged eleven tickets last month.
Ask three things. Which system owns the customer record? Does the help desk match on an account ID, or on a company name someone typed? And is the link a supported integration or a script your IT person writes and then has to look after?
Installed systems often connect by reading each other’s databases directly. It’s fast, and it breaks at every upgrade. Hosted tools connect through the vendor’s published interface, which is slower to set up but survives upgrades because the vendor has to keep it working for every customer.
What happens when the hosted vendor goes down?
It will go down. On 6 January this year Salesforce’s service was largely unavailable for about 38 minutes on a Tuesday lunchtime, and Computerworld reported that its status site was down too, so customers couldn’t even find out what was happening.
That is the risk you take on with hosted software. Your installed server can fail too, of course, but then your own IT team is working the problem and you can walk over and ask how long. With hosted, you wait. So plan the wait before it happens: a paper or spreadsheet fallback for logging calls, a named person who checks the vendor’s status and tells the floor, and a rule for re-entering the tickets afterwards so nothing logged during the outage disappears.
What should the hosted contract say?
Uptime percentages sound alike and aren’t. Over a 30-day month:
| Uptime promised | Downtime allowed per month |
|---|---|
| 99% | 7.2 hours |
| 99.5% | 3.6 hours |
| 99.9% | 43 minutes |
A 38-minute outage fits inside a 99.9% month. At 99%, the vendor could be down for a full working day every month and still meet the contract. Get the figure in writing, find out whether planned maintenance counts against it, and make sure maintenance windows fall outside your peak hours, not theirs.
Then look at what you get when they miss. Service credits are normally a small slice of one month’s fee, which doesn’t come close to what an outage costs your customers. I covered the same gap with the BlackBerry outage in December: the credit is the vendor’s penalty, and your customers bear most of the real cost.
Who should still install?
Install if you run a large desk and plan to keep the tool for years, if your customers’ data can’t leave your building for contractual reasons, or if the help desk has to sit deep inside other systems you already host yourself. In those cases the server is worth owning.
Everyone else should give hosted a serious trial. And before you sign, ask the vendor for one thing almost nobody asks for: the name of a customer who left them and took their data. How easily that customer got out tells you more about the vendor than anything in the sales demo.
Hutch Morzaria is a Director-level CX and Support Leadership professional with 19 years of experience building global support organizations across SaaS, Fintech, and enterprise technology. He has hired dozens of support and CX leaders across his career and holds ITIL Expert certification across V3 and V4.
