TL;DR: Most upselling training fails because it hands staff a script instead of a mindset. This five-module framework trains the mindset first, which is why the numbers actually move. Run it in sequence. The role-play at the end is not optional.
I spent a decade behind the bar. I've watched servers deliver upsell lines so wooden you could build furniture with them. "Can I start you off with some appetizers tonight?" delivered in the same tone as "do you want fries with that." The table feels the friction. The server feels the friction. Nobody wins.
Here's the thing nobody says out loud: the script isn't the problem. The script is a symptom. The problem is that the person delivering it doesn't actually believe the thing they're saying. They're performing a pitch, and guests have been pitched enough times to recognize the performance in about four seconds.
I've also watched servers who never once used a scripted line move checks by 30%. Not because they were charismatic outliers. Because they knew the menu cold, they were actually listening when the guest talked, and the suggestion they made was a real one. It felt like a favor because it was.
That's the difference this training is designed to build. Not a better script. A staff that's good enough at their job that the upsell is a natural extension of doing the job well.
Operators who try to shortcut this by skipping to "here are the phrases to use" will get short-term compliance and long-term mediocrity. The framework below is sequenced the way it is for a reason.
Before a single upsell word comes out of anyone's mouth, the guest has to feel like the server is actually present. This is the foundation everything else sits on.
Warm first impressions matter more than most training programs admit. Eye contact, using a guest's name when you have it from a reservation, a greeting that sounds like a person and not a policy. These are not soft skills. They're the mechanical preconditions for everything downstream.
Active listening is where most staff actually fall short. The guest who mentions they're celebrating something, or that they don't like anything too heavy tonight, or that they've been here before and always get the same thing: those are all live data points. The server who catches them and responds to them is now operating with information the menu doesn't have. That's a real competitive advantage. The server who's already thinking about table twelve misses all of it.
The practical drill here: teach staff to repeat back one thing the guest said. Not parroting, but reflecting. "So you're leaning toward something lighter tonight." It signals that the guest was heard, and it creates the opening for a recommendation that actually fits.
This module does the reframe that makes the rest of the training stick.
The frame shift is this: upselling is not something you do to a guest. It's something you do for one. The distinction sounds small and it's enormous.
Language matters here in a specific way. "Do you want dessert?" is a yes/no question with a default answer of no. "Have you had the chocolate lava cake? It's one of the things we get the most regulars coming back for" is an invitation. The guest can still say no. But they're being invited into a conversation, not processed through a checkout.
The suggestive versus hard sell distinction is worth taking seriously. Hard selling creates tension. The server is pushing, the guest is resisting, and the table's energy shifts. Suggestive selling is closer to what a knowledgeable friend does: here's what I'd order if I were you, here's what's actually worth the price. The server who operates from that posture will consistently outperform the one running the script.
Role-play matters here and most training programs rush through it or skip it. Have staff practice the language shift with a partner before they're in front of a real table. The first few times feel awkward in practice, which is much better than feeling awkward in front of guests.
This one's counterintuitive for a lot of servers. The instinct is to perform knowledge: here's what I know about this dish, here's what the chef told me. The guest wants the opposite. They want to feel like they made a great choice. Their experience. Their story.
Customization is one of the cleanest tools here. "Would you like to add the truffle oil?" is a question, but it's also a door. The guest who says yes now has a dish that's theirs. That's a higher-value interaction than a clean order taken efficiently.
The narrative approach works for a simple reason: a dish with a story is more memorable than a dish without one. Not every item needs a dissertation, but one true, specific detail about origin or process is often enough. "The pasta recipe came with our chef from his family in Sicily" is three seconds of context that changes what the guest experiences when they eat it. They're not just eating pasta. They're in the story now.
This module trains staff to think about the guest's experience as the thing they're actually selling, and the food and drink as the vehicles for it. That's a different orientation than "we need to move the duck confit."
"No" is not a dead end. Operators often treat it that way in training, so staff treat it that way on the floor.
The first job when a guest pushes back on a suggestion is validation. Not agreement necessarily, just acknowledgment. "I get it, it's a heavy dish" costs nothing and defuses the friction that would otherwise build. The guest is not fighting the server anymore. Now there's room for an alternative.
The practical application: staff should have two or three genuine alternatives in their head for every item on the menu they're likely to suggest. If the guest says the steak sounds like too much, where do they go? Not to a blank stare and "okay, no problem." To something real. "If you want something lighter but still substantive, the halibut has been getting a lot of love this week."
This is where menu knowledge pays off. You can't offer an alternative you don't actually know. Which connects this module directly back to Module 1 and the listening work. The server who was paying attention knows the guest wants something lighter. The alternative they offer is going to land because it's the right one, not a random pivot.
The smaller adds are where a lot of check growth actually lives. Not the upsell to the premium entree. The glass of wine with the main, the side that actually makes sense with what they ordered, the limoncello at the end.
Menu knowledge is non-negotiable for this to work. A server who confidently says "this pairs well with the Cabernet" because they've tasted the Cabernet and eaten the dish lands very differently than a server who says it because that's what the pairing card says. Guests feel the difference.
The dessert and after-dinner category gets neglected in most operations. The check is effectively written in the guest's head once the entrees clear. The server who makes one clean, confident offer at the right moment can reopen that budget. Not aggressively. One offer, delivered like a recommendation from someone who knows what's good, is enough.
The reinforcement layer matters too. After each shift, staff spend a few minutes surfacing what worked and what didn't. Flash cards on menu knowledge. Quick quizzes on the language shifts. Success stories from the week. This is not corporate training-speak. It's the difference between a one-time lesson and an actual change in behavior.
If you can only do one module this week, do Module 2.
Not because it's the most important foundational skill, but because the frame shift it creates unlocks every other module. Staff who understand the difference between a pitch and a recommendation start listening differently. They start engaging differently. The rest of the framework has soil to grow in.
The biggest mistake operators make is skipping to Module 5 because they want to move the numbers immediately. You'll get a spike, you'll get pushback from staff, and it'll fade. The modules are sequenced the way they are because the mindset has to come before the mechanics.
Get Module 2 running. Do the role-play. Let it sit for a week. Then add Module 1 and Module 3 together, because the genuine engagement and the guest-as-hero orientation reinforce each other naturally.
Role-play is the first place this falls apart. Staff hate it, especially in front of peers. The instinct is to rush it, make it optional, or do it once and call it done. Operators who skip the role-play get staff who have heard the information but can't actually execute it under pressure. Run the role-play. Keep it short. Make it not miserable by leading with a bad example yourself first.
Manager follow-through is the second failure mode. This training doesn't work as a one-and-done. The feedback loop, the flash cards, the success story sharing: all of it requires someone on the management side to run the ritual consistently. If it happens once during training week and then disappears, the behavior reverts. The system needs a keeper.
The third failure: implementing this on a staff that doesn't trust management. If the team already feels like training is something done to them rather than for them, this framework will feel like more surveillance. Before running this, the operators who will get the most from it are the ones whose staff actually believes the house is on their side. If that's not true yet, fix that first.
Nothing in these five modules is technically hard. The hard part is the sustained commitment to running the system after the initial training high wears off.
Q: Does this work in fast casual or is it just for fine dining?
The language scales down, but the principles don't change. A guest who feels heard and gets a genuine recommendation will spend more and come back more often whether they're ordering at a counter or sitting down for a four-course meal. Adjust the vocabulary. Keep the mindset.
Q: What if my staff resists the upselling training because they've had bad experiences with pushy sales pressure from previous managers?
That resistance is information. It means they've been trained to do something that felt wrong and it didn't work. Start with Module 2 explicitly and name the distinction out loud. You're not asking them to pitch. You're asking them to recommend. Show them the difference. The skeptics usually come around once they see the frame.
Q: How do I measure whether this is actually working?
Track average check per server, not just per shift. You'll see the spread narrow over time as the lower performers move up. You can also track dessert attachment rate and add-on rate per table. Those are cleaner signals than total revenue because they isolate the behavior you're training.
Q: My staff turns over every few months. Is it worth investing in this kind of training?
Yes. Two reasons. First, even partial training moves the numbers for the time someone is on your floor. Second, staff who receive real training turn over less. The investment in their development signals that they're worth developing. That signal matters more than most operators think.
Q: Do I need to hire a trainer for this, or can my floor manager run it?
A strong floor manager can run this. The materials are in the modules. What you need from leadership is buy-in on the role-play and consistent follow-through on the reinforcement layer. If your manager will actually run the feedback loops, you don't need outside help. If that's not realistic, get outside help.
If you're a restaurant operator who wants the full training system built out and running in your operation, that's what ASM Command on jlittrell.com is built for. Documented, automated, off your plate.
If you want more playbooks like this one, The Ops Wire is where they live. One system per week, in your inbox.
Jason Littrell spent 10 years behind the bar in NYC (including Death & Co) and served as USBG NYC president. He now runs his hospitality consulting firm entirely on AI. He hosts the Hospitality Strategy Lab podcast and writes The Ops Wire newsletter.
Jason