How to choose an answering service for your small business

Choose an answering service the way you would choose a supplier you cannot easily leave: from evidence you gathered yourself, in a fixed order. One week of your own call log, a shortlist of three, the same six test calls placed to each, and one scoresheet. Rate cards come last, because a plan sized for a call pattern you never measured is how a small business ends up paying for four hundred minutes it does not use. BotsDontSleep answers 24/7 at $0.60 CAD per answered minute with $0 setup and no contract, which is also what makes the test cheap — a week of test calls costs less than lunch.

The criteria themselves — coverage, capability, price, contract, escalation, transcripts — are set out in how to choose the right answering service. This post is the procedure: how to actually run the comparison when you are the owner, you have no front desk, and you can spare about two hours across two weeks.

Why does choosing an answering service go wrong for small businesses?

Because the decision gets made from rate cards and review sites rather than from your own phone line.

The sequence most owners follow is: search, open six tabs, compare monthly prices, pick the cheapest plan that looks big enough, sign. Every step is reasonable and the order is backwards. You cannot tell which plan is big enough without knowing what your line actually produces, and you cannot tell whether a service can do the job without hearing it attempt the job.

A business with no receptionist is also the hardest case to size from a brochure. Your volume is low enough that plan minimums dominate the bill, uneven enough that a monthly average hides the Monday spike, and mixed enough — enquiries, wrong numbers, the supplier who calls every week — that a per-call price and a per-minute price produce very different invoices from identical traffic.

Two weeks of deliberate work fixes all three. Week one you measure. Week two you test.

What should you measure before you talk to anyone?

Five columns, one week, on paper or in a spreadsheet. Every call that rings your business line, whether or not you answered it.

• Time and day. You are looking for the shape of the week, not the total.
• Answered or missed. Count the missed ones honestly. Owners consistently underestimate this, because the calls you never heard are the calls you never heard.
• Roughly how long it ran. Thirty seconds, two minutes, ten minutes.
• What it was. Booking, quote, status check, supplier, wrong number, robocall.
• What it was worth. Your average job value if it was a real enquiry, blank if it was not.

At the end of the week you have four numbers that decide everything after: calls per week, the share you missed, your average call length, and the share of traffic that was never a customer. A shop taking 40 calls a week where 14 are robocalls and wrong numbers has a very different bill on a per-call plan than on a per-second one, and no review site can tell you which you are.

Two minutes per call is the planning figure we use when a business has no log at all. Treat it as an assumption to replace, not a measurement — your own average will be different, and it is the number that moves the bill most.

Which kind of service does your call log point at?

Four things can answer a business line, and your log usually points at one of them before you have compared a single price.

Read the last row against your own four numbers. If the misses cluster between 6pm and 8am, you are buying after-hours coverage and should not pay for a daytime plan to get it. If your misses are at 11am on a Tuesday because you were under a car, you are buying overflow, which is the case our page for auto shop call answering in Hamilton is built around. If the log is mostly people trying to book, you are buying a booking capability, and that is the one most services quietly do not have — a dental clinic does not need a better message taker, it needs the slot filled before the caller hangs up.

How do you build a shortlist of three without reading twenty review sites?

Take one of each type that fits, and stop at three.

Three is enough because the differences that matter show up in the test calls, not in the feature grids, and because you will not honestly run six test scripts. Pick them by the type your log pointed at, not by search ranking: one staffed service, one AI service, and one of whichever type your log makes the second-best case for.

Most review sites you will find are published by competitors of the services they rank, ours included — the comparison pages on this site exist to sell you something, and so do theirs. Use them for published prices and claimed capabilities, both of which are checkable, and ignore the verdicts. Our own Ruby comparison is written that way on purpose, including the part about when Ruby is the better buy.

Which test calls should you place?

Six, from your own phone, to every provider on the shortlist. Place the same six in the same order so the results are comparable, and write down what happened while it is fresh.

1. Book the most common thing you sell. Not a generic appointment — the actual service, named the way your customers name it. Watch for whether a time is offered and confirmed, or whether you are told somebody will call you back.
2. Ask something it cannot know. A price that depends on seeing the job, or a policy you have never published. The correct answer is "I don't have that, let me get someone who does." An answer that sounds confident and is invented is the disqualifying result, and you will only find it by asking.
3. Say the word emergency. Confirm the call reaches the person you nominated, and ask what happens if that person does not pick up. Escalation that ends at a voicemail box is voicemail with extra steps.
4. Call with noise, or in the other language your customers use. A running tap, a parking lot, a bilingual caller who switches mid-sentence. This is where staffed services depend on who is on shift and AI services depend on what they were built for.
5. Call at 11pm on a Sunday. Whatever the coverage claim says. Until you have dialled it at an awkward hour, 24/7 is a sentence on a website.
6. Read what the call produced. The transcript, the message, the calendar entry. Is it something you could act on tomorrow without phoning the customer back to ask what they meant?

Ours answers on +1 437-494-9110 and the agent is called Maryann. Run the identical six on us.

What should you ask before you sign?

Five questions, and the answers are usually not on the pricing page.

What is the billing unit? Per answered minute, per call, or a monthly plan with included minutes. This single choice moves the bill more than the headline rate does. Checked 25 September 2026: Ruby publishes minute plans — 50 minutes at $250 USD a month, 100 at $395, 200 at $720, 500 at $1,725 — which works out to between $3.45 and $5.00 USD for every included minute. Smith.ai publishes per-call plans, starting at $300 USD for 30 calls with additional calls at $11.50 each. Abby Connect publishes minute plans from $165 USD for 50 minutes. Check each provider's site for current figures; all three price in US dollars, and the Canadian arithmetic is worked through in what an answering service actually costs over a year.

What is the increment? A 45-second wrong number costs $0.45 on per-second billing, a full minute on rounded-minute billing, and a whole call charge on a per-call plan. If a seventh of your traffic is robocalls, that difference is the largest line in your comparison.

What happens on overage, and what happens on underuse? Ask for the overage rate in writing. Smith.ai publishes $11.50 per call over its Starter allowance; Ruby's pricing page does not publish an overage rate at all, which is not a criticism — it is a question to ask before signing rather than after. Unused minutes almost never roll over.

What is the term, and how do I leave? Abby Connect states it is month to month with a free trial on every plan; Smith.ai states it is month-to-month with a 30-day money-back guarantee of up to $1,000 for new clients, excluding overage, and a 10% discount for subscribing twelve months. We have no contract and no minimum, so there is nothing to cancel. Whatever you are told, get the notice period in writing.

Who owns the recordings and transcripts? You want every call, not a summary someone else wrote, and you want to keep them if you leave.

How do you score three providers without kidding yourself?

One sheet, five columns, scored 0, 1 or 2 per test call.

Did it answer. Did it understand what you asked. Did it complete the task or only capture it. Did the escalation reach a person. What the call cost on that provider's billing. Six calls, five columns, sixty points at most — and the total matters less than being forced to write a number in every cell instead of recalling an impression a week later.

Two rules keep it honest. Score immediately after each call, before the next one. And weight the booking call and the escalation call double, because those two are the reason you are buying anything — a service that scores well on politeness and badly on those two has failed the test it was given.

When should you not buy an answering service at all?

Four cases, and they are common enough that this section is here rather than buried.

If you take fewer than about twenty calls a month and you answer nearly all of them, no service will find you money. The phone in your pocket is already doing the job and a scoresheet will not change that.

If the first conversation is itself the service you sell — a consultant, a specialist whose clients chose a named person — a screening layer in front of it costs more than the missed calls do.

If nobody in your business will spend an hour writing down your wording, your prices, your hours and your escalation rules, do not buy yet. Every service on your shortlist, staffed or AI, is only as good as that hour, and skipping it produces a bad month you will blame on the provider.

And if your problem is that you never call people back, an answering service moves the problem rather than solving it. The message will be captured accurately and sit unread. Fix the callback habit first; it is free.

What should the first month look like after you choose?

Narrow, then wide.

Forward only after-hours and overflow calls for the first month — the calls that were already reaching voicemail. Nobody's working day changes, the risk is close to zero, and at month end you hold a transcript of every call you used to lose. Then count how many were real enquiries. If thirty were, you have an evidence-based case for forwarding the line full time. If three were, you have learned something worth knowing for the price of a few dollars in answered minutes.

Read every transcript for the first two weeks, not a sample. That is where you find the wording that needs fixing, and it is the only fortnight in which you will be motivated enough to do it.

Competitor pricing and terms above were checked on the providers' own pricing pages on 25 September 2026 and change without notice; confirm current figures on their sites before deciding.