iGaming / Knowledge atlas
iGaming Knowledge Atlas: 11 Working Methods
Eleven iGaming working methods on one map: promotion rules, provider offers, player journeys, reporting and evidence, plus seven decision questions.
Before a promise becomes a problem, find the gaps.
iGaming / Knowledge atlas
Eleven working methods from operator and provider work. The offers are history. The method still works.
New to iGaming? Start here.
Provider (The game studio): Makes the games, and hands out promotion offers to push them.
Merchant / operator (The gaming site): Runs the site, takes deposits, and gives players promotions.
Player (The customer): Deposits, plays, claims rewards, and asks support when stuck.
In the middle. I turn provider offers into the few a merchant can actually run, make sure what players are promised is written clearly and priced, and check the results afterwards.
Promotion: An offer to players: bonus, rebate, free spins, lucky draw.
Turnover: Total amount bet. Bet 100 ten times and turnover is 1,000.
Wagering requirement: How much must be bet before a bonus can be withdrawn.
GGR: Gross gaming revenue: bets minus winnings paid. The site's gross take.
Rebate: A percentage given back to players. What it is a percentage of decides the cost.
FTD: First-time deposit: a player depositing for the first time.
Funnel: The steps from seeing an ad to depositing. People drop out at every step.
Channel: Where players come from: ads, SMS, messaging apps, email.
Segment: A group of players sorted by behaviour or market.
Back office: The admin system where promotions are set up.
CS: Customer support.
Placement: A spot on the site, like the homepage banner, that offers compete for.
Eleven methods, sorted two ways.
Rules & cost (01, 09): What exactly are we promising players, and can we afford it? Every promotion is a small contract and a budget line. If the languages say different things, or nobody priced the prize, it comes back later as complaints or losses.
Partners & delivery (02, 11, 07): Which provider offers do we take, and can the team actually ship them? Providers offer a lot. Saying yes is easy; delivering needs placement, setup, support, and an owner for every step.
Players & service (03, 05): Who should we contact, and what does a stuck player need to hear? Not every player should get a message. When a reward is late, the page, the system and support must tell the same story.
Data & evidence (04, 06, 08, 10): Does this number really show what people say it shows? Report names and formulas often disagree. Before believing "the campaign worked", check what was counted, at which level, and whether it ever went live.
01 · One offer, every language, the same rule
A promotion is not a poster. The copy players read, the back-office setup, the support script, the reward calculation and the payout must all say the same thing, in every language.
Think of two coupons: one says "ten purchases in your birthday month", the other "ten purchases in the past year". That is not a translation style; it decides who qualifies. Drafts in three languages can each read smoothly and still disagree on eligibility, game category or percentage.
• Line the languages up field by field: identity and region → start and end → eligibility and exclusions → minimum, cap and calculation base → products and wallets → wagering → claim, verification and payout → expiry and disputes.
• Any difference is a new version with a number, not a quiet edit to one language.
• "The English version prevails" settles disputes; it does not stop players reading their own language. Fix the text anyway.
Built from historical drafts. It shows the drafts disagreed, not that a live page was wrong.
02 · "The provider supports it" is only step one
A feature existing is not the same as it being usable. Ask separately: does the current integration support it, is this merchant entitled, which placement or time slot does it need, can the commercial terms all hold at once, who approves, and who delivers after launch.
Picture a shopping mall window: two brands both want the single main entrance spot. Each deal is attractive on its own; you still cannot promise both.
• Keep one condition sheet per offer: provider and product, mechanic, technical dependency, merchant eligibility, placement, cost, validity, approver, delivery status.
• A blank field is "unknown", never "allowed by default".
• Two offers that need the same exclusive slot are decided together, before anyone promises either.
The allocation tool behind this card was a prototype; it did not cover every contract.
03 · Define the player state before you message anyone
Registered, first deposit, active, gone quiet, asking for help: these are different events. "Do not contact" is a valid branch, not a failure.
Journey maps that put the observation window, email events, support follow-up and rewards on one page turn what each team privately knows into states everyone can discuss.
• Trigger event and data source.
• Observation window and the condition that allows contact.
• What help the player actually needs.
• The condition that stops contact, the owner, and the next review date.
Older flows that kept adding offers based on wins and losses are kept only as reflection material, never as a method.
04 · A report label does not guarantee the formula
Before reading a number, trace its numerator, denominator, currency, time window, sign and raw data range. A column called ROI may not be a profit return at all.
Registrations, first deposits, activity, deposits, stakes, rewards and channel cost often sit on one sheet. Ratios get familiar names; month-on-month comparisons quietly include an unfinished month or a second currency.
• Metric card: the question it answers, formula, unit and currency, period, de-duplication rule, data owner, and the usual misreading.
• People are not events, deposits are not revenue, clicks are not first deposits.
• Without a control or attribution evidence, do not write "the campaign drove growth".
A method card. It certifies no historical profit or growth figure.
05 · Wallets, rewards and support: "where is it now?"
For the player a campaign is not finished when the back office is configured. They need to know which wallet holds the money, when the reward lands, why they cannot withdraw yet, and who fixes an error.
Page plans split registration, deposit, transfer, withdrawal, bet history and reward status into screens; the support Q&A lists the same moments as complaints. Read together, a marketing promise turns into a chain of player actions and explanations.
• Every important state answers: where am I, what happens next, how long, who can help.
• Page text, system status and the support script must match. Internal terms are not answers.
• Support scripts state facts the player can check, never unprovable reassurance such as "we are the most trusted".
Shows the requirements existed; it does not prove every screen shipped or lifted conversion.
06 · Channels and content have levels of evidence
Reach, click, registration, first deposit and long-term activity are different levels. Content is the same: "we tested this" must answer how, when, and where the result is.
Weekly reports mixed paid ads, SMS, messaging apps, email and calls. One page logged bounced email; another separated clicks from registrations. Clean the data before you compare channels.
• Name the level a number belongs to before comparing two channels.
• First-person review claims need a test record behind them.
• Old channel efficiency is not today's channel efficiency.
No current keyword or channel research is included here.
07 · Cross-team delivery: every hand-off must be checkable
From idea to something a player can claim, content, calculation, design, back office, support, payments and the public page hand over in order. Each hand-off needs a verifiable input and a done state.
A monthly routine ran research → calculation → copy → design → upload → back-office setup, with support, payments, third parties or logistics depending on the reward. Exhibition checklists put materials, site, video, owner and deadline on one page.
• Approved rule version and artwork version.
• Back-office setup, test account and test result.
• Support explanation, payout owner, exception handling, end or take-down time.
• One field unconfirmed means not launched. No "close enough".
Checklists show how work was organised, not commercial results.
08 · A prototype is not a result
Plan, working prototype, approved rule, live record and checked result are five different kinds of evidence. A prototype is valuable because it exposes conflicts before money is spent; it is still not an outcome.
A prototype can show cost and reward maths, sharing and a mini-game while its starting balance and thresholds disagree and its back-office binding is still a TODO. A landing brief asked for a participant counter that always went up. Both are lessons in stopping unsupported numbers before launch.
• Evidence ladder: idea → prototype → approved rule → live record → checked result.
• Without a live record, call it a design or a prototype.
• A number without a source never becomes social proof.
Only the last rung allows careful talk of impact, and even then correlation is not causation.
09 · Wheels and red packets: price the expected cost first
A spin-the-wheel or lucky-draw campaign is a probability table, not a fun banner. Before launch, know the expected cost of one entry, whether the top prize will really be drawn, and whether the worst case breaks the budget.
Like a scratch card: printing "phone" on one prize does not mean stocking one phone, but at 0.5% with a thousand players a day, that is five phones a day.
• Probabilities add up to 100%, including "try again".
• Expected cost = Σ(prize × probability) × expected entries.
• Cap the top prize or accept that it will be won; never advertise a prize nobody can draw.
• When stage odds change, the player-facing rules change with them.
A cost model only. It is not advice on making people deposit more, and draw promotions are regulated differently by market.
10 · The structure of a brand health check
Diagnosing a brand runs in one order: current data → segments → operations experience → ranked recommendations → timeline and team KPIs. Every page answers one decision.
A fifty-page review looked at two markets separately: registration versus first deposit, unique depositors, deposits and withdrawals, game-type preference and contribution, spending power, segment call lists, promotion results, and processing times, then proposed channels, site changes, promotions and KPIs per team.
• Current state: where does the funnel drop most?
• Segments: which markets, products and players carry the value?
• Experience: deposit and withdrawal times, top support questions.
• Five to eight recommendations ranked by impact and cost, each with an owner and a KPI.
"Healthy" thresholds in old reports had no source and are not industry benchmarks. Tactics aimed at players who are losing are blocked, not reused.
11 · Six offers in, two offers out
When a provider hands over six promotion offers, pass two to the merchant. Too many options and the merchant stalls, or says yes to everything and executes none of it well.
A menu with sixty dishes freezes the table; a waiter who recommends two signatures gets an order. The person curating does half the merchant's thinking for them.
• Fit: the merchant's players, market and currency.
• Feasible: placement, integration and staff the merchant actually has.
• Compatible: the two do not compete for the same slot, period or players.
• Measurable: a clear metric and data source afterwards. Two, not one, so the merchant decides by comparing.
Eden's own practice, stated first-hand; acceptance rates were not recorded.
Three numbers that change the decision.
The rarest prize is the budget. A five-slot wheel. The phone lands on 0.5% of spins and eats 82% of the cost.
Same 10%, 25× the cost. One month, one player group. "10% back" depends on what the 10% is of.
More clicks, fewer first deposits. Same month, same budget. A gets 3× the clicks; B brings 2× the first deposits.
Illustrative numbers for teaching. Not historical data.
Six years in the files.
2020 · Campaign rules in many languages: Campaign collections, multilingual terms and support handling: rich detail, and visible version drift.
2021 · Agency side: Serving operator brands as an agency: rebrands, three-language terms and support scripts, reward probability tables, a full brand health check.
2022 · Running operations: CRM and page flows, monthly routines, weekly reports: acquisition, pages, support, back office and retention joined into one system.
2024 · Reporting: Monthly reporting and the dimensions behind it.
2025 · Provider-side B2B: Provider conditions, merchant offers, promotion prototypes and exhibitions: surfacing conflicts before anyone promises. Six offers in, two out.
2026 · From files to a system: Thousands of old working files reduced to methods, a handbook and consulting work.
Seven questions before any decision.
Can this offer be promised to players? Who can join, in which region and wallet, from when to when, how it is calculated, the cap and exclusions, who pays out and who handles disputes. Compare every language line by line. Outcomes are only three: consistent and approved, go to testing; fields missing, return to the owner; contradictions, pause and name one rule owner. Never guess the approver's answer.
Is the provider feature really usable? Does it exist, does the current integration support it, is the merchant entitled, is the placement free, what are cost and validity, who approves? Separate "technically possible" from "commercially promisable". Any unknown stays "to confirm" and nothing is promised.
What is the next step in this player's journey? Which verifiable state is the player in, what is the reason to contact, how long is the window, when do we stop, who handles service issues? If the data source, contact condition or stop condition is unclear, do not automate the message.
Does this report support the conclusion? What question does the metric answer; are numerator, denominator, de-duplication, currency, window, cost and refunds consistent? Discuss change only when the definition is traceable; otherwise fix the definition first.
Can this page or prototype count as a result? Idea, working prototype, approved rule, live, or checked result? Each level needs its own evidence. Contradictory numbers inside a prototype are a finding, not a win.
How do I hand this decision to the next person? Nine fields: the question, evidence visible then, what was missing, options, trade-off, approver, version adopted, verification result, and the date or condition for review. Keep my recommendation separate from the team's decision. Unknown is written as "unknown".
When do I stop digging through old material? When an old draft has no reliable end date, label it "historical terms, current status unverified" and stop. Go back to the original only to publish, review a result or reuse a clause.
Where the methods become work.
Promotion terms audit (3–5 days per 10 campaigns): Multilingual terms disagree, support cannot explain them, players complain. Clause-by-clause comparison: eligibility, amounts, products, wagering, dates, exclusions; Conflict list and proposed fixes.
Reporting definitions review (1–2 weeks): Teams' numbers never match; ROI and retention mean different things to different people. Metric cards: numerator, denominator, currency, window, de-duplication; A report structure the next person can use.
Provider offer triage (1–2 weeks): Too many provider incentives, clashing placements, promises nobody can keep. Condition sheet and resource-conflict map; A pre-commitment checklist and the two offers worth taking.
Brand operations health check (2–4 weeks): A new brand or a relaunch, and no agreement on what to fix first. Funnel diagnosis and segments; Deposit, withdrawal and support experience; Ranked recommendations with team KPIs.
What I do not take on: tactics aimed at players who are losing, fabricated social proof, legal or licensing opinions, or raw player data. Former employers, brands and providers are never named.