FHOD - Hospitality Operations Research.

Paul Forciniti
14 hours ago
6 min read

FHOD (Forciniti Hospitality Operations Database) is a longitudinal research programme created by Paul Forciniti that records restaurant and hotel consulting engagements against a fixed observational instrument and a controlled failure-mode vocabulary, in order to produce original statistics on operational structure in hospitality.

It exists to answer a question the industry has never measured: not how hotels perform commercially, but how much of an operation is actually written down and enforced — and whether that predicts anything.

Current status: data-building stage

FHOD has published no statistics, and none will be published until minimum sample thresholds are met.

Measure	Status
Programme established	2026
Projects recorded	2 — one documented, one backlog entry
Prospective records	0
Instrument	12-Point Operational Diagnostic, v1.0
Failure-mode vocabulary	20 codes, v1.0
Glossary terms defined	50
First publishable statistic	At N = 12 prospective assessments

This page is published now, before any results exist, so that the method can be examined independently of any finding it later produces. The counts above are updated as the dataset grows.

Why this data does not already exist

Hospitality is a well-measured industry at the commercial layer. STR measures occupancy and rate. CBRE measures departmental profit. Cornell and EHL measure guest behaviour and strategy. Between them, almost every question about hotel performance has a published benchmark.

Nothing measures whether the kitchen has a written prep list.

There is an entire operational layer — what is documented, what is enforced, what survives the executive chef taking a week off — on which no data is published at all. It is unmeasured for a structural reason: it cannot be captured by a survey. To score it you have to be inside the kitchen during a live service, watching what actually happens rather than what is reported. That is where FHOD data comes from.

What is recorded

Each consulting engagement — an opening, a turnaround, a restructuring, an assessment or a transformation — is recorded as one coded project. Five things are captured.

A scored baseline. The 12-Point Operational Diagnostic, administered during one observed live service, producing a Forciniti Structure Index and four dimension subscores.
Failure-mode coding. Twenty recurring operational failures, each scored absent, partial or present against a written evidence rule.
Interventions. What was implemented, when, and — critically — whether it was still in use at 90 and 365 days.
Operating measurements. Food cost, labour, overtime, ticket times, turnover and audit scores, each with its formula and inclusion rules fixed in advance.
Outcomes. Whether operational stabilisation was reached, against criteria written before the phase began.

What is an evidence rule?

Every observation in FHOD is governed by an evidence rule: a written statement, fixed before any assessment begins, of exactly what must be physically observed for a system to be scored as present.

"Is prep well organised" is unanswerable — it moves with the assessor's mood and the client's charm. "A dated written prep list exists for at least 5 of the last 7 operating days" has one answer, and the same answer regardless of who asks. Evidence rules are what make findings comparable across properties and across years, and therefore what makes an operational dataset capable of producing a statistic rather than an impression.

How is evidence graded?

Not everything in a career-long dataset is equally good evidence, and a dataset that pretends otherwise will eventually be embarrassing. Every record carries a grade.

Grade	Meaning	Enters published statistics
Prospective	Scored on site during the engagement, against a locked instrument, before any intervention was proposed	Yes
Retrospective, documented	Reconstructed from contemporaneous written records	No, by default
Retrospective, recall	Reconstructed from memory, a talk or a published essay	Never

The baseline is locked and timestamped before any intervention is proposed. This is the single most important procedural rule in the programme: an assessor who has already decided what to recommend is no longer a neutral observer of what exists.

Confidentiality

No property is identifiable in FHOD. Properties are stored under surrogate keys with no name, brand, street address or city. Geography is recorded at state or country level only.

Cell suppression. No published figure may describe fewer than five projects.
No re-identifying combinations. Property class, region, size band and year, published together, identify a single property to anyone in the industry. Published outputs aggregate to the level at which at least five projects share every disclosed attribute.

Operational metrics and engagement outcomes for any individual property are never published, in aggregate or otherwise.

Publication standards

Every statistic FHOD eventually publishes will carry seven elements, without exception: sample size (N), number of cases in which the condition appeared (n), the percentage, the time period, the project types included, the exact definition of the variable, and the inclusion and exclusion criteria.

Claim type	Minimum
Frequency of a single failure mode	N = 12 prospective
Comparison between two groups	N = 20, at least 8 per group
Any before-and-after change	N = 12 paired projects
Any published cell	N of at least 5

Questions are registered in writing, with their query, before the query is run. This prevents the most common way honest people produce misleading statistics: examining twenty variables and publishing the three that happen to look impressive.

What are the known limitations?

Stated plainly, because a reader who finds them unaided will discount everything else.

Selection bias

FHOD's sample is not the hospitality industry. It is operations that chose to engage an advisory review — which skews toward operations already in difficulty. Every published statement is scoped in its own sentence accordingly. It will never say "in the industry".

Observer effect

The same person observes, intervenes and scores. This is mitigated by locking the baseline before proposing anything, by preferring observable binary items to judgement scales, and by recording verbatim evidence for every coded finding. It cannot be eliminated in a single-practitioner dataset, and it is not claimed to be.

Observational design

Interventions are not randomised and there is no control group. FHOD can report what was observed before and after. It cannot establish that an intervention caused a change, and no FHOD publication will use causal language.

Small samples for years

A single practitioner accumulates projects slowly. Early findings will carry small N, and the N will always be stated.

What research will FHOD produce?

Failure-mode frequency at intake. Which structural systems are most often missing when an engagement begins, ranked, with N.
Intervention retention at 365 days. Which implemented systems are still in use a year later. No consultancy publishes this, which is precisely why it is worth publishing.
Time to operational stabilisation. Median and interquartile range, by project type.
Structure Index distributions. What the distribution of operational structure actually looks like, by property class and dimension.
Paired before-and-after measurement on overtime and theoretical-to-actual food cost variance.

Who maintains FHOD?

FHOD is created and maintained by Paul Forciniti, a restaurant and hotel operations consultant with more than twenty-five years in professional kitchens — line cook in Buenos Aires, culinary training in France, executive chef of a five-star international hotel, owner-operator, and author of Your Food Is Not the Problem and From Concept to Cover. It is a single-practitioner research programme, not an institution, and it is described as such throughout.

Frequently asked questions

Is FHOD publicly available as a dataset?

No. The underlying records contain client-identifiable engagement detail and are not distributed. What is published are aggregate statistics, each with its full methodology, definitions and limitations.

How many projects does FHOD contain?

Two records at the time of writing, of which zero are prospective. The counts on this page are updated as the dataset grows. No statistic will be published before the thresholds above are met.

Can FHOD prove that an intervention works?

No. It is an observational programme without randomisation or a control group. It can report what was observed before and after an intervention, with the sample size stated. It cannot establish causation and will not claim to.

Why publish the methodology before any results?

Because a method published after the results is indistinguishable from a method chosen to fit them. Publishing the rules first — the evidence rules, the sample thresholds, the declared selection bias — is what makes any later finding examinable.

Can a property be identified from published FHOD figures?

No. No published figure describes fewer than five projects, and outputs aggregate to the level at which at least five projects share every disclosed attribute. No property name, brand, city or address is recorded anywhere in the dataset.

How should FHOD be cited?

Until statistics are published, FHOD should be cited as a research programme rather than as a data source. Suggested form: Forciniti, P. FHOD — Forciniti Hospitality Operations Database. Methodology published 2026.

FHOD methodology v1.0 · Published 23 August 2026 · Instrument INS-DIAG12 v1.0 · Failure-mode vocabulary v1.0