GM Interview Question Sheet

Twelve questions, in this order

Forty five minutes, two people in the room, and about twenty two minutes of genuine measurement time once introductions and courtesies are paid. That is room for twelve questions asked properly, not sixty asked badly. The order matters: the sensitive instrument fires before anyone knows what is being measured, errors come after rapport is built, and the close flatters and generates design input.

1. Take me through last month, date by date. When did you close, when did the pack leave you, and when did the owner have it?
   Deployment  Three separate dates. Do not let close, submission and owner receipt merge into one. The register measures the lag to the owner, and no owner date was obtained in interview 1.
2. And the month before. What date did it leave you then?
   Deployment  One instance is an anecdote. Two consecutive instances are the beginning of a distribution, and two that differ is the finding.
3. When you plan your month end, how many days do you allow between sending the pack up and getting it signed off?
   Deployment  Ask this cold, here, before anything else touches head office. It sits entirely inside his own planning, so it returns the corporate turnaround as a by-product and costs him nothing to answer honestly. Asked later, after the subject has been raised, the answer is contaminated. This is the only head office question in the set.
4. Between the accounting system closing and you having a P&L you can work with, what happens? Take me through it.
   Deployment  Count the manual steps silently. Never name a tool. If they name a spreadsheet, that is theirs. If we name it, it is ours and worth nothing.
5. Take one number, RevPAR for June. Walk me from the system to the page the owner read.
   Results  Count the hands yourself rather than asking how many people type it. Asking for the count presupposes typing and presupposes plurality, and if the honest answer is that it is linked, we have already told him we do not believe him.
6. What was the last thing your owner asked you after he got the pack? Was the answer already in it?
   Results  Replaces "what pushback do you get from owners", which produced "make more money, reduce cost" and told us nothing. If the answer was already in the pack, that is volume-without-analysis evidenced rather than alleged.
7. For last month, what date were the numbers final, and what date did the last piece of commentary land?Finance, not the GM
   Results  Two dates give the narrative queue length without anyone estimating it. The financial manager owns this pain and the blame lands on the system, not on him.
8. How many times did you have to ask before that last piece arrived?Finance, not the GM
   Deployment  Converts a duration into a chase count, which is a number the person doing the chasing can actually give.
9. Think of the last time a number had to be corrected after the pack went out. What happened, and who spotted it?
   Assess and Refine  The mandatory one. It was skipped entirely in interview 1. The wording presupposes the event, so admitting one costs nothing, and it never uses the word fault. Who caught it matters more than how many there were: if the answer is consistently head office, properties have no detection capability of their own.
10. Most packs get built by rolling last month's file forward. Which line most often gets missed here?
    Deployment  The preamble presupposes roll-forward and presupposes something gets missed, which makes the admission free.
11. Your pack has N sections. Who decided that list?
    Approach  Ask in identical words at all six properties. N comes from the artefact where we hold it. If the origin stories differ between two hotels of the same brand, there is no group approach, only local ones.
12. If you had to cut it in half by next month, what would go, and whose permission would you need?
    Approach  "Nobody's" or "I do not know" means the format is orphaned, which caps everything built on top of it.

The sixty second close, worth more than any question above

The full bank, by friction point

The twelve questions on the previous tab are what fits. This is the reservoir to dip into when an answer opens a door, when a property turns out to be different, or when there is time in hand. Each block gives the opening invitation, the pointed follow-ups tagged to a RADAR element, and the artefact to ask for instead of accepting an assertion.

EFQM RADAR, and why it fits this job

Four questions asked of any process in a fixed order. The shorthand of results, approach, deployment and review maps onto EFQM's official wording: Results, Approach, Deploy, Assess and Refine. Current edition is the EFQM Model 2025, which merged the old direction and execution matrices into one.

Two rules for using it live

• Run it backwards in the room. RADAR is taught left to right for designing a process. For diagnosing one it runs right to left: start with what came out and when, which is checkable, then who did it, then why it is designed that way, which is the only place opinion is legitimate, then what changed, which is hardest to fake. Opening with "tell me about your process" invites the rehearsed answer.
• The master rule caps everything. A criterion is capped at the score of its rationale. Translated: a beautifully executed, endlessly polished pack built on an unsound reason for existing is worth only what the reason is worth. Huge deployment effort sitting on top of weak purpose is exactly what F6 and F10 describe, and RADAR is how that gets stated defensibly rather than rudely.

The rolling wave, complete: all six interviews held

The wave closed on Monday 27 July with Addis in the morning and Dubai in the afternoon. Six sites, four countries, three brands, every introduction Jad made on 11 July converted. What follows the summary block is the position as it stood after three interviews, kept for the record; the per-person tabs carry each later interview in full.

Where the wave landed, 27 July

Position after three interviews, kept for the record:

Provable now, cross-site

The calibration rule, now in force

The design fork Thursday decides

Still missing after three interviews

• Any hours split, numbers versus narrative. Zero for three. The single most important number in Measure.
• Has an error ever reached the owner. Asked zero times in three interviews, despite being mandated after each. Meanwhile the Dar artefact proves four defects DID reach the owner while testimony claimed errors are internal only.
• The corporate review email thread. Requested zero times. One forwarded thread settles the loop's speed, content and round-trip count at once.
• Currency mechanics. EUR packs from dirham books, TZS books under a EUR pack: who converts, what rate, is budget restated. Unavoidable with the birr on Thursday.
• Round-trip count with corporate, owner receipt dates, and who at ONOMO controls the Sage chart of accounts.

Retired, stop spending minutes

• The open month-end narration: three tellings, same shape. Go straight to artefacts.
• "Would you cut the pack": three refusals. It belongs in the owner interviews.
• "Does a manual Excel step exist": established three times. Ask only mechanics, hours, who touches it.
• Corporate-first routing, Rohit as recipient, email as channel: confirmed twice each.

How the month runs at The Pelican

1. Close, 30th to the 3rd
   Long hours, and the finance team is interrupted by live operations such as requisitions rather than being ring fenced.
2. Generate the P&L. Two of them, Rooms and F&B
   Out of SunSystems. Two separate statements at this property, not one split by department. The pack build therefore starts with a merge.
3. Pull to Excel, hand key the statistics, work offline
   The ledger export is "not seamless". Figures such as room counts are typed by hand so the ratios calculate. This is the data integrity weak point, and it is the mechanism, not a chart of accounts problem.
4. Circulate the P&L to all heads of department
   Before the pack is touched.
5. Hold an HOD critique meeting on the numbers
   The off-map step, and the most important structural find of the interview. The numbers are challenged and agreed in a room before the document exists. The analysis event is a meeting; the document is the write up of a judgement already reached.
6. Roll last month's document forward, each manager overwrites their section
   Copy forward confirmed as standard practice, multi hand. No copy forward error reported here, but the mechanism that produces them is present.
7. The Financial Controller chases heads of department for one to two days
   The binding delay. Queue time, not writing time. This is what Philippe himself wants automated.
8. Finance review, then review with the GM, amend, final review
9. Corporate validation with Rohit and Jad, then submit on the 7th
   To head office, not to the owner. Jacob's own walkthrough left the corporate step out entirely; Philippe added it seven minutes later.
10. Pack to the owner, then a meeting date is agreed fresh each month
    No standing date. So the commercially meaningful lag, month end to owner decision, has an uncontrolled tail nobody has measured.

What he claimed, and what it is worth

The owner, and why it matters for the design

A Ghanaian property developer, best read as Devtraco Group though the transcript garbles it three ways, so confirm the spelling before it appears anywhere. Fractional ownership: 134 suites sold to about 100 partial owners, dividends paid annually, and Aleph reports to exactly one owner representative with no asset management team behind him. It is their first hotel; their background is luxury real estate.

What he asked for, in his own words

• Scope: AI reads the numbers and writes the departmental commentary heads of department currently write by hand. He raised it himself. He did not ask for the P&L, the deck build or the rollup to be automated.
• Human in the loop, his own boundary: the front office manager or head of department validates. He never proposed removing the GM or finance gates.
• His before and after: today one to two days of chasing; with AI "10 minutes and another maximum one day for them to validate".
• Success: "Speed and accuracy", 100 per cent accurate, corporate finalisation down from one to two days to ten minutes, and no questions back from the owner.
• Data sources he named: STR and Lighthouse.

What the interview failed to get

• Any hours figure at all. Never asked, in any form. The single most important number in Measure does not exist for this property.
• An owner receipt date, a slide count, a round trip count, and the identity of the PMS and POS, which cannot be the ledger yet supply occupancy, ADR, covers and channel data.
• Whether an error ever reached the owner. The question was in the guide and was skipped.
• He and Jacob had never heard of Claude. They use paid Gemini and ChatGPT for email triage. The property that volunteered to pilot does not know what it would be piloting on.

The pipeline, end to end

• Opera and Simphony feed Sage daily, but the interface drops the statistics: segmentation, room nights, occupancy are picked manually from Opera and keyed into Excel.
• Materials Control has no interface to Sage: literal double entry. Every receipt, purchase and issue is posted twice, and the P&L cost of sales disagrees with the MC report every month. A pure waste stream, and a self-contained quick win.
• Sage to the Aleph template: the mapping mismatch above.
• Departments feed a Teams-shared Excel (Kirsten revenue, Emmanuel payroll last, HODs training to Emile personally, since there is no HR person).
• Gosbert copies it into PowerPoint as pictures and re-copies on any change.
• Excel P&L to Rohit by email around the 7th; his three or four questions are on the P&L, not the pack, and are business-why questions.

Claims, and what they are worth

The rollup, in three phases

• Phase 1: he received three property packs, authored the fourth himself, and consolidated all four alone. "Awful days." "Stuck behind my office spending one week doing just tables." No assistant, no finance help.
• Phase 2: finance built a consolidation Excel that pulls from per-property sheets, but the output still crosses into the deck as screenshots, and all commentary is his.
• Phase 3, June: for the first time the regional finance manager prepared the tables; HR and sales plug their own slides; he adds commentary, checks, and harmonises formats. "This month really it was lighter." And the finance-prepared tables still contained wrong data he caught and bounced back. Delegation moved the labour, not the error rate.

The structural findings

• No finance manager in any Morocco property. Finance sits at HQ Casablanca outside his hierarchy; property finance staff are, in his own apologised-for words, "bastard positions... no responsible who can lead them". Escalated for four years; embedding FMs is now "being thought about".
• The region's whole quality gate is one man's CV. He checks every hotel's numbers because his GMs cannot, explicitly so the asset manager never spots a discrepancy live. If he leaves, there is no check at all. The sharpest scaling argument in the programme for a machine verification layer.
• "Accuracy", the discovery of the interview: an ONOMO-legacy consolidation system already holds per-hotel KPIs and P&L and can consolidate regionally, and the pack pipeline bypasses it into hand-built Excel and screenshots. An incumbent tool that exists and is ignored means the failure mode may be adoption, not absence of tooling. Someone must ask Jad or Rohit what Accuracy is.
• Owner governance is now a trichotomy: Accra's weekly informal visits, Dar's sub-30-minute green-light meetings, Morocco's formal per-hotel presentations to a professional in-house asset manager. One-size owner review design is dead; build the data layer once, flex the presentation layer.
• His own escalation to Bani and Jad, before our programme existed: spending "between one week and 10 days preparing presentation... for data from the past" is not logical. F10 has an internal sponsor; quote him.
• His asks: AI training for GMs and HODs (not a tool: "we have many AI tools but we can be lost"), an embedded FM per major hotel, and no retyped data ("on typing we're going to have errors").

The deck is not universal

Grand Iva sends no monthly PowerPoint. The artefact leaving the property is the Excel P&L workbook, "about seven or eight sheets", to the regional finance office. Ali confirmed the same for the other two Saudi hotels. The consolidated deck is built at regional level.

• F2 is not a property-tier defect. It belongs to whoever builds the deck. In Saudi that sits above the property and we have never seen it. Untested, not absent.
• F6 splits into written and verbal. Written narrative here is zero. But there is a couple of hours of discussion with the Saudi VP, HOD explanations and Ali's owner sit-down, and none of it is written anywhere, so it cannot be reused, audited or rolled up. The target changes: not "automate a narrative nobody wants" but "capture verbal justification that leaves no artefact".
• F10 is not conceded. Seven or eight sheets of numbers with zero commentary is arguably volume without analysis in its purest form.
• The estate has two artefact architectures, not one process with variants. A future state built only for property-builds-deck will not deploy to the Saudi cluster.

F11, new: deadline-driven error deferral

A fourth owner model, and a fifth artefact

A local Saudi family with many businesses, cement, laundries, petrol stations, for whom this is the only hotel. No asset manager, no owner-side hospitality expertise. Ali spent four months explaining every purchase request.

Fredrik Reinisch

GM Marriott Marquis Dubai Creek · Area GM Aleph Dubai · also Hilton Dubai Creek and Movenpick Grand Al Bustan

Monday 27 July, 17:00 GST. Online, 45 minutes, both of us joining. The last GM interview. (Moved twice: from Fri 24 July 11:00, then 13:30.)
Jad has confirmed the NDA covers sharing files, so this is the artefact interview.

Added after Addis, 27 July: seven questions the last five interviews created

These are new or sharpened since this tab was written. If time runs short, 1, 4 and 7 cannot be recovered later, because this is the last GM conversation.

1. Who physically builds each of the three decks?
At four sites finance builds the pack. At Addis the GM builds it himself and refuses to delegate, because preparing it is how he comes to own the numbers before facing the owner. Fredrik is the only person who can tell us whether authorship varies inside one portfolio. Ask plainly: for each of your three hotels, who opens the file and types the commentary, and is it the same person in all three? Then: one finance team across the three, or three?

2. Do you send Aleph a forecast twice a month?
Addis submits a forecast file twice monthly, updating forecast one and forecast two. No other site has mentioned this in five interviews. Either it is a local habit or an estate-wide obligation we have missed entirely. One question settles it. If yes: what triggers each one, and who reads it?

3. Two workbook traps, now named and testable.
Addis's finance director demonstrated both live on screen. Ask whether either exists here.
· Does your P&L workbook have a switch between forecast and actual? At Addis, forgetting to flip it silently runs the whole month's comparison against an un-actualised forecast.
· Is there a tab somebody updates by hand each month that everything else reads from? At Addis it is the statistics tab, and one wrong entry propagates into every downstream number.

4. The trial balance. Ask for it by name.
Addis's workbook reportedly carries a trial balance tab. DH holds no trial balance for any property, and that is the single biggest constraint on the programme: without one, nothing we check can be called verified against source, only internally consistent. Ask: does your monthly workbook carry a trial balance tab? If yes, that is the artefact to leave this call with.

5. What does each owner read first, and what do they skip?
One owner told us through his GM that he is not interested in the expense lines at all, only revenue, receivables and cash. That is the clearest materiality signal any owner has given, and it cuts against a pack organised around GOP. Fredrik has two very different owners: a first-time merchant-family joint venture and an establishment-linked institutional group. What does each turn to first, what do they never ask about, and do the two decks differ because of it?

6. Does anything in your reporting differ by BRAND, or only by OWNER?
The franchise point is settled: two GMs have now said unprompted that the franchisor is not the audience. Do not spend time establishing it again. Spend it on the consequence: across a Marriott, a Hilton and an Accor property, is there a single line in the monthly reporting that changes because of the flag, or is the flag irrelevant to everything except which systems the numbers come out of?

7. THE CLOCK. Five interviews, five misses, last chance.
We still have no close date, no submission date and no hours figure from anywhere in the estate. Script it, ask it early, and ask it as hours of the person who builds the file, not days of the GM:
"Of the total hours that went into last month's pack, roughly how many were pulling and checking numbers, and how many were writing words?"
Then the anchor that needs no memory at all: "Could you forward me the email you sent last month's pack with?" The header settles the date without anybody having to recall it.

The evidence close: do exactly what worked at Addis

Four interviews ended with goodwill and nothing else. The fifth ended with three files offered, because they were named out loud on the call while a screen was being shared. Repeat it precisely.

Name these five before the call ends:

1. Last month's pack for one property, native file, his choice which.
2. The Excel P&L workbook behind it, all tabs, not a PDF.
3. The trial balance, if the workbook carries one (question 4).
4. The forecast file, if question 2 turns up a twice-monthly submission.
5. The transmittal email for last month's pack, forwarded as-is.

Do not end on "send us whatever you have". That produced nothing four times. One named file per ask, and a date against each. And if he offers to screen-share, take it: both Addis mechanisms were found that way, not by asking.

Three brands, two owners, one operator

This is what makes Fredrik unlike every other interviewee, and he is now the sixth and final conversation, so everything still unresolved after five sites lands here. He is not a Marriott GM who also has a pack. He runs a Marriott, a Hilton and an Accor property at once, for two different owners, under one operator. Three franchise regimes, three system stacks, three loyalty programmes, three quality regimes, and only one reporting obligation that actually matters. He is the natural experiment the whole programme has been missing: everything we have triangulated across four countries, he lives in one city.

The questions, rebuilt

1. In a normal month, list every document that leaves each of your three hotels, and who receives each one.
   Deployment  Artefact-first, and never name the artefact. At Abha we asked how many slides were in the PowerPoint and burned four minutes discovering there wasn't one. With three brands and three owners he may have three different shapes. Let him enumerate.
2. Where does the reporting rigour actually sit for you: the brands, the owners, or Aleph?
   Approach  Expect owners and Aleph. It validates him, corrects our record from his own mouth, and opens the real ground. If he says the brands, that is a finding in itself.
3. Walk me across your three hotels. Do Marriott, Hilton and the Movenpick each report to Aleph the same way, or three different ways? Where do they differ most?
   Deployment  The question only he can answer, and the spine of the interview. One person, three brands, three system stacks, two owners. Everything else on this tab hangs off how he answers this. Let him range; do not narrow it to Marriott.
4. Do all three send Aleph the same template? Do the three run different accounting systems, and does the same problem show up in all three?
   Deployment  The fastest confirmation available of the biggest finding in the programme. Four interviews found three ledgers failing against one template. He can settle in a sentence, across three brands at once, what took us four countries to triangulate.
5. For each brand, what does the franchisor actually ask you for, versus what it already holds in its own systems?
   Results  Tests the data-source reframe across all three. If the answer is "they already have it", the follow-up writes itself: could the Aleph pack be fed from that data instead of re-keyed? And whether Marriott, Hilton and Accor differ here is itself useful.
6. Take one hotel's ledger export and the Aleph template side by side. Which lines have to be moved, added or re-coded before they agree?
   Deployment  Then: is there anything the template relies on that your system simply does not carry? Dar's defect was missing cost centres; Abha's was a missing analysis dimension. Different failure modes, so ask both.
7. Where do you think that template came from, and who maintains it? Has anyone ever asked Aleph to change it?
   Assess and Refine  Two sites have absorbed this defect for years without ever escalating. He is senior enough to know who owns the file, and that is the single most valuable answer available tomorrow.
8. You report to two different owners here, West F5 for the Creek pair and Dubai Developments for the Movenpick. Do they want the same thing, and how differently do the two conversations run?
   Results  A contrast no other property offers, inside one Area GM: a first-time merchant-family JV versus an established, royal-linked institutional owner. How the reporting flexes between the two is a direct read on what the future-state has to serve. Also ask whether either has an asset manager, and which pages they actually work from. He closes the owner-governance range: we have had a weekly-coffee owner, a green-light meeting, an asset manager, and a first-time family.
9. Take June, one property. On the 30th, who sat down first and for how long? Then who? Hours per person. And of that, how much was pulling numbers versus writing words?
   Results  Four interviews, zero hours splits. Aim it at whoever builds the pack, not at him. This is the last chance to get the number the whole Measure phase rests on.
10. Has an error ever reached the owner? What was it, who caught it, what happened?
    Assess and Refine  Asked zero times in four interviews. Put it at minute 25 so it cannot be squeezed again. Then the new one from Abha, in the same words: when a correction lands after the pack has gone, does it get fixed in that month or carried into the next?
11. As Area GM, do you consolidate across the three, and do you ever receive a consolidated deck back from the region?
    Deployment  Abha ships a workbook and someone at regional builds a deck nobody at the property sees. Three properties feed a deck we have never asked to look at. He is exactly the person who would know whether that happens in Dubai too.
12. When you send yours up, how long before you hear back? Is that the same for the other two, or different?
    Results  Identical words at every property, so the answers stay comparable. The second half is his alone: if the flagship is answered in minutes and the smaller properties wait days, the corporate loop is a capacity queue, not a service level, and it will not survive the estate doubling.
13. Across every brand and complex you have run, what is the best owner reporting you have seen, and what made it good? What would you steal for Aleph?
    Approach  He has run more brands than anyone we will speak to: Al Habtoor City into three Hilton brands, six brands at Yas Plaza, JA, EDITION, now Marriott, Hilton and Accor together. He is the only interviewee who can tell us what good looks like from the inside, across the whole industry, not one flag.

Drop, to buy the time

• The full step-by-step process narration. We have four, and they are broadly identical. Confirm the ledger-to-Excel bridge in one question.
• Generic owner-type discovery. Ask only the two discriminating questions: asset manager yes or no, and which pages they work from.
• Most generic SOP scoping, now run four times. Keep the customisation percentage and the delivery-mode question only.
• Career and market discovery. Best-documented CV in the set: use it for the opener, not for airtime.

Live technical questions, credible and current

• All three brands are mid-migration on their core systems at once. Marriott is replacing its property management system across 2026 to 2027, Hilton is moving off OnQ, and Accor is consolidating on Opera Cloud. Where are his three in that? A system change mid-build is a real risk to anything we design, and he is carrying three of them.
• He staffs revenue management across three brands, each gated to its own certification and toolset (Marriott One Yield, Hilton GRO, Accor IDeaS). How he clusters that across one team is the same shape of problem as clustering the reporting, and his answer previews how hard the reporting cluster will be.

Where the six stand

Owed, and outstanding (swept Wed 22 July)

• Pelican pack: IN. Philippe replied Tuesday evening with the pack attached, cc Jad and Rohit. His wording says "(Blank)": verify whether it is the real May pack or a blanked template before booking it as the artefact. The file sits in the Aleph correspondence folder.
• Dar samples: PROMISED for close of business Wednesday. Emile confirmed he and Gosbert will send the named list. Chase Thursday morning if silent.
• Abdeljalil chase: sent Tuesday 17:35, no reply yet. His June owner presentation is Wednesday; the finance-draft version of the deck is the time-critical item.
• Owner interviews are moving. Jad sent "Your valuable input needed" on Tuesday to Mehdi El Baqali, asset director of MHC (Maghrib Hospitality, the Michlifen and Marchica owner), cc Abdellah Essonni of Aleph, proposing a 30 minute owner conversation covering both the BRM and SOP work. Nitin acknowledged; awaiting the owner's signal, then coordinate times. Register item 11 unblocking. Note the new Aleph name: Abdellah Essonni, plausibly the Morocco regional VP Abdeljalil mentioned.
• Fredrik: RESOLVED. Jad forwarded the introduction again on Wed 22 and Fredrik replied within the hour, offering online or in person. Friday 24 July at 11:00 proposed. The drafted nudge to Jad (id 33) is now redundant and should be discarded.
• Received dates from Rohit for the last six packs per property. Still the highest-value unsent email in the programme.
• Ask Jad or Rohit what "Accuracy" is, and who at ONOMO controls the Sage chart of accounts.
• Confirm garbled names before anything client-facing: Devtraco, African Hotel Development, the Masira property, "Squad Hotel".
• We owe Dar a return and Gosbert is waiting. Interview meeting reports from the recording assistants (Read.ai for both Tuesday interviews, Fireflies for the SOP scoping call) are in the inbox if a second transcript source is ever needed.