Research
Growing your restaurant by improving service quality.
How to "make every ticket count" with Tico.
A new guest costs something to earn — marketing, word of mouth, a first impression that has to land. A returning guest costs almost nothing and spends with less hesitation, because they already trust you. Repeat business is the single most powerful lever on a restaurant's bottom line — industry figures (POS/marketing vendors, cited for magnitude not precision) put acquiring a new customer at roughly 5–25x the cost of retaining an existing one, who also spends 25–67% more per visit.
So the real question isn't whether service quality matters — every operator already believes it does. The real question is which specific parts of the experience actually earn loyalty. That turns out to be answerable.
What the research shows
DINESERV is the standard framework hospitality researchers use to measure what actually drives restaurant service quality. Rather than treating "good service" as one vague feeling, it breaks the guest experience into five measurable dimensions and tests, across real guests, which ones actually predict whether someone comes back.
Two of the five matter more than the rest for whether a guest becomes a repeat guest:
- Assurance — a server's demonstrated knowledge and confidence: do they know the food and drinks, present them well, handle a question without hesitating. A guest asks if a dish can be made dairy-free — a server with real Assurance answers immediately; one without it says "let me check" and walks off. Same eventual answer, different guest experience. Assurance consistently ranks among the top predictors of guest loyalty, and separate research on real recommendation behavior found it directly moves what a table orders and spends.
- Empathy — the ability to read an individual guest and shape the interaction around them, rather than running the same script at every table. Adjusting recommendations to a table's situation — celebrating, in a rush, undecided — and asking the right questions to shape recommendations around what this specific guest actually wants. A table mentions it's a birthday, and a server with real Empathy picks up on it, maybe suggesting a dessert to share unprompted. Empathy ranks alongside Assurance as a top predictor of loyalty — guests remember feeling personally looked after far more than they remember any single procedural detail.
The other three dimensions — Tangibles, Reliability, Responsiveness — cover cleanliness, order accuracy, and speed, and are what most Steps of Service programs are built to encode. That's because they can be tested: a manager can watch a shift, run a checklist, catch a miss. Real, ongoing work — but testable work, with a clear right answer.
Assurance and Empathy get essentially none of that same structure, even though they're a leading driver of repeat business — not because restaurants don't care, but because a checklist or quiz structurally can't reach them. A menu quiz can confirm a server knows a dish's ingredients, a small, static slice of Assurance, but it can't test confident presentation, off-menu knowledge, or on-the-spot customization. Empathy is harder still: it's entirely about reading a specific guest in the moment, which by definition can't be captured by any static test. That's not a training oversight to fix with a better quiz — it's a structural limit of the format itself.
What actually closes the gap
Two questions worth asking directly: do servers actually differ on Assurance and Empathy, and can they close that gap just by working the floor long enough?
- Yes, likely substantially. Sales research finds individual knowledge alone explains nearly half of performance variance between salespeople. We're extrapolating that to restaurants, not citing a study that tested it there directly.
- Only partially, on their own. Skill-development research consistently shows unstructured experience plateaus — fast early gains, then flat, regardless of tenure. Sustained improvement past that point takes structured practice with real feedback.
- The likely reason: fear. Guiding a recommendation without being sure the depth is there feels costly to get wrong in front of a paying guest, so the safer default is to undersell. The research behind this documents fear specifically around suggestive selling; extending it to the whole skill bundle is our own reasoning, not a separately tested finding.
If these are live, demonstrated skills rather than facts to memorize, the fix has to be live and demonstrated too: repeated rehearsal of real guest interactions, with real feedback on how it landed. One clear example is suggestive selling — actively guiding a guest toward specific dishes, drinks, or choices, rather than simply taking whatever order they arrive with. Done well, it's much bigger than a scripted line: reading and matching the guest's vibe throughout the visit, and becoming their knowledgeable ally for the whole interaction. That's a rehearsal problem, not an information problem.
The Tico solution
Tico is built to be that rehearsal space, through two connected pieces. (This is our thesis for closing the gap, built on the research above — not yet a controlled result showing this specific approach moves the needle.)
- Insights — real visibility into where each server actually stands on Assurance and Empathy, tracked from real practice, not a vague manager impression.
- Train — a set of learning tools built to replace flash cards, not resemble them. It brings a restaurant's menu data, back-of-house knowledge, and POS data on what guests actually order into a single knowledge base that powers AI-driven roleplay a server can practice with on their phone, in minutes — with proposed mechanisms, like skill tiers and visible progress a server can show a manager, for making that practice something staff want to do, not one more mandatory task.
The research, methodology, and full citations behind these findings are in the appendix below, for anyone who wants to go deeper.
Appendix: Supporting research
Everything below backs the findings above. Skip to whatever's relevant — it isn't meant to be read start to finish.
What DINESERV measures, and why
DINESERV (Stevens, Knutson & Patton, 1995) is a restaurant-industry adaptation of SERVQUAL, factoring the guest experience into five dimensions:
- Tangibles — the physical, visible elements: cleanliness, furnishings, staff appearance.
- Reliability — doing what you said you'd do, consistently and accurately: correct orders, consistent quality.
- Responsiveness — speed and willingness to help when a guest needs something.
- Assurance — staff knowledge and confidence — whether guests trust the person serving them actually knows what they're doing.
- Empathy — reading and responding to what this specific guest wants, not treating every table identically.
Tangibles, Reliability, and Responsiveness are procedural — checklist behaviors you can audit directly. Assurance and Empathy are demonstrated competencies, shown in the moment rather than checked off a list — which is exactly why they're harder to train.
From trust to behavior: why Assurance isn't just a feeling
DINESERV establishes that Assurance and Empathy predict satisfaction and loyalty. A separate peer-reviewed study goes further and asks whether perceived server knowledge changes what a guest actually orders. Borchgrevink and Susskind's Consumer Acceptance of Server Recommendations found that perceived server knowledge has a real, measurable persuasive effect on whether a recommendation is followed — guests weigh a suggestion by how knowledgeable they believe the server to be, not just by whether the suggestion is objectively good.
Note: full effect sizes from this study weren't directly accessible in this research pass; treat the finding as directional and peer-reviewed, not a precise statistic.
Suggestive selling is Assurance and Empathy in practice
A genuine, well-informed recommendation — grounded in real dish knowledge and real attention to what this guest wants tonight — is Assurance and Empathy, demonstrated live. Pairing is the clearest example. Naming a technically correct pairing is a knowledge test with one right answer. Pairing well is a small conversational loop: asking what this guest wants tonight, then threading the recommendation through the answer.
Why it doesn't happen: a fear gap, not a motivation gap
Industry commentary (practitioner sources, not peer-reviewed research) consistently points to fear of rejection as the reason most servers avoid suggestive selling — a guest's "no thanks" often gets processed as personal rejection rather than a neutral answer. The gap isn't unwillingness to give guests a great experience; it's never having had a low-stakes place to build the knowledge, delivery, and comfort with occasional rejection that suggestive selling requires.
The DINESERV research in detail
Two independent studies applying DINESERV/SERVQUAL to full-service restaurants report the standardized beta coefficients below.
Study 1 — Modified DINESERV Structural Equation Model (2023). American Journal of Multidisciplinary Research and Innovation. n=102, structural equation modeling.
| Dimension | β (→ service quality) |
|---|---|
| Empathy | 0.656 |
| Assurance | 0.641 |
| Tangibles | 0.634 |
| Reliability | 0.589 |
| Responsiveness | 0.561 |
Note: full-service status of this sample is unconfirmed; its pattern (Empathy and Assurance highest) is consistent with Study 2 below.
Study 2 — Sudan Restaurant Industry (Diab et al., 2016). Marketing and Branding Research, 3, 153-165. n=387 across 4 full-service restaurants in Khartoum, Sudan. Full text.
| Dimension | β (Satisfaction) | β (Loyalty) |
|---|---|---|
| Assurance | 0.31 | 0.17 |
| Empathy | 0.23 | 0.24 |
| Tangibles | 0.21 | 0.23 |
| Reliability | 0.16 | 0.07 (weak) |
| Responsiveness | 0.02 (n.s.) | 0.04 (n.s.) |
The summary: averaged impact across both studies
Each study's betas were normalized to its own 100%; Study 2's two models were averaged together first, then averaged against Study 1.
| Dimension | Average share of impact |
|---|---|
| Empathy | 24.8% |
| Assurance | 24.4% |
| Tangibles | 23.6% |
| Reliability | 16.2% |
| Responsiveness | 11.0% |
Empathy and Assurance rank first and second, with Tangibles close behind. As a pair, Empathy and Assurance account for 49.2% of total impact against 50.8% for the other three combined — essentially even, not lopsided. The defensible claim is a ranking: the two highest-ranked individual levers, not two dimensions that dwarf everything else.
Two additional studies (a Malaysian fast-food sample and an institutional university-dining sample) were reviewed and set aside as not full-service.
What menu quizzes test — and the larger part of Assurance they miss
Product knowledge quizzes are a real but narrow test of Assurance. A guest doesn't experience Assurance as a correct ingredient list recited back; they experience confidence, fluency, and the sense the server knows more than what's printed. Four things are bundled into this one dimension, and a quiz only reaches the first:
- Can you name it? Static recall — what a quiz already tests.
- Can you present it with confidence and desire? Delivery. No recall test touches this.
- Do you know what's not on the menu? Off-menu knowledge — substitutions, seasonal items.
- Can you adapt it on the spot? Customization — allergies, substitutions, cross-menu recommendations.
A printed menu is a sales tool aimed at the guest, not a training document aimed at staff — expecting a quiz built from it to measure real Assurance is a category error. Assurance means the server knows more than what a guest could read on the menu themselves.
References
- Stevens, P., Knutson, B., & Patton, M. (1995). Dineserv: A tool for measuring service quality in restaurants. Cornell Hotel & Restaurant Administration Quarterly, 36(2), 56-60.
- Borchgrevink, C.P., & Susskind, A.M. (2006). Consumer acceptance of server recommendations. International Journal of Hospitality & Tourism Administration, 7(4), 21-41.
- Modified DINESERV structural equation model study (2023). American Journal of Multidisciplinary Research and Innovation. journals.e-palli.com
- Diab, D.M.E., Mohammed, H.E., Mansour, E.H., & Saad, O. (2016). Investigating the impact of key dimensions of service quality on customers' satisfaction and loyalty: Evidences from the restaurant industry in Sudan. Marketing and Branding Research, 3, 153-165. academia.edu
- Fah, B.C.Y., & Kandasamy, S. Service quality and customer satisfaction among hotels/restaurants in Malaysia (fast-food sample) — set aside: not full-service.
- Kim, H.S., Joung, H.W., Yuan, Y.H.E., Wu, C.K., & Chen, J.J. (2009). Institutional DINESERV, university dining. Journal of Foodservice, 20(6), 280-286. Set aside: not full-service.
- Parasuraman, A., Zeithaml, V.A., & Berry, L.L. (1988). SERVQUAL: A multiple-item scale for measuring consumer perceptions of service quality. Journal of Retailing, 64(1), 12-40.
- Repeat-guest spend and acquisition-cost figures: industry-reported estimates from restaurant POS and marketing-technology vendors (e.g. ChowNow, Toast, Paytronix, Restroworks) — practitioner/vendor sources, cited for directional magnitude only.
- Ericsson, K.A., Krampe, R.T., & Tesch-Römer, C. (1993). The role of deliberate practice in the acquisition of expert performance. Psychological Review, 100(3), 363–406. General expertise research, not restaurant-specific; later replications found smaller effect sizes than originally reported.
- Sales-performance variance and salesperson knowledge: general sales-performance literature reporting that individual knowledge accounts for a substantial share (cited estimates near 50%) of variance in salesperson performance. Cited as a directional parallel, not a restaurant-specific finding.