Key takeaways
- Pooling is the real case for a central line: in the example, 20 stores need 20 agents at peak in one queue instead of 40 covering their own phones.
- Keep each store's local number on its listings and forward it; Google asks for a number that connects to the individual location[2].
- Most central lines fail on menu sync. Agents should key orders through the same online ordering every store uses, so 86s and prices match.
- A center rarely pays for itself on store payroll alone. Write service levels down and measure the revenue it recovers.
Why groups move phones off the floor
Somewhere between the fifth and the tenth location, the phone stops being a store problem and becomes a group problem. Each restaurant answers differently, the same caller hears three different pickup times from three stores, and the operations director learns about a lost catering order from a review. Speed is what guests notice first: 94% of consumers say speed is critical when ordering off-premises[3]. A central line promises one standard for every call. Whether it saves money depends on arithmetic most groups never run.
The arithmetic is pooling. Take an example store with 14 calls in its peak dinner hour, at about three and a half minutes each, keeps a phone busy for roughly 49 minutes of that hour. To answer 80% of those calls within 30 seconds, the standard queueing formula for call centers (Erlang C) says the store needs two people free to pick up. Send the same calls from 20 stores into one queue and you need 20 agents, not 40, because a quiet minute at one store covers a busy minute at another.
| Units | Peak-hour calls | Agents if each store covers itself | Agents in one pooled center | Pooled agents' busy time |
|---|---|---|---|---|
| 1 | 14 | 2 | 2 | 41% |
| 5 | 70 | 10 | 7 | 58% |
| 10 | 140 | 20 | 11 | 74% |
| 20 | 280 | 40 | 20 | 82% |
| 40 | 560 | 80 | 37 | 88% |
| 100 | 1,400 | 200 | 88 | 93% |
The table also shows where the gains stop. Pooling saves the most seats on the way from one store to twenty. After that the savings flatten, and in the example agents at 93% busy finish one call with the next already waiting all hour, so real schedules add seats for breaks, training and absence. The hard parts at scale are menu accuracy, routing and keeping agents who sound like they know the restaurant.
Three ways to centralize, compared
| Question | In-house center | Outsourced contact center | SI Employee on forwarded lines |
|---|---|---|---|
| Who answers | Your hires, trained on your menu | Vendor agents, often shared across clients | An SI Employee answering in each store's name from approved information |
| Voice orders | Keyed into your ordering system | Keyed in if the vendor has access to your system | Not taken by voice; the official ordering link is texted when the caller agrees |
| Reservations | Booked directly if agents have system access | Booked or passed on, depending on the contract | Taken as a request for the host to confirm |
| What you staff | Agents, a supervisor and a schedule | A manager on your side who runs the vendor | No agents; a manager path for handoffs, with a backup |
| Menu and 86 updates | Same company as operations, so the fastest | Needs a feed or portal and a turnaround promise | Updated in the approved information |
| Hours | As long as you staff it | Often around the clock | All day, after hours or overflow, on the schedule you set |
| Best fit | Large groups with heavy voice ordering | Groups that want coverage without hiring | Groups whose calls are mostly questions, bookings and catering leads |
The third column is narrower than the others, on purpose. An SI Employee does not key voice orders into a POS. It answers questions from the information you approve, takes reservation requests (it never promises a table), captures catering leads for your catering manager, texts the official ordering link when a caller agrees, and hands complaints and allergy questions to the manager path you choose. Each call arrives as an emailed summary, with the recording and transcript in the account, and a manager can listen in live, whisper guidance or take over. Call details can also go to your own systems by webhook. A group with heavy voice ordering can pair it with a small human order desk.
Staffing an in-house center
Forecast by half-hour
Pull eight weeks of call logs from every store and build a half-hour curve for each weekday. Tracking call volume by hour shows the method.
Time the calls
Measure average handle time by call type: talk time plus the time to key the order. A phone order with a full read-back runs far longer than an hours question.
Pick a service level
Write the target down before hiring, for example 80% of calls answered within 30 seconds.
Run the queue math
Feed calls and handle time for each half-hour into an Erlang C calculator to get the agents needed per interval.
Add the real world
Add seats for breaks, training, absence and turnover, then build shifts that cover the lunch and dinner peaks with part-timers.
Name a supervisor
Someone who listens to recorded calls, keeps the menu notes current and owns the link to store managers.
Pay is the other half. Customer service representatives earned a median $21.53 an hour in May 2025[4], against $15.00 for restaurant hosts[5]. An agent who takes orders for 20 restaurants all evening is doing the first job, not the second, and you will be hiring against contact centers for the same people. Employer costs tell the same story: compensation cost $34.94 per hour worked in administrative and waste services, the industry group that includes outsourced contact centers, and $20.03 in accommodation and food services[6].
Source: US Bureau of Labor Statistics, 2026[6]
Show the data
| Industry | Cost per hour worked |
|---|---|
| Accommodation and food services | $20.03 |
| Administrative and waste services | $34.94 |
| All private industry | $46.89 |
Example. A 20-unit group with 70 calls per store per day handles 1,400 calls a day. At 3.5 minutes each, that is about 82 hours of talk time. If agents are busy 70% of the hours they are paid for, the center pays for about 117 agent hours a day, roughly 15 eight-hour shifts plus a supervisor. At $21.53 an hour before benefits and payroll taxes, the agent hours alone come to about $2,500 a day. The talk time does not vanish when you centralize; it moves. What vanishes is the waiting by the phone at 20 host stands and the interruptions.
Routing: keep each store's number
Do not replace store numbers with one brand number. Google's guidelines ask for a phone number that connects to the individual location, and for a local number instead of a central call center helpline whenever possible[2]. Keep each store's number on its Google profile, location page and receipts, and forward it to the center all day or on overflow. The center sees which number was dialed and answers with that store's name. A brand-wide ordering number can sit in the website header and the app, never on store listings. Phone routing for multi-location groups covers the routing tree in detail.
- Caller
- System
- Your team
01Caller
Guest dials the store
The number on that store's Google profile, website page and receipts.
02System
Forwarded with a store tag
The center sees which store number was dialed before anyone answers.
03Your team
Agent answers as the store
Uses the store's name and sees its menu, hours and current quoted times.
04Your team
Order keyed online
Entered through the store's own online ordering, so prices and 86s match.
05System
Ticket prints at the store
The kitchen sees it like any online order, with the pickup time.
06Your team
Exceptions go to the GM
Complaints and large catering go to the store manager, then a backup.
Example call path for an in-house or outsourced center taking voice orders.
Menu sync decides whether guests trust the line
The classic central-line failure is an agent selling the short rib that store 12 ran out of at 7:15, or quoting 20 minutes while the kitchen is at 45. Guests blame the restaurant, not the call center. Five habits prevent most of it:
- Agents enter orders through the same online ordering every store uses, so prices, modifiers and 86s flow through without a second menu to maintain. Keeping menus, prices and 86s in sync shows how.
- Store managers have one button or one message that marks an item 86 everywhere, including the agents' screen.
- Quoted pickup times come from each store's current ticket times, not a fixed script.
- Each store sends its specials and closures (patio shut, parking lot repaved) to the center before the dinner shift.
- A weekly sample of recorded orders is checked against the printed tickets. Phone orders versus online orders explains where errors creep in.
What a center saves, and what it does not
Run the store side honestly before you sign anything. The calculator counts the store hours a central line takes off the floor: the walk to the phone, the call, and keying the order.
| Input | Example |
|---|---|
| Calls per store per day | 70 |
| Minutes per call | 4 |
| Share of calls moved to the central line (%) | 70% |
| Store labor cost per hour worked | $20.03 |
| Calls a month | 1,519 |
| Staff hours on these calls | 101 |
| Hours handed off | 71 |
| Monthly value of that staff time | $1,420 |
How it is worked out: Calls a day × 21.7 working days × minutes per call ÷ 60 × share handled = hours; hours × hourly cost.
Assumptions
Example, per store. Multiply by your unit count and compare with the center's full cost.
Example. For the 20-unit group, the store side comes to about 65 hours a day, worth roughly $1,300 at that rate. The center's agent hours cost about $2,500 a day before benefits. So the payroll case alone does not close. The gap has to be covered by revenue: calls the stores were missing, larger tickets from agents who follow an upsell script, and catering leads that now reach a person. Measure those for a month before the switch and a month after, store by store, or the center turns into a cost line nobody can defend.
Service levels to put in writing
Whether the center is yours or a vendor's, write the targets down. A useful yardstick comes from an unlikely place: federal rules for Medicare drug plan call centers require an average hold of no more than 2 minutes, 80% of calls answered within 30 seconds and a disconnect rate no higher than 5%[1]. A dinner-rush order line should be at least that good.
- Answer speed: the share answered within 30 seconds, counted from the end of any greeting and reported by half-hour, not only as a daily average.
- Abandonment: calls that hang up in the queue, by store and by half-hour.
- Order accuracy: a weekly sample of recorded orders checked against tickets.
- Menu changes: an 86 or price change live on the line within a set number of minutes of the store reporting it.
- Escalation: complaints and catering above a set size reach the store manager the same shift, with a named backup.
- Recordings and data: who owns recordings and caller details, how long they are kept, and how you get them back if you leave.
Ask outsourced vendors about staffing stability too. Employment at US telemarketing bureaus and other contact centers fell 7.5% in 2025, to about 294,000 people[7]. Find out how many agents will know your menu, how long they usually stay, and who retrains them when they leave. Restaurant answering services lists more questions for the contract.
Choosing by group size
- 5 to 10 units. Pooling saves a few seats. Fix routing and menu sync first, and cover overflow from outside before you hire a center. Opening a second location is the right moment to set routing rules.
- 10 to 40 units. The pooled queue starts to pay in seats. A hybrid fits many groups: a small order desk for voice orders, with questions, reservation requests and catering leads answered on forwarded store lines.
- 40 to 100 units. An in-house center can keep agents busy across the day. The work shifts to forecasting, quality review and holding managers or vendors to the written service levels.
Franchised groups have an extra layer, because the brand may set or supply the phone setup; see phone answering standards for franchise restaurants. For a smaller group's version of the same problem, the three-restaurant group example models how catering and event calls would reach one events manager.
Frequently asked questions
Will a central phone number hurt our Google listings?
It can work against them. Google asks for a number that connects to the individual location and a local number instead of a central helpline whenever possible[2]. Keep store numbers on listings and forward them to the center.
How many agents does a 30-store group need at peak?
With the example's assumptions (14 peak-hour calls per store, 3.5-minute calls, 80% answered within 30 seconds), Erlang C gives 29 agents in one queue, against 60 if every store covered itself. Add seats for breaks and absence, and use your own call data.
Should the center quote catering?
Usually not. Have it take the event date, guest count, service style and callback details, then route the lead to the catering manager, who quotes. Prices and menus change too often for a script.
Can a franchisor require franchisees to use the brand's call center?
That depends on the franchise agreement and disclosure document, which list required suppliers and fees. Read those before assuming either way; franchise phone answering standards covers what to check.
What should we measure in the first month?
Answer speed and abandonment by store and half-hour, order accuracy from a weekly sample, catering leads passed to managers, and phone-order revenue per store against the month before the switch.
References
Legal Information Institute (Cornell Law School), eCFR text (2026). 42 CFR 423.128 - Dissemination of Part D plan information (call center requirements) (opens in a new tab) ↑
Google Business Profile Help (2026). Guidelines for representing your business on Google (opens in a new tab) ↑
National Restaurant Association (2025). From Trend to Transformation: Off-Premises Dining Now Essential for Restaurant Consumers, Operators (opens in a new tab) ↑
US Bureau of Labor Statistics (2025). Occupational Employment and Wage Statistics, May 2025: customer service representatives, median hourly wage (opens in a new tab) ↑
US Bureau of Labor Statistics (2025). Occupational Employment and Wage Statistics, May 2025: hosts and hostesses, median hourly wage (opens in a new tab) ↑
US Bureau of Labor Statistics (2026). Employer Costs for Employee Compensation, Table 4: private industry workers by industry group, June 2026 (opens in a new tab) ↑
US Bureau of Labor Statistics (2025). Quarterly Census of Employment and Wages, 2025 annual averages: telemarketing bureaus and other contact centers (NAICS 561422) (opens in a new tab) ↑