Advisor guide · Platform Power
How Platform Power works
Platform Power turns what you learn about a client’s platform into scores, ratings, and a modernisation plan a board can trust. Your judgement is the input, and the framework’s job is to make that judgement consistent, comparable, and defensible. This guide covers the framework first, then the studio itself, from a first prospect to the commission on a finished engagement.
The problem it solves
Why Platform Power exists
One question sits under every brokerage and fintech engagement: can this platform create value and hold onto it? Answering it well means looking at three things at once. There are the economics, which say whether the business makes money. There is the strategic position, which says what stops a rival taking that business away. And there is the technology, which says whether the plumbing is an asset or a liability waiting to surface. Most reviews look hard at one of the three and wave at the other two.
Platform Power scores all three on one comparable scale, works from graded evidence, and prices the gaps without pretending to more precision than it has. It does not replace your judgement. It takes your judgement and makes it consistent, so that two advisors reach the same score from the same facts; comparable, so that this platform can be set against its peers and against its own position last year; and defensible, so that it survives the technical due diligence a board or an acquirer will put it through.
Provenance
Where the framework comes from
The strategy lens was not invented here. Power is Hamilton Helmer’s 7 Powers, the modern canon on durable competitive advantage, and it is used as he wrote it: seven structural sources of power, each of which is real only when a genuine benefit is protected by a barrier that a competitor cannot cheaply cross.
Infrastructure comes from the infrastructure deep-dive lineage, which is a structured technology assessment of nine modules and fifty-one subcomponents running from the front end through to liquidity. Its job is to turn the question of whether the plumbing is an asset or a constraint into evidence-graded ratings rather than opinion. Business is the hard economic register: assets under administration, revenue, unit economics, and growth. Bringing strategy, technology and economics under one graded, uncertainty-aware method is the whole idea. Nothing is scored on instinct, and nothing is priced by dividing a score gap into pounds.
The pipeline
How it works, end to end
- Gather evidence and grade it
You learn how the platform works, from documents, interviews, dashboards, or your own inspection, and you record how sure you are on a four-point evidence scale. That grade is not paperwork. It is what decides how wide the ranges on the final scores are.
- Rate each element against a rubric anchor
Every subcomponent gets a maturity level against a written anchor that describes what that level looks like in practice. Every power gets a benefit and a barrier. Every business metric gets a value read against its normalisation curve. You make the judgement, and the anchor is what keeps two advisors reaching the same answer from the same facts.
- The engine computes the three lens scores
Business, Power and Infrastructure are each computed as an evidence-weighted blend in which the weakest critical part caps the result, so a module cannot score above its own bottleneck. Nothing is guessed or filled in with a zero. An element you have not assessed stays marked as unassessed and widens the uncertainty instead.
- The engine models the uncertainty as a range
Rather than report a single figure that looks more precise than the evidence supports, the engine re-scores the assessment many times over, moving each input within the confidence its evidence grade allows, and reports the spread. Thin evidence produces a wide range, which is the honest result.
- Rules turn the scores into the words a board reads
The rating a client sees for each module, from Basic to Frontier, comes from rules rather than from rounding the score. A module cannot be rated Advanced while a critical part of it is Basic, however high the arithmetic runs. The rating is what you defend in the room, and the score underneath is what orders your fix list.
- The value bridge prices the gaps
Each upgrade is costed in pounds, the cash-flow levers it moves are valued as a net present value on the client's own baselines, and its strategic worth is stated in words. The three are never collapsed into a single figure, because dividing a score gap into money is the claim that fails technical due diligence fastest.
The shape of it
Three lenses, one headline
An assessment looks at a platform through three lenses and reports a composite headline over them. Each lens has a letter, which is the short form you will see on the wizard and in reports once the full name has been given.
- Business
What does this platform achieve economically? Assets under administration, revenue, margins, growth, and the cost of acquiring a customer, taken as hard numbers.
- Power
What stops a competitor taking the business away? Hamilton Helmer's seven Powers, each scored on the benefit it creates and on the barrier that protects it.
- Infrastructure, the technology Layer
Is the plumbing an asset or a constraint? Nine modules and fifty-one subcomponents, from the front end a customer touches through to liquidity.
- Platform Value
The composite headline that blends the other three. The figure a client tends to remember, though, is the bottleneck: the weakest critical link, which strength elsewhere never fully hides.
Sitting on top of the lenses, the client also sees the Platform Power triad: Economic, Perceived, and Defence value. Each is reported as one of four words, from None through Emerging and Established to Wide, and never as a decimal.
Reading the lenses
What the letters mean
A platform, in this framework, is the whole operating system of a brokerage or fintech: its economics, its strategic position, and the technology that runs it. Platform Power is how much durable value that whole creates, which is why the headline figure is called Platform Value. The letters stop being arbitrary once you read them together.
- B, for Business
- The economic reality. Does the platform make money, and is that improving? Assets under administration, revenue, unit economics and growth, all normalised so that a £2bn platform and a £50m one can be compared fairly.
- P, for Power
- Taken straight from Helmer’s seven Powers. A power counts only where a real benefit is protected by a barrier that a rival cannot cheaply cross, and the engine scores it at whichever of the two is weaker.
- L, for the technology Layer
- The infrastructure layer that sits under the business, covering nine modules from the front end through to liquidity. The question it answers is whether the plumbing is an asset that protects the business or a constraint that will surface as a problem later.
- V, for Platform Value
- The composite headline, blending the other three. The figure a client actually remembers is usually the bottleneck instead: the weakest critical link, which no amount of strength elsewhere fully hides.
The Platform Power triad re-reads the same evidence as three plain words a board understands immediately. Economic value asks whether the platform is worth money. Perceived value asks whether customers feel it. And Defence value asks whether it can be protected.
Rating infrastructure
The four maturity levels
Every infrastructure subcomponent is rated at one of four levels, against a written anchor that describes what that level looks like in a real firm.
| Basic | It barely exists. Manual, unreliable, or absent, and the people who depend on it feel the pain regularly. |
|---|---|
| Developing | It exists, but you would not trust it under pressure. There are gaps, workarounds, and single points of failure. |
| Advanced | It reliably does its job. It is automated, monitored, and documented, and there is still room to improve it. |
| Frontier | It is a competitive weapon rather than merely adequate. Not every firm needs this, so it should never be presented as the universal target. |
Two further states sit outside the ladder, and both are honest answers. Not Applicable means the subcomponent is out of scope for this firm, and it drops out of the arithmetic entirely, with the remaining weights rebalanced around it. Not Assessed means it is in scope but you have no evidence yet. It is never scored as a zero, though it does cap what the module can reach and it widens the uncertainty range. Marking something Not Assessed is professionalism rather than failure.
How sure are you?
Evidence grades drive the ranges
Every rating carries a second mark for how you know. The four grades climb from the weakest form of evidence to the strongest, starting with what a client said, then what you probed in an interview, then an artifact you were shown, and finally something you observed yourself. Each step is a step from someone having claimed it toward you having seen it.
- E1Client-said
The client told you, and that is all you have. It is an unverified claim.
What counts: A statement made in a meeting or written on a form, such as “our uptime is 99.9%”, with nothing behind it yet.
- E2Interview
You probed the claim in a structured interview with the person who owns the thing.
What counts: You asked how it works, since when, who runs it, and what breaks, and the answers held together under follow-up.
- E3Artifact
You saw the thing itself, in the form of a document, dashboard, configuration, runbook, or metric.
What counts: A screenshot of the monitor, the incident log, the architecture diagram, or the actual number.
- E4Observed
You watched it work, or inspected it yourself. This is the strongest evidence available.
What counts: You saw a deployment run, watched a failover, or drove the system and confirmed it behaves the way it was described.
The grade is not administrative overhead, because it drives the output directly. Ratings carried on E1 evidence produce wide ranges, and E4 ratings produce tight ones. A higher grade counts for more because it is closer to the thing itself: a claim can be wrong and a document can be out of date, but something you watched work is hard to argue with. In practice the difference between a £25k assessment and a £75k one is largely the evidence grade you manage to reach.
Scoring the seven Powers
Benefit and barrier, with the weaker side winning
For each power you record two things. The benefit asks whether there is a real economic advantage here. The barrier asks what stops a competitor copying it. The power is then scored at whichever of the two is weaker. A brilliant benefit with no barrier scores None, because competitors will simply copy it, and this is the single most common thing new advisors get wrong.
Take a broker whose onboarding converts twice as well as anyone else’s. The benefit is plainly real and shows up in the acquisition cost. Ask the barrier question, though, and the answer is that a competitor could hire the same design agency and match it inside two quarters. The power scores None, and what you have found is a head start worth defending rather than a moat worth paying for.
Every power is always scored, and Not Applicable does not exist for powers. A power that is irrelevant to this business is simply weak for this business, and that is itself information worth reporting.
Helmer’s seven
The seven Powers, one by one
Each power has a distinct benefit, meaning the advantage the leader enjoys, and a distinct barrier, meaning the reason a rival cannot copy it. Each also tends to arise at a particular stage of a business’s life. Origination is where the model is still being formed, Take-off is rapid growth, and Stability is maturity. Knowing which stage a platform is in tells you which powers are even available to it.
- Scale EconomiesTake-off
- Benefit
- Unit costs fall as volume grows — the leader can price where a sub-scale rival loses money.
- Barrier
- A challenger would have to buy share at a loss to reach the volume that makes the economics work.
- Example
- A broker spreading fixed clearing, market-data and compliance cost across millions of accounts. Zero-commission economics only close at scale — a small entrant can't match the price without burning capital.
- Network EconomiesTake-off
- Benefit
- Each new user makes the platform more valuable to every other user; the leader's value compounds.
- Barrier
- A rival must reproduce the whole installed base at once to match the value proposition — not just the product.
- Example
- A social- or copy-trading platform: more traders means richer leaderboards and more strategies to copy, which pulls in more traders. A clone with an empty network offers little.
- Counter-PositioningOrigination
- Benefit
- A newer, superior business model the incumbent can't adopt without hurting its existing business.
- Barrier
- The incumbent's own economics deter imitation — copying you would cannibalise their core revenue.
- Example
- A commission-free entrant monetising order flow. A commission-based incumbent can't simply match it without vaporising its per-trade revenue line.
- Switching CostsTake-off
- Benefit
- The value a customer forgoes by leaving — deeper the lock-in, the more pricing power the leader holds.
- Barrier
- Rivals must compensate the customer for the pain of switching, eroding the rival's own economics.
- Example
- Transferring an ISA or portfolio in specie is slow and fiddly; unrealised CGT positions, recurring investments and linked accounts all raise the cost of moving away.
- BrandingStability
- Benefit
- A durable attribution of higher value (trust, safety, prestige) to an otherwise comparable offering.
- Barrier
- Brand is slow, costly and uncertain to replicate — built over years, not bought.
- Example
- A decades-old, trusted name commands AUM and trust a new app must earn slowly — most acute where investors are choosing who to hand their life savings to.
- Cornered ResourceOrigination
- Benefit
- Preferential access to a coveted asset — talent, IP, a deposit base, a scarce licence — on attractive terms.
- Barrier
- By definition the resource isn't available to rivals; they can't obtain it at any reasonable price.
- Example
- An exclusive market-maker relationship, a scarce licence in a hard-to-enter jurisdiction, or proprietary routing that rivals simply cannot obtain.
- Process PowerStability
- Benefit
- Embedded organisation and activity sets that yield superior product or cost, refined over time.
- Barrier
- Hysteresis and opacity — rivals can only match it through an extended, uncertain commitment.
- Example
- A deeply optimised in-house order-management and risk stack built over years. A rival can't buy equivalent execution quality and cost off the shelf.
The rule from the previous section applies to every one of these. The score is whichever of benefit and barrier is weaker, so a textbook benefit sitting behind a barrier a rival can cross in a quarter is not a power at all.
Reading the outputs
Reading the outputs
What a modelled range is
Every input you enter carries an evidence grade that says how well supported it is. To turn that into an honest picture, the tool re-samples each input within the confidence its grade allows and re-scores the whole assessment many times over. The result is not one number but a spread of possible scores. That spread is the modelled range, and it is how the assessment shows its own confidence: strong, well-graded evidence produces a narrow spread, and thin or weakly graded evidence produces a wide one.
P10, P50, and P90, in plain words
From that spread of re-scored results, three points are reported. P50 is the median: the value that half of the re-scored results fall below and half fall above, so it is the middle of the spread. P10 is the value that ten percent of the results fall below, and P90 the value that ninety percent fall below. Together, P10 to P90 is the likely range: the assessment is fairly confident the true score sits between them. A wider P10 to P90 gap means less certainty, because the evidence behind it was thinner or more weakly graded.
The headline is one number, and it never moves
The headline always quotes the deterministic score: the single value the engine computes directly from your inputs, the same value a finalised assessment locks in. The range from the re-scoring is shown around it as the honesty band, so you can see how much the evidence lets the number move. The headline and the range never disagree, because the headline is the point and the range is the context around that same point. When an input has no evidence grade, the score cannot be re-sampled, so it is shown as a labelled point rather than a falsely narrow range. Quote the range to a technical audience, because the single number on its own overstates how precise the assessment is.
What the rating word is for, and what the score is for
Each module carries a rating in words, from Basic to Frontier, and that word comes from rules rather than from rounding the score. The word is what a client understands and what you defend in a meeting. The continuous score underneath it, which is more precise, is what decides which weakness to fix first. So a module can carry a respectable-looking number and still be rated Developing, because a critical part of it is weak. That is the design rather than a fault. Use the word to communicate and the score to decide the order of the work.
The bottleneck sets the score
The weakest critical element caps the whole. A module cannot be rated Advanced if a critical part of it is Basic, however strong everything else is, in the same way that a chain is only as strong as its weakest link. This is why the headline word can sit below what a plain average would suggest, and why the fix that moves the score most is usually the bottleneck rather than the part that is already strong.
The value bridge: how a gap is priced
Money is kept in three separate layers that never collapse into one number: the cost to upgrade, in pounds; the cash-flow levers it moves, each expressed as a net present value on the client’s own baselines, in pounds; and the strategic value, stated in words. The Upgrade Priority Index says what to do first, and the value bridge says what it is worth. Platform Power never says that a score gap is worth a given number of pounds, because dividing a score directly into money is the one move that fails technical due diligence.
What gets an assessment rejected
Mistakes to avoid, and why they matter
- Scoring from memory instead of reading the rubric anchor.The anchor is what makes two advisors agree. Without it you are recording an impression, and an impression does not survive challenge.
- Presenting E1 evidence as though it were certain.A wide range is a perfectly respectable result. False precision is not, and it is the thing a technical reviewer tests first.
- Guessing at a rating to avoid marking something as not assessed.Everything else here is built to survive scrutiny, and a guess quietly poisons that. It is the one input the framework cannot recover from.
- Scoring a power on its benefit while leaving the barrier unexamined.An advantage a rival can copy next quarter is a head start, not a power, and the score is meant to tell those two apart.
- Presenting Frontier as the target every module should reach.Frontier means the capability is a competitive weapon. Most firms should be at Advanced in most modules and should spend the money elsewhere.
- Quoting a point score without the range around it.The point on its own claims more precision than the evidence supports, and it hides the very thing that tells a client how much to trust the number.
- Converting a score gap directly into pounds.Score points and currency are kept in separate equations on purpose. A sentence such as “these twelve points are worth £2m” fails due diligence on the spot.
The maths, briefly
Scoring, explained in full
Platform Value is a weighted average of the three lens scores. Written out, that is V = θ_B · B + θ_P · P + θ_L · L, where each θ (the Greek letter theta) is the weight given to one lens. The three weights add up to one, and they are not the same for every kind of firm, because what drives value differs by segment. These are the weights the engine uses today.
| Operating model | Business (θ_B) | Powers (θ_P) | Infrastructure (θ_L) |
|---|---|---|---|
| Retail brokerage (draft) | 0.30 | 0.30 | 0.40 |
| Wealth advisory (starter) | 0.45 | 0.30 | 0.25 |
| Exchange / infrastructure (starter) | 0.30 | 0.37 | 0.33 |
In wealth advisory the franchise economics lead, so Business carries the most weight and Infrastructure is trimmed, because there it is largely hygiene whose cost is already priced into the Business figures. For an exchange the moat matters most, so Powers is the top term. The retail weights are a uniform draft and are still awaiting elicitation, which is why they are labelled as draft on screen. The Customer Proposition index is reported alongside Platform Value rather than folded into it.
The full and reviewable account of the maths, covering the composite, module maturity, the uncertainty model, and how Hamilton Helmer’s framework maps onto it, is in docs/ATLAS-Scoring-Explained.md.
When a client puts a CTO or a diligence team across the table, the document to hand them is the technical white paper, docs/ATLAS-White-Paper-v1.md. It carries the formal model, the validation evidence, and a limitations register that states what is not yet proven — written expecting the other side to check it.
Working the app
From a first prospect to a paid engagement
Everything above is the framework. This second half is the practical walk through the studio itself, from a first prospect to a finalised assessment, a client deliverable, and the commission that follows.
Getting started
Sign in and find your way around
Access is by invitation. Once you are in, the dashboard is home, and each section on it carries a one-line description of what it is for.
- Sign in with the email address you were invited on.The Bruntsfield mark in the top left always returns you to the dashboard, so you cannot get lost.
- If the framework is new to you, read the first half of this guide before you open an assessment.It covers the three lenses, the four maturity levels, and the evidence grades, which is everything the wizard assumes you already know.
- Then pick a section: Pipeline, Assessments, Engagements, Workbench, or Earnings.
Pipeline
Work your pipeline
Your board holds ten stages, from first contact through to a contracted engagement, with a weighted forecast across them. The forecast is expressed as a count of deals rather than in pounds, so that an early conversation is never mistaken for booked revenue.
- Add a prospect by typing a company name and pressing Add prospect.It lands in the first stage.
- Move a card forward as the relationship progresses. The board decides which moves are legal and reverts one that is not.A card that has sat in one stage too long is flagged as stale, so nothing goes quiet on you.
- Read the forecast strip: prospects, open deals, and expected won deals, with a win probability for each stage.An empty stage reads as empty on purpose. It is not a loading glitch.
Assessment
Run a Platform Power assessment
The seven-step wizard scores a company across the seven Powers, Platform Value, and the nine infrastructure modules. It saves every edit as you make it, so a partial assessment is always safe to leave and resume.
- Start an assessment for the subject company, then work the steps from left to right: Business Metrics, Powers, then the Infrastructure Deep Dive.Your work saves on its own, and the badge moves from Saving to All changes saved.
- Rate what you know. For each subcomponent pick a maturity level, then record how sure you are with an evidence grade, from E1 for something you were simply told up to E4 for something you watched work.That grade is what widens or tightens the uncertainty carried on the score.
- Leave anything you have not looked at marked as Not Assessed, and never guess.Not Assessed is a proper state rather than a gap. It lowers your coverage and widens the range, but it is never counted as a zero.
- When the assessment is ready, finalise it. Finalising locks the inputs and produces an immutable, versioned scoring run.
Review and sign-off
Send it to John, then finalise
One gate stands between a scored assessment and a finalised one. Nothing goes to a client that John has not read and signed off. You handle this from the Summary step, and it usually takes one message and a day.
- When you are happy with the assessment, press Send to John for review on the Summary step.It appears on his review queue in the Workbench, with your name and the date you sent it.
- Wait for the approval. The Summary step shows where it stands, so you do not need to chase.Once he approves, the step says so and gives you the date he signed it.
- Finalise and lock. From there you can generate a client deliverable.
- If you change anything after the approval, send it again.An approval covers the version it was given for. Editing after sign-off means John has not seen what you are now proposing to send, so the approval stops applying and the record goes back on his queue.
Reading a finished score
Read the result honestly
A finalised assessment opens straight to Summary and Interpretation, which is the answer rather than another form. Every element of that screen is arranged to show the confidence and its limits at the same time.
- Read Platform Value as the headline, and quote it with its range rather than on its own.Business, Power and Infrastructure each carry a range of their own, computed the same way.
- Check the coverage line and the overall uncertainty before you rely on any figure. A line reading 1 of 51 rated, 2% of applicable, uncertainty very high means the number is directional at best.The more subcomponents you assess, the tighter and more defensible the ranges become.
- Find the likely constraint, which is the weakest module. It is usually the number a client remembers, and it is where the next pound of effort pays back most.The module bars are ordered weakest first for exactly this reason.
- Read the Platform Power triad as words. Economic, Perceived and Defence value are reported as ordinal ratings such as Emerging or Established, never as decimals.The section on reading the outputs, above, explains why those two kinds of result are kept apart.
The client report
Write the report the client actually reads
The client report is the document that represents the firm. You write six sections in your own words, the score fills in the arithmetic, and the client reads it as a branded PDF or a web page you send them a link to. It is the newest surface in the studio and the one worth learning first.
- From the engagement, open a deliverable and press Client report.The editor opens, headed with the client's name so you always know whose report you are writing.
- Write the six sections: the business, where the advantage sits, what is holding it back, what to do about it, what that is worth, and the technical appendix.Each section saves as you go. The report cannot be produced while any of them is empty, and the refusal names the ones still blank.
- Quote a figure only after declaring it. Any number in your prose has to come from the run.If you type a number the assessment did not produce, the editor refuses and says which one — that guard is what stops a report claiming something the score does not support.
- Send it to John if the record is client-facing, exactly as with the assessment.The words a client reads are approved separately from the scores. An edit after approval withdraws it, same rule as before.
- Download the branded PDF, or create a share link for a client who would rather read it in a browser.The PDF is named for the client and the month. The link shows the same content, and you can revoke it at any time.
Deliverables
The document packs
Alongside the client report, an engagement generates .docx packs — a Platform Power Report, an Executive Summary, a Heatmap. These are the internal and technical path: fuller, denser, and drafted by AI for you to approve. Nothing AI-written reaches a client without your sign-off. If you are sending something to a client to read, start with the client report above.
- On the engagement, pick the document and an audience. The audience defaults to Internal draft.An internal draft generates in one click, for your eyes only.
- For a client document, choose Client-facing. A review step then names the document and lists the release gates it has to clear before you confirm.Cancelling backs out with nothing sent. The extra step is deliberate friction on the one action that leaves the building.
- Review each AI-drafted section and approve it. A pack is not client-ready while any section is still pending.Download re-checks every gate again before a single byte is written.
Earnings
See what you have earned
Your commission, workshop recovery fees, and projections sit in one place, with every figure disclosed plainly and a statement you can download.
- Read the cards: earned year to date, pending, invoiced, paid, and projected unpaid.
- Commission lines appear as engagements and recovery fees are recorded against you.You can download a statement whenever you need one.
Workbench
Sharpen your practice
Certification, the practice arena and power drills sit together in the Workbench. This is how you earn the assessor level that unlocks high-stakes ratings, and how you keep it.
- Work the certification ladder, and clear your next action from the bench queue.
- Run drills and practice-arena sessions to stay calibrated between engagements.Levels and streaks here measure your craft rather than your client or deal activity, and they are visible only to you.
The rules that never bend
Principles
Four commitments shape everything above, and they are the reason a Grassmarket assessment holds up in the room.
Honest about uncertainty
Ranges, coverage, and confidence are shown as plainly as the headline number, because a score you cannot yet defend ought to look like one. This is why the tool would rather show you a wide range than a tidy figure it cannot support.
Two tracks, kept separate
Continuous scores rank what to fix first. Rule-based gates produce the rating word a client reads. The value bridge prices work in pounds. Score points and currency never appear in the same equation, because the moment they do the arithmetic stops meaning anything.
A person approves every AI draft
Deliverable prose and practice feedback are drafted by AI, and a person has to approve each one before it counts. Nothing drafted by a model reaches a client on its own, and the approval is recorded against the document rather than assumed.
Fail loudly rather than fill a gap
A missing input or an unmet gate stops the work and says why. The platform never guesses a value, substitutes a default, or quietly fills a gap on your behalf, because a silent fallback is indistinguishable from a real answer once it is in a report.