How to Design a Scavenger Hunt in Singapore

What a scavenger hunt is in Singapore, and how it differs from a route race

A scavenger hunt in Singapore is an open-list format in which teams are handed a set of items to find, tasks to complete or places to reach, each carrying a point value, and are then released into a defined area with a deadline and no prescribed order. At least six local operators sell some version of it, usually under the names treasure hunt, city hunt, heritage trail or photo hunt, and published prices for hunt-style formats sit between $25 and $130 per person depending on whether the hunt is app-led, facilitated or fully bespoke.

The distinction from an Amazing Race is the whole reason this page exists, and it needs stating before anything else.

A route-based race is a following exercise. Teams receive one clue, that clue resolves to one station, the station releases the next clue, and the sequence continues until the route is exhausted. The organiser controls the order. Teams cannot skip, cannot reorder, and cannot decide that a station is not worth their time. The skill being tested is execution speed and clue-solving inside a corridor somebody else built.

A scavenger hunt is a prioritisation exercise. Teams receive the entire list at minute zero. Every item is available immediately. Nothing directs them anywhere. The organiser controls only what is on the list, what each item is worth, and when the clock stops. The team decides the order, the route, which items to attempt and, critically, which items to abandon unread.

That single structural difference cascades through every other design decision on this page:

• Scoring changes shape. A race can score on elapsed time because every team covers the same ground. A hunt cannot, because no two teams attempt the same set of items. Hunts must score on points, and the point values become the design.

• Failure looks different. A race fails when teams queue. A hunt fails when the list is completable, because then every team collects everything, every team ties, and the format has tested nothing.

• The organiser's workload moves. A race puts staff at stations. A hunt puts staff at the finish line, checking evidence, and that verification burden is the main hidden cost of the format.

• The team behaviour under test is different. A race surfaces speed and coordination. A hunt surfaces whether a group of six adults can agree on what not to do while a clock runs.

If your brief says the team should learn to make trade-offs under time pressure with incomplete information, you want a hunt. If it says the team should move through a designed sequence of challenges, you want the route format instead, and running a hunt will disappoint you.

Running the hunt from brief to hand-in

The sequence below assumes one organiser, no hired facilitators and a group in the twenty to forty range. It is written so that an HR lead can run it without buying anything except prizes.

1. Define the boundary before you write a single item. Draw the area on a map and name it in words a participant can enforce on themselves: one mall and its immediate street frontage, one heritage precinct between four named roads, one office building including the lobby and the two food outlets on level one. Every item you later write must sit inside that boundary. A hunt with a fuzzy boundary produces a team that walks to the next MRT stop and a scoring dispute you cannot win.

2. Write the item list to the point structure in the next two sections, not the other way round. Decide the tiers, decide how many items sit in each tier, and only then invent the items.

3. Time-test the hardest tier yourself. Walk out and complete two of your top-tier items with a stopwatch. If your own time is materially under your estimate, the tier is mispriced and the hunt will end in a tie.

4. Print the list, and print it in a deliberate order. Ordering is a design lever. A list printed in descending point order pulls greedy teams straight into the expensive items. A list printed in random order forces teams to read the whole thing before deciding, which costs them clock but rewards the team that reads.

5. Brief for ten minutes, and spend most of it on the rules that create disputes: the boundary, the deadline, the proof standard, whether splitting is permitted, and the late penalty. Say out loud that finishing the list is not possible. Teams that do not hear this sentence will spend the first fifteen minutes planning to do everything.

6. Form teams and give each one a two-minute planning window before the clock starts. This window is where the exercise you actually want to observe happens. Do not skip it to save time.

7. Start every team simultaneously from one point. Unlike a route race there is no staggering problem, because there is no fixed sequence to queue behind. Simultaneous starts also make the deadline a single fixed instant for everyone, which simplifies scoring enormously.

8. Stay at the finish point and receive evidence as it arrives, rather than in one lump at the end. Teams that hand in three items at minute thirty give you thirty minutes of quiet checking time you would otherwise not have.

9. Close the clock hard. Announce a five-minute warning, then stop. Apply the published late penalty without discussion. One organiser who negotiates at the deadline teaches every future group that the deadline is negotiable.

10. Score in front of the group, tier by tier. Read out each team's claimed items, confirm or reject each one, and let the totals build in public. Rejections announced publicly are accepted; rejections announced privately are contested.

11. Show every team the full point table afterwards, including the items nobody attempted. This is what turns the debrief from a chat into a post-mortem, because teams can see precisely what they left on the board.

Writing a list where the points force a choice

The failure mode of nearly every scavenger hunt is a list where the sensible strategy is obvious. If everything is worth ten points, teams collect whatever is closest and the winner is whoever started nearest the middle of the map. If the expensive items are also the most efficient items, teams sweep the expensive items and the format tests nothing but stamina.

A list forces triage when three conditions hold at once.

Condition one: the list must be unfinishable by a wide margin. Not slightly unfinishable. If the total time to collect everything is 110 percent of the clock, teams will attempt everything, fall short by two items, and the result will hinge on luck rather than judgement. Aim for a total collection time of roughly three times the available clock. At that ratio no team can plausibly hold the ambition of completing the list, and the planning conversation shifts immediately from sequencing to selection.

Condition two: the point-per-minute rate must not rise monotonically with tier. This is the part almost everyone gets wrong. If your sixty-point items take twice as long as your thirty-point items, the sixty-pointers are exactly as efficient and the choice is a coin flip. If they take less than twice as long, the sixty-pointers dominate and there is no choice at all. The list only becomes interesting when the highest tier is deliberately the worst value per minute, so that the big number on the page is a trap for teams that read points instead of rates.

Condition three: something must make the trap tier occasionally correct. A tier that is always wrong is not a decision, it is dead paper. Give teams a reason to touch it. The cleanest device is a set bonus: designate four items spanning all three tiers as a named set, award a bonus for holding all four, and price the bonus so that the set as a bundle beats every other line on the list. Now a team that reads carefully finds a route into the expensive tier that a team scanning for big numbers will miss.

Practical rules for the items themselves:

• Write items with a binary completion state. "Photograph your whole team touching the same object" is binary. "Find something interesting" is not. Every non-binary item is a scoring argument you have scheduled for yourself.

• Vary the cost axis, not just the difficulty. Some items should be expensive because they are far away, some because they need a stranger's cooperation, some because they need thought, and some because they need every team member present. Teams with different compositions will then correctly reach different conclusions, which is what you want.

• Include at least two items that are cheap and nearly free. They give a losing team something to do in the last five minutes and they stop the final stretch from being dead time.

• Include at least one item that is quietly impossible. A shop that closes at five, a mural currently under scaffolding. The learning is not cruelty; it is that the team which checked feasibility early spent two minutes and saved twenty.

• Never award partial credit. Partial credit destroys the arithmetic, because it converts an abandoned expensive item from a total loss into a cheap consolation, and abandonment cost is the entire engine of triage.

The maths of an open list: item count, spread and clock

Work in three numbers: the clock, the saturation ratio, and the tier rates.

The clock. Published Singapore hunt durations cluster between 1 and 3 hours, with 1.5 to 2 hours the most common band for app-led public trails and 2 to 4 hours for facilitated corporate versions once briefing and debriefing are counted. One published run-of-show breaks a half-day into a 15 minute brief, 15 minutes of team preparation, 1 to 3 hours of hunting and a 15 minute debrief. Whatever the total, the number that matters for the maths is the hunting window alone, not the room booking.

The saturation ratio. Add up your own honest completion time for every item on the list, then divide by the hunting window. Below 1.5 the hunt is a sweep and everyone ties. Between 1.5 and 2.2 strong teams get close to completion and the result turns on pace rather than choice. Between 2.5 and 3.5 the format works: no team can attempt more than about a third of the board, so selection dominates. Above 4 the list stops feeling like a list and starts feeling like a menu nobody read, and teams disengage from the ones they cannot reach.

Item count. Local operators publish anywhere from 10 to 20 checkpoints for a facilitated hunt up to 30 or more challenges on app-led formats. For a self-run hunt the useful range is 18 to 30 items. Below 18 the choices are too few to be interesting and teams can hold the whole list in their heads. Above 30 the reading cost alone eats the planning window, and you will find that teams simply ignore the back half of the sheet.

Tier spread. Three tiers is right. Two tiers gives a binary choice that a team resolves in ten seconds. Four or more tiers blurs the distinctions and teams stop computing rates at all. Set the tiers roughly an order apart in feel rather than exactly: a low tier, a middle tier at about 2.5 times the low, and a top tier at about 6 times the low. Then price the time so that the middle tier is the efficiency winner, the low tier is close behind, and the top tier is the worst rate on the board.

Distribution of items across tiers. Weight the count toward the bottom. A useful split for 24 items is 12 low, 8 middle, 4 top. This shape matters: a large low tier means the board never runs out of legal moves, so a team is never stuck, and the small top tier means the trap is visible but not dominant.

The late penalty. Announce a per-minute deduction rather than a disqualification. A deduction roughly equal to one low-tier item per minute is enough. Deductions preserve the triage decision right at the deadline, because a team holding a nearly finished middle-tier item must now compute whether three minutes of overrun costs less than the item is worth. Disqualification removes that decision entirely and just produces an argument.

A worked 90-minute hunt for 30 people in five teams

Room booking is 90 minutes: 10 minutes to brief and form teams, a 75 minute hunting clock, and 5 minutes to score in the room. The debrief sits outside this window. Thirty participants in five teams of six.

The board: 24 items across three tiers.

Plus one set bonus of 40 points for holding all four items marked with a star, which are one top-tier item, one middle-tier item and two low-tier items. Maximum theoretical board value is 600 points.

Check the saturation ratio. Total collection time 240 minutes against a 75 minute clock gives 3.2. Inside the working band, and comfortably impossible.

Check the rates.

The middle tier is the best line on the board. The top tier, carrying the largest number printed on the sheet, is the worst. That inversion is deliberate and it is the whole design.

Check the set bonus. The starred bundle costs 30 plus 9 plus 4 plus 4, which is 47 minutes, and returns 60 plus 25 plus 10 plus 10 plus the 40 bonus, which is 145 points. That is 3.09 points per minute, better than any single tier, and it is the only legitimate reason on the board to touch the top tier.

Now five teams, five strategies, one clock of 75 minutes.

Team Sweep reads the sheet top down, sees the sixty-pointers, and decides the big items are where the game is. It completes two top-tier items in 60 minutes, starts a third, abandons it after 12 minutes with no partial credit, and has 3 minutes left, which is not enough for even a low-tier item. Total 120 points.

Team Balanced hedges. One top-tier item in 30 minutes, four middle-tier items in 36 minutes, two low-tier items in 8 minutes. That is 74 minutes for 60 plus 100 plus 20. Total 180 points.

Team Filler decides the small items are safe and takes everything cheap. All 12 low-tier items in 48 minutes, then three middle-tier items in 27 minutes. That is 75 minutes for 120 plus 75. Total 195 points.

Team Triage computes the rates during the two-minute planning window, identifies the middle tier as the best line, and refuses the top tier entirely. Seven middle-tier items in 63 minutes, three low-tier items in 12 minutes. That is 75 minutes for 175 plus 30. Total 205 points.

Team Reader is the only team that reads to the bottom of the sheet and notices the star marks. It takes the starred bundle in 47 minutes for 145 points including the bonus, then spends the remaining 28 minutes on three middle-tier items for 75. That is 74 minutes. Total 220 points.

Final standings: Reader 220, Triage 205, Filler 195, Balanced 180, Sweep 120.

The arithmetic that matters is the gap between first and last. Team Sweep worked the entire 75 minutes, never stopped moving, and completed the two hardest things on the board. It finished 100 points behind a team that never attempted a single top-tier item and 85 points behind a team that simply read the whole sheet before starting. The team that tried to collect everything did not lose because it was slow. It lost because it never made a decision, and the point values were built so that not deciding is the most expensive thing a team can do.

Two secondary readings are worth putting in front of the group at the debrief. First, the winner took only 220 of a possible 600, which is 37 percent of the board, and that is what a well-tuned hunt looks like. Second, the spread between second, third and fourth is only 25 points, meaning several different reasonable strategies performed similarly. That is also correct. You want one strategy to be clearly wrong and several to be defensibly right, not one to be obviously optimal.

What counts as proof, and what each kind costs you to check

Every item on the list needs a completion test, and the test you choose determines how much of your own evening you spend adjudicating. The four options are not equivalent.

A code. The item exists at a location, the location carries a short alphanumeric code on a card, a plaque, a receipt or a sticker, and the team writes the code down. Verification is a lookup against your master sheet and takes seconds. It cannot be faked without visiting, unless one team messages the code to another, which is why codes should be unique per item and never repeated across teams. This is the cheapest proof in existence and it should be your default for anything location-based. Its limitation is that it only proves presence. It cannot prove that anything was done.

An object. The team brings back a physical thing: a stamped receipt, a leaflet, a specific item purchased under a spending cap, a signature from a named person. Verification is a glance. Objects prove acquisition rather than presence, which makes them good for tasks involving a transaction or an interaction. They fail when the object is obtainable elsewhere, so specify provenance in the item wording rather than the object type.

A photograph. The most common proof in local hunts and the most expensive to check. A photo proves something happened in front of a camera. It does not prove where, when, or that the whole team was present, and it invites the compliance argument: the item said the whole team, four people are in frame, the fifth says she took the picture. Every photo item you write is a potential dispute. The mitigation is to specify the frame requirements precisely in the item text, including who must be visible and what must be identifiable in the background.

A video. Worst verification cost by a wide margin, and the only proof that establishes duration and sequence. A ten-second video is ten seconds of your attention that a photo would have taken one second. Five teams with six video items each is thirty clips, which is a genuine twenty minutes of scoring you must schedule. Use video only where the task is inherently temporal: a performance, a full sequence completed without a break, an interaction with a stranger that a still image could stage.

Budget your verification explicitly. Take the number of items you expect to be claimed across all teams and multiply by your realistic check time: 5 seconds for a code, 5 for an object, 20 for a photo requiring a compliance judgement, 45 for a video. In the worked example above, five teams claiming an average of eight items each is 40 claims. If they are all photos, that is roughly 13 minutes and your 5 minute scoring window is fiction. If they are codes and objects, it is under 4 minutes.

The single most useful design move is mixing proof types by tier. Put codes and objects on the low tier, where volume is high and you cannot afford per-item judgement. Reserve photos for the middle tier. Allow video only on the top tier, where there are four items and the claims will be few. This lines your verification cost up with your claim volume rather than against it.

Appoint a second checker if the group is above about forty. Not to double the throughput but to make rejections defensible. A single organiser rejecting a photo is making a call; two organisers agreeing is applying a standard.

The heritage variant, and what separates a good clue from trivia

The heritage version ties items to places and to facts about those places. It is the most popular hunt variant sold in Singapore because the raw material is free and abundant: the precincts around Chinatown, Kampong Glam, Little India, Tiong Bahru, Joo Chiat, Balestier and the Civic District are dense with plaques, conserved shophouse features, five-foot ways, temple architecture and street naming that reward looking. Published heritage-format durations run up to about 3 hours with group sizes from 10 to 500, and app-led public heritage trails publish figures around $30 per person for 1.5 to 2 hours across roughly 4 kilometres.

The design problem is that heritage hunts collapse into a quiz with walking attached. What separates a good heritage clue from trivia is where the answer lives.

A trivia question has its answer in a search engine. "In what year was this building completed?" is answerable from a phone anywhere in Singapore, which means the item does not require presence, does not require observation, and rewards whoever has data roaming. You have written a general knowledge test and then made people walk to it for no reason.

A good heritage clue has its answer only on the object. The distinguishing property is that a team standing in front of the thing can answer it and a team sitting in a coffee shop cannot, no matter how well they search. Four reliable constructions:

• Count something. How many of a specific feature appear on this facade. Counting requires physical presence and produces a single unambiguous number, which makes it self-verifying and free to check.

• Read something small. The last two words of a plaque, the name of the donor in the smallest line, a date carved rather than printed. Published sources give the headline; only the object gives the fine print.

• Compare two things that are near each other. Which of these two adjacent buildings has the taller parapet, which of two shrines faces the road. Comparison forces observation and cannot be looked up because nobody has published the comparison.

• Ask what changed. Give teams an old photograph of the spot and ask what is now standing where a named feature was. This one is worth its weight, because it produces the thing heritage hunts are supposed to produce and rarely do: a group of colleagues standing in a street arguing about what used to be there.

Three further rules for heritage lists.

Verify every item on foot within the fortnight. Shophouse frontages are repainted, plaques are removed during works, and businesses close. An item that no longer exists is not a hard item, it is a broken one, and the team that spends twenty minutes on it will be justifiably annoyed.

Respect the sites you are sending people to. Places of worship, cemeteries and conserved residential streets are not props. Write items that require observation from the public realm rather than entry, do not write anything that requires touching a religious object or interrupting a service, and put a line in the brief about noise and photography in residential lanes. An organiser whose hunt produces a complaint has produced a worse outcome than an organiser whose hunt was boring.

Do not require paid entry to score. If an item is inside a ticketed attraction, either cover the tickets for every team or move the item outside. A hunt where the winning move is to spend money is a hunt you have lost control of.

Weight heritage items by walking distance, not by obscurity. In the heritage variant the natural cost axis is travel, which means your top tier is simply the far end of the precinct. That maps cleanly onto the triage maths: the far items are worth more and cost more per minute, so the trap tier builds itself.

The photo variant, and the moderation problem it creates

The photo hunt replaces find-and-record with photograph-and-interpret. Items become prompts rather than targets: a shot demonstrating a named concept, a composition containing three specified elements, a recreation of a supplied reference image, the entire team reflected in a single surface. It is the most reliably enjoyable variant and the one most likely to produce something the company actually keeps, and it is also the variant with the worst structural flaw in the whole format.

The flaw is that photo items are usually judged rather than checked, and judgement does not scale.

A checked item has a defined pass condition. A judged item has a quality ranking. The moment you write "best photograph of teamwork" you have committed to comparing five submissions against a standard you have not published, in front of the people who made them, immediately after they made them. There is no version of that which the losing teams experience as fair.

Four ways to keep the photo variant and lose the moderation problem.

Make most photo items binary, not comparative. Convert every prompt you can into a checklist. Rather than "the most creative shot involving a stranger", write "one photograph containing all six team members and at least one person not from your company, all faces visible". That is a pass or fail in two seconds, and teams still have to be creative to obtain it.

Where you genuinely want a judged prize, take it out of the scoring. Run the points hunt on binary items only, and award the creative prize separately as a named side award with no bearing on the standings. Teams accept a subjective side award cheerfully. They do not accept a subjective award that decided who won.

If judged items must carry points, publish the rubric before the start and cap their share. Three named criteria, each worth a stated number of points, judged independently, and no more than about 15 percent of total board value in judged items. A published rubric converts an argument about taste into an argument about a criterion, which is a conversation you can actually win.

Never judge alone and never judge live. Collect all photo submissions, score them after the clock stops, and have two people score judged items separately before comparing. If the two scores differ by more than one band, that item was badly written and you should tell the group so.

Two operational points specific to photo hunts.

Set the content rules in the brief, not afterwards. No photography of children who are not in the group, no photographs inside shops without staff permission, nothing that requires trespass, nothing that blocks a walkway. Say it once at the start and you will not have to say it again.

Decide the ownership question before you collect anything. Tell participants where the photographs will end up: an internal channel, a company page, nowhere. People behave differently in front of a camera when they know the answer, and asking permission after the fact from thirty colleagues is a task nobody wants.

Should teams be allowed to split up?

This is the question every hunt group asks within ninety seconds of the start, and the answer you give shapes the exercise more than any single item on your list. Treat it as a design lever with three settings.

Setting one: splitting forbidden, all members present for every item. Enforced by writing the requirement into the proof standard rather than into the rules, which is the trick that makes it work. If every photo item requires all six faces visible and every code item requires the code written on a single sheet held by the team, splitting gains nothing and the rule enforces itself. This setting maximises the shared experience and the argument quality, because every decision is made by the whole group. It also lowers the ceiling: a team of six moving as one covers far less ground, so scale your saturation ratio down or the board will feel out of reach.

Setting two: splitting permitted without limit. The natural instinct of any efficient group, and the setting that quietly destroys the point of the exercise. Six people split into six ones, each collects independently, the team total is the sum, and nobody has worked with anybody. You have run six parallel individual hunts with a shared scoreboard. If the brief was about coordination or shared decision-making, this setting fails the brief entirely, and it fails it invisibly, because the participants will report that they had a good time.

Setting three: splitting permitted with structure. The most interesting setting and the one to use by default. Permit splitting but constrain it so that splitting is itself a decision with costs.

• Sub-team minimum. No group smaller than three, so a team of six can be two threes but never six ones. Coordination survives inside each sub-team.

• Mandatory regroup points. Two fixed times, for example minute 25 and minute 50, at which the whole team must be physically together at the start point. A team that splits must now plan its return, which converts splitting from a free optimisation into a routing problem.

• Anchor items. Designate three or four items, usually in the top tier, that require the full team. These are the items around which the regroup has to be planned, and they are what stop the split from being permanent.

• Split declaration. Teams must declare their split at the start and cannot change it. This forces the allocation conversation to happen once, properly, in the planning window, rather than continuously by text message.

The reason setting three is usually right is that the decision to split, and how, becomes the richest thing the team does. Splitting genuinely is the correct move for a large board, so forbidding it makes teams do the wrong thing. Allowing it freely makes it costless and therefore uninteresting. Making it legal, structured and expensive is what produces a real conversation about capability, coverage and trust.

One practical note. If splitting is permitted, the two-minute planning window is not enough. Give five minutes. A team that must decide its own structure needs longer than a team that only decides its route.

Handling the team that plays the rules instead of the game

Open formats attract rule lawyers the way closed formats do not, because in a hunt the rules are the only surface and there is no marshal watching. Expect it, design for it, and above all decide your posture before the day rather than in the moment.

The most common exploits, in rough order of frequency.

Item wording taken literally. The item says a photograph of the team at a named landmark, so the team photographs itself in front of a poster of that landmark inside the mall. This is not cheating, it is your writing. The fix is entirely on your side: write the constraint you actually mean. If you mean the physical location, say so in the item.

Buying the answer. Paying a stranger for a code, tipping a shop assistant for a signature, ordering a required object for delivery. Address it with a single blanket rule in the brief: no item may be completed by payment to a third party, and any spending cap applies to the whole hunt rather than per item.

Remote completion. One team member never leaves the start point and completes phone-based items while the others walk. If your rules permit splitting, this is legal and you should let it stand. If your rules require presence, your proof standard should already have prevented it, and if it did not, the gap is in the proof standard.

Inter-team trading. Two teams exchange codes or agree to divide the board. This is the exploit that genuinely breaks a hunt, because it converts a five-team competition into a two-team alliance and the arithmetic collapses. Prevent it structurally: issue per-team unique codes, or vary the item wording slightly per team so that a shared answer is visibly wrong.

Deadline arbitrage. Teams work out that the late penalty is cheaper than an item and start deliberately overrunning. This one is not a problem, it is the format working correctly. A team that has computed the trade-off has done exactly what you designed the penalty to elicit. State the penalty in the brief, apply it exactly, and say so at the debrief.

Your posture, and the one rule to publish.

Do not attempt to enumerate every prohibited behaviour in the brief. You cannot, the list will be long and boring, and every gap you leave will read as permission. Give the teams one clause instead, in plain language: the organiser's decision on whether an item is complete is final, and any completion that defeats the evident purpose of an item scores zero. Then use it sparingly, publicly and with a reason stated.

Three principles for using it well.

Rule against the exploit, never against the team. Say "that item was written badly and the poster does not count for anyone, including the two other teams who did the same thing" rather than "your team cheated". The first is a correction, the second is an accusation, and only one of them survives the following Monday.

Apply rulings retroactively to everyone. If you disallow something at minute 60, disallow it for every team including those who already banked it. Uneven application is what turns a rules dispute into a grievance.

Reward the good lawyer explicitly. Some rule-reading is exactly the skill you are testing. A team that spots the set bonus, notices that the low-tier items cluster on one street, or works out that one item is impossible and skips it has read carefully and thought clearly. Name that in the debrief. The distinction you want the group to hear is between reading the design and evading it.

Finally, log every dispute and rewrite the item. A hunt run twice with the same list should have fewer arguments the second time. If it does not, you are treating symptoms.

How many people does a hunt need, and how should teams be cut?

Minimum viable group is about 12, in three teams of four. Two teams is not a hunt, it is a comparison, and with only two data points the group cannot tell whether a strategy was good or the other team was unlucky. Three teams is the point at which a strategy ranking becomes visible.

Maximum is set by your verification capacity, not by the venue. This is the constraint people miss. A hunt area can absorb hundreds of people, and published local capacities for hunt-style formats run from 10 to 500 and, in app-led versions, to several thousand. But every team you add adds claims to check. One organiser checking code and object proof can handle roughly ten teams. One organiser checking photo proof can handle about five. Above that you need either a second checker per five additional teams, or an app that scores submissions automatically, or a proof standard that removes judgement entirely.

Team size of four to six is the working band, which matches the published local range: operators publish teams of 2 to 8, with 4 to 6 the most common recommendation and 5 to 10 the widest band found. The reasoning for a hunt specifically differs from the reasoning for a station-based format. In a station format team size is capped by how many people the station accommodates. In a hunt it is capped by how many people can hold one conversation while walking. Above six, decisions get made by the two loudest members and the rest follow. Below four, splitting becomes impossible and the coordination content disappears.

How to cut the teams. Three approaches, each with a different consequence.

• Mixed by department, assigned by the organiser. The default for a corporate hunt, and correct when the brief mentions silos. It guarantees the planning conversation happens between people who do not usually plan together.

• By existing team, self-selected. Correct when the brief is about an intact team that already works together and needs to see its own decision-making from the outside. The debrief is far richer because the behaviour on display is their actual behaviour.

• Balanced for local knowledge. Specific to hunts and worth doing deliberately. In a heritage or city variant, one participant who grew up in the precinct is worth more than any other advantage on the board. Ask in advance, and spread that knowledge across teams rather than letting it concentrate.

Handle the odd numbers by uneven teams, not by spectators. Five teams of six and one team of seven is fine. One person marshalling because the numbers did not divide is not.

How long a hunt runs, and where the time actually goes

The published band for facilitated corporate hunts in Singapore is 2 to 3 hours,commonly quoted alongside per-head rates of $50 to $90. App-led public trails publish 1.5 to 2 hours, and the shortest published corporate format is around 1 hour. Bespoke designed hunts publish 2 to 4 hours end to end.

The hunting clock is not the event length, and confusing the two is the most common scheduling error. A published 3 hour hunt with a 15 minute brief, 15 minutes of team preparation and a 15 minute debrief has a hunting clock of 2 hours 15 minutes. Your saturation ratio must be computed against the clock, not the booking.

Where the time goes in a 90 minute booking:

The shortest hunt that still works is about 45 minutes of clock, with a board of around 12 items and a saturation ratio near 2.5. Below that the planning window becomes a larger fraction of the exercise than the hunt, and teams sensibly skip planning altogether, which removes the behaviour you wanted.

The longest hunt worth running is about 3 hours of clock, and only outdoors with a large area. Beyond that, fatigue rather than judgement decides the result, and in Singapore heat and afternoon rain both become material. Two practical local adjustments: schedule outdoor hunts to start before 10am or after 4pm, and write the boundary so that at least a third of the items are under cover, so that a downpour reduces the board rather than ending the event.

Add a hard buffer between the clock stopping and the results. Ten minutes of nothing, with drinks. It gives you scoring time, it lets teams tell each other what happened, and it means the results are announced to a room that has already decompressed rather than one still catching its breath.

Kit, permissions and ground to cover

What you must have.

• A printed board for every team. Paper, not a screen. Six people cannot read one phone, and the whole planning conversation depends on everyone seeing the list at once. Print at least two copies per team if splitting is permitted.

• A defined boundary, drawn and distributed. A marked map or a written description in four named streets.

• A master answer sheet for the organiser, with codes, pass conditions and point values, which nobody else sees.

• A single timepiece of record. Announce that your phone is the clock and that no other clock counts. This retires one entire category of dispute.

• A contact number on every team sheet, plus one nominated phone per team that must stay reachable.

• A scoring sheet with teams as columns and items as rows. Build it before the day. Scoring live into an unstructured sheet is how results get delayed.

What you can usually improvise or borrow. Team identifiers such as coloured lanyards or wristbands, an envelope per team for objects collected, a whistle or a horn for the start, pens, and a shared photo album or messaging group per team for submissions.

What you must actually buy or hire. Prizes, printing, the codes themselves if you are laminating or affixing anything, and any tickets if an item sits inside a paid venue. Published add-on costs that catch organisers out include venue hire at $15 to $50 per person, photography at $600 to $1,500, catering at $30 to $80 per person, transport at $20 to $40 per person and custom design fees of $500 to $2,000. If you are quoting a hunt to a budget holder, quote it with these lines visible.

Permissions, which the outdoor variants genuinely require.

• Private property is private. Malls, office lobbies and food halls are managed spaces. If your items send thirty people into a mall to photograph shopfronts, tell the management office in advance. Most will say yes and some will ask you to avoid certain hours, and either answer is better than being asked to leave mid-hunt.

• Do not place physical markers in public without checking. Affixing anything to public infrastructure is not yours to do. Use features that are already there, and put your codes on things you control or on cards held by a staffed counter.

• Religious and residential sites need explicit care, as covered in the heritage section.

• Public transport is a boundary decision. Deciding whether the MRT is inside or outside your boundary changes the effective size of your area by an order of magnitude. Decide, state it, and price the items accordingly.

The indoor version needs almost none of this, which is the main reason organisers choose it. A single building removes weather, transport, permissions, safety marshalling and the risk of losing a team, at the cost of the route choice that makes the format interesting. That trade is discussed honestly in the venue section below.

Turning the difficulty up or down

Difficulty in a hunt is not the difficulty of individual items. It is the tightness of the decision. Adjust these levers rather than making the items harder.

Easier version. Lower the saturation ratio to about 2.0 so that a strong team feels close to a meaningful share of the board. Print the list in tier order with points visible and the tier totals shown, so the arithmetic is doable without effort. Extend the planning window to five minutes. Permit splitting freely. Use codes and objects only, so no team ever has a claim rejected. State the set bonus explicitly in the brief rather than hiding it in a star mark. This version suits a mixed group, a first hunt, or an audience that will disengage from anything that feels like a test.

Standard version. Saturation ratio 3.0 to 3.2. Three tiers with the middle tier as the efficiency winner. Set bonus marked but not explained. Two-minute planning window. Splitting permitted with a minimum sub-team size and one mandatory regroup. Mixed proof types by tier. Late penalty published. This is the worked example above.

Hard version. Change exactly three things and no more.

• Hide the point values until the clock starts. Teams get the item list in the brief and the points at minute zero, which means the planning conversation has to happen on incomplete information and then be revised under time pressure. This single change is the most punishing lever available and it is also the most true to life.

• Make the board dynamic. Announce at minute 30 that one named tier has doubled in value, or that three specific items have expired. Every team must now re-plan mid-hunt, and the team that built a rigid route rather than a flexible one discovers the cost of its own commitment.

• Introduce a claimed-item lock. The first team to claim a given top-tier item removes it from the board for everyone else, announced live to all teams. Now the top tier is not just expensive, it is contested, and the decision acquires a game-theoretic layer: attempting an expensive item that another team is also attempting risks 30 minutes for nothing.

What not to do when making it harder. Do not shorten the clock without changing the board, which just converts judgement into rushing. Do not add obscure items, which converts the exercise from prioritisation into guessing. Do not add more tiers.

Where hunts fall apart on the day

The list turns out to be completable, and everyone ties. The single most common failure and the one that voids the format. Cause is always the same: the organiser estimated item times optimistically from their own knowledge of the answers. Fix is prevention. Time-test two items from your top tier yourself with a stopwatch, then multiply every estimate on the board by 1.4 to account for the fact that participants do not know the answers. If you discover mid-hunt that teams are sweeping, add a live bonus item worth a large amount at a far corner of the boundary; it restores a decision immediately.

Scoring drags and the room dies. Second most common, and entirely predictable from your proof mix. Fix by moving proof types down the tiers as described above, by receiving evidence progressively rather than at the deadline, and by building the scoring grid in advance. If you are already in trouble on the day, score the low tier by team self-declaration on the honour system with spot checks on two random claims per team. It is not perfect and it is much better than twenty minutes of silence.

Teams split into individuals and the exercise evaporates. Fix by rule, and specifically by proof standard rather than by instruction, since a proof standard enforces itself and an instruction does not. If it is happening live and your rules did not prevent it, announce a mandatory regroup at a fixed time and make one remaining item a full-team item. Both changes are legitimate mid-hunt because they apply to everybody.

A team goes quiet or goes out of bounds. Fix with the nominated phone rule, a stated boundary, and one check-in requirement partway through. The check-in is the important part: a team that must be reachable at minute 40 cannot be lost at minute 75. If a team does go dark, stop worrying about the hunt and find them; nothing else on this page matters more.

An item turns out to be broken. A shop is shut, a plaque has gone, a code has been removed. Fix by voiding the item for all teams the moment you learn of it, announcing it to everyone simultaneously, and giving any team that already banked it the points anyway. Never let a broken item stand for the teams that got lucky.

A dispute at the deadline sours the result. Fix by ruling in public, on the item rather than the team, retroactively for everyone, and quickly. Most disputes are actually about the perception that another team got away with something, which is why public rulings resolve them and private rulings do not.

One person carries the whole team and everyone else follows. Very common in hunts specifically, because the format rewards local knowledge and one person often has it. Fix by spreading local knowledge across teams during formation, by requiring a rotating navigator so the same person cannot lead every leg, and by including items whose cost axis is not knowledge but presence or interaction, which redistribute contribution.

The photo review turns into an argument about taste. Fix by making photo items binary, keeping judged items out of the standings, and publishing any rubric in advance.

Which skills the format really exercises

Be precise about this, because a hunt is often sold as building things it does not build.

It exercises prioritisation under a hard constraint, which is the format's real and distinctive contribution. Very few team activities force a group to say out loud what it is choosing not to do. A hunt forces it in the first two minutes and then punishes the group arithmetically for getting it wrong. That is a rare and genuinely transferable thing to observe.

It exercises resource allocation across people. If splitting is permitted with structure, the team must assess who is good at what, who knows the area, who will actually get a stranger to cooperate, and then commit to that assessment before it has any evidence. The commitment is the exercise. Teams that allocate by seniority rather than capability find out within twenty minutes.

It exercises decision revision. Every hunt produces the moment where the plan is visibly failing and the team must choose between finishing what it started and cutting the loss. The sunk cost is real, it is measured in minutes already spent, and the correct answer is usually to abandon. Watching who says it out loud is the most useful observation of the whole event.

What it does not build.

It does not build trust. Trust requires dependence on somebody else's honesty or competence in a situation where you are exposed. A hunt distributes tasks; it does not make anybody vulnerable. Groups that need trust work need a different format, and no amount of debriefing a scavenger hunt will produce it.

It does not build psychological safety, and can erode it. A format that publicly ranks decisions is not a place where a quiet person will risk a bad suggestion. If your brief mentions that people do not speak up in meetings, a competitive hunt is the wrong intervention.

It does not develop deep collaboration. The natural equilibrium of a hunt is parallel work, not joint work. Even with full-team rules, six people completing an item together are mostly waiting for one person to do it. If the brief is about how the team works on a problem together, run something with a single shared artefact instead.

It is a weak icebreaker for genuine strangers. The planning window demands an immediate group decision from people who have not yet learned each other's names, which usually resolves into whoever is most confident deciding alone. If the group has never met, run something short and low-stakes first and hunt second.

The 15 minutes afterwards: questions worth putting to the room

The scoring is not the debrief. Show the teams the full board, including the untouched items, and then work through these in order. Give each team two minutes on the first three among themselves before opening it up.

On the decision.

• What did you decide not to do, and at what minute did you decide it?

• Who said it first, and what did the rest of you say back?

• Did you compute anything, or did you go on instinct? If you computed, what did you compute?

• Look at the board now. Which item that you skipped should you have taken?

On the plan surviving contact.

• How long did your original plan last?

• What was the first thing that broke it?

• When you revised, did the whole team revise or did one person change course and the rest follow?

• Was there a moment you should have abandoned something and did not? What kept you going?

On the split, if splitting was permitted.

• How did you decide who went where? Be honest about whether it was capability or convenience.

• What did you lose by splitting that you did not anticipate?

• If you ran it again with the same six people, would you split the same way?

On the result.

• The winning team took roughly a third of the available points. Does that change how you read your own score?

• The gap between second and fourth was small and the gap to last was large. What separated last place from the rest?

• Which of tonight's decisions looks like a decision you make at work?

One closing question for the room, and the only one that reliably transfers: what is the item on your actual list at work that everybody is still attempting because nobody has been willing to say it is not worth the points? That question is the reason to run this format rather than another one, and it is worth protecting the full fifteen minutes to get to it.

Other names for the same format, and its close cousins

Common aliases. Treasure hunt, city hunt, urban hunt, discovery trail, heritage trail, photo hunt, photo challenge, quest, and, confusingly, amazing race. The same underlying open-list format carries at least ten different names locally, most of them branded product names invented by a single operator, which means comparing quotes by name is close to useless.

What actually distinguishes it from adjacent formats.

• Versus a route-based race. The race directs teams from station to station in a fixed order; the hunt hands over the whole list and lets teams choose. If a quote describes clues that unlock the next location, you are buying a race regardless of what the product is called.

• Versus a parallel station format such as a mini olympics or telematch. Those open every station at once and assign teams to them on a rotation, so no team chooses where to go and every team plays everything. A hunt's entire content is the choosing.

• Versus an escape room. An escape room is a single shared puzzle chain in one space with a binary outcome. A hunt is many independent items across a large space with a scored outcome. Escape rooms test collective problem-solving; hunts test allocation.

• Versus a quiz or trivia trail. If the answers are searchable from a chair, it is a quiz with walking attached, however heritage-flavoured the questions are.

• Versus a photo competition. If the standings are decided by judged quality rather than by checked completion, it is a competition, not a hunt, and it should be scheduled and debriefed as one.

What to ask a vendor, since the product names tell you nothing. These five questions separate the formats faster than any brochure.

1. How many items are on the list, and how many can a good team realistically complete in the time?

2. Are the point values uniform, or tiered? If tiered, is the top tier the best value per minute or the worst?

3. Who verifies completion, how, and how long does scoring take after the clock stops?

4. Are teams directed to locations in sequence, or given the whole list at once?

5. Is any part of the score judged rather than checked, and if so, is the rubric published to participants beforehand?

An operator who cannot answer question two is selling a flat list, which means the format will not produce triage no matter what the marketing says. An operator who cannot answer question three is going to hand you a slow finish.

Fitting a hunt into PLAYON

PLAYON does not run a scavenger hunt, and the venue cannot host the full format. That needs saying at the top rather than the bottom. A scavenger hunt lives on route choice across a large area, and this is one indoor floor. What the venue does support is a compressed variant that keeps the triage mechanic and drops the navigation, which is a real activity with real value and is not the same thing as the format described above.

What the venue genuinely contributes: the verification problem disappears.

Most of the nine attractions usable as hunt targets score themselves and display the result. That means proof of completion is a number on a screen rather than a photograph you have to adjudicate, which removes the single most expensive part of running a hunt.

Three exceptions to the automated-proof argument, stated plainly. Throwing Competition does not publish automated scoring, and the mode is chosen by a staff member, so it needs a person with a tally sheet. Air Hockey publishes no leaderboard, so its result is match-local and somebody has to record who won. Bowling publishes no round timer, being frame-paced rather than clock-paced, so a bowling target has to be defined in frames rather than in minutes. Everything else on the list posts a number without you.

How the triage mechanic survives indoors: concurrency replaces distance.

Outdoors, the cost of an item is how far you have to walk. Here, every target is a minute away, so distance carries no information. The published concurrency figures do the work instead, because they determine how long a six-person team takes to get everyone through.

• Five-Screen Basketball takes exactly five players simultaneously. A team of six needs two passes or leaves one player out.

• Air Hockey is strictly one against one. Six people through an air hockey target is three sequential matches, which makes it the most time-expensive target on the floor per unit of team.

• Bowling takes up to six across two lanes, turn-based between frames. A whole team fits, but frames are sequential within a lane.

• Archery takes up to four in rotation. A team of six splits or sits two out.

• Rubber Ball Shooting is turn-based on the station, and the number of stations is not published, so treat it as sequential.

• Colorful Grid is simultaneous, with a group of six referenced on the page and no exact maximum published.

• King of Guns is simultaneous, with the maximum not published.

• Reaction Testing concurrency depends on station configuration, with head-to-head rounds available.

• Throwing Competition is turn-based, with the maximum not published.

That gives you the cost axis you need. Price a target by how many team-minutes it consumes rather than by how hard it is, put the high-concurrency targets in the cheap tier and the strictly sequential ones in the expensive tier, and the triage arithmetic from the worked example transfers intact. A target requiring all six players through a two-player station is genuinely expensive, and a team has to decide whether the points justify it.

A compressed 60-minute indoor version for 30 people. Five teams of six, nine targets, a board of about 15 items built by setting two or three score thresholds per attraction at different point values. Saturation should sit near 2.5, which on a floor this size means the board fills up fast. Score thresholds rather than completions, so that a team can bank a low-tier target quickly and a high-tier threshold on the same attraction stays available as a separate item.

The honest limitations, in order of importance.

Route choice, which is the heart of the format, is almost entirely absent. Nine targets on one floor, all within a minute of each other, means there is no routing problem to solve. What remains is selection and sequencing, which is a genuine decision but a narrower one than an outdoor hunt offers. If the brief specifically says the exercise should involve navigation, planning a route or covering ground, this building cannot deliver it and an outdoor precinct hunt will.

Five teams and nine targets means contention. In an outdoor hunt teams rarely meet, because the area is large. Here, two teams wanting the same attraction at the same time is not an edge case, it is the normal state. You have two options and you must choose one before the start: allow queueing and accept that waiting time becomes part of the cost calculation, which is realistic but frustrating, or impose a one-team-per-attraction lock and make target availability itself the scarce resource, which is more interesting and turns the hunt into a live contest for the board. The lock is usually the better choice, and it is the closest indoor equivalent to the claimed-item lock in the hard version above.

Nothing aggregates the attraction scores into a team total. Each board shows only the players who used that machine. No venue-level total exists, and no normalisation across attractions exists either. One nominated scorer has to carry the numbers into a sheet. It is much less work than checking photographs, and it is not nothing.

The heritage variant is impossible here and the photo variant is thin. If the brief asks for cultural discovery, a sense of place or learning something about Singapore, this is the wrong building and no amount of design will change that. Take that brief to a heritage precinct.

Attire and age rules need to be in the calendar invitation. Colorful Grid and Reaction Testing both require grip socks and prohibit bare feet. Five-Screen Basketball, Rubber Ball Shooting and Reaction Testing require rings, watches, bracelets, bags and drinks to be removed. Archery, Air Hockey, Throwing Competition, Colorful Grid, Five-Screen Basketball and Reaction Testing all state that under-8s or anyone under 120cm needs an accompanying adult, which matters for a family day rather than a corporate group but should be checked if the event mixes the two. Where minimum heights are signposted at a zone entrance rather than published, confirm the figure before you build the board around that attraction.

What this venue is right for. A corporate group that wants the triage decision, an automated scoreboard, no weather risk, no permissions to obtain, no props to hire and a hard 60 to 75 minute window, and that is honest with itself that the walking half of the format is not on offer. If the brief instead says the team should be out in the city making routing decisions across a heritage precinct for two hours, book that, and treat the indoor version as the wet-weather fallback it is well suited to being.

Questions HR asks about scavenger hunts

How much does a scavenger hunt cost per person in Singapore? Hunt-style formats are published locally between $25 and $130 a head. A commonly published band for a facilitated two to three hour hunt is $50 to $90. App-led public trails publish around $30. Minimum event fees from $2,000 are common below thirty participants.

How is a scavenger hunt different from an Amazing Race? Order. A race gives teams one clue at a time and directs them from station to station in a fixed sequence. A hunt hands over the whole list at once and lets teams choose what to attempt and in what order.

How many items should be on the list? Between 18 and 30 for a self-run hunt. Local operators publish 10 to 20 checkpoints for facilitated formats and 30 or more challenges on app-led ones. Below 18 the choices are too few, and above 30 teams stop reading the back half.

How do I stop teams from completing everything on the list? Make the total collection time about three times the hunting clock. Add up your honest completion time for every item, divide by the minutes available, and target a ratio between 2.5 and 3.5. Below 1.5 every team sweeps the board and ties.

Should the biggest items be worth the most points per minute? No, they should be worth the least. If the top tier is also the most efficient, sweeping it is the only sensible strategy and there is no decision left. Make the middle tier the best value so the largest printed number is a deliberate trap.

What proof of completion should I ask for? Codes and objects wherever possible, because both verify in seconds. Reserve photographs for a middle tier and video only for a handful of top-tier items. Verification time is the hidden cost of a hunt, and photo-heavy boards routinely double the scoring window.

Should teams be allowed to split up? Allow it with structure rather than banning it. Set a minimum sub-team size of three, fix one or two mandatory regroup times, and designate a few items that require the whole team present. Unrestricted splitting turns one team into several solo hunts.

How long should a scavenger hunt run? Two to three hours is the published local band for facilitated hunts. What matters is the hunting clock alone. A three hour booking with briefing, planning and debriefing leaves roughly two hours and fifteen minutes of actual hunting.

What team size works best? Four to six people. Published local recommendations converge on 4 to 6, with 2 to 8 the widest band found. Above six, decisions get made by the two loudest members. Below four, splitting is impossible and the coordination content of the format disappears.

What makes a heritage clue good rather than trivia? The answer must live only on the object. Ask teams to count a feature, read the smallest line on a plaque, compare two adjacent buildings, or identify what changed since an old photograph. Anything answerable from a phone in a coffee shop is a quiz.

How do I judge a photo hunt fairly? Make photo items binary rather than comparative. Write a checklist condition such as all six members plus one stranger with faces visible, and keep any judged creative prize out of the standings as a named side award. Give teams any rubric before the start.

What do I do about a team that games the rules? Give every team one clause saying any completion that defeats the evident purpose of an item scores zero, then rule against the exploit rather than the team, apply the ruling retroactively to everyone, and announce it publicly. Log the dispute and rewrite the item.

Does a scavenger hunt build trust? No. It builds prioritisation under time pressure and forces a team to say out loud what it will not do. Trust requires dependence on somebody else's honesty in a situation where you are exposed, and a scored hunt supplies neither condition.

What is the smallest group a hunt works with? Twelve people in three teams of four. Two teams is a comparison rather than a competition, because the group cannot tell whether a strategy was good or the other team simply got lucky. Three teams is where a strategy ranking becomes visible.

Can a scavenger hunt run entirely indoors? Yes, in compressed form. A single venue removes weather, permissions and transport risk, and self-scoring attractions remove most verification work. The trade-off is that route choice, which is the core of the format, largely disappears, leaving selection and sequencing.