Layered demo

Three layers for two audiences on a quantitative platform

A portfolio platform has to convince a business buyer and a quant, and they cannot be shown the same thing. I packaged the pitch, the library and the notebooks as three separate entry points.

Pitch, library and notebooks — one per audience
3 layersPitch, library and notebooks — one per audience
The library ships with a test suite and CI, not as a notebook dump
Tested and containerisedThe library ships with a test suite and CI, not as a notebook dump
Results always shown next to the conventional approach, never alone
Against a baselineResults always shown next to the conventional approach, never alone

The client

Institutional portfolio management — an audience split between decision-makers who need the economics of a systematic approach and quantitative researchers who will judge the whole thing by the optimiser and the backtest.

The engagement

One capability presented as three layers, each aimed at a different reader, with an explicit instruction about which to open first.

The problem

Quantitative capability is uniquely easy to demonstrate badly. Show the business audience a notebook and they see nothing; show a quant a dashboard with a flattering equity curve and they stop listening, because they know exactly how easy that curve is to manufacture. One artefact cannot do both jobs, and pretending otherwise loses both rooms.

What I did

I separated the layers and told people where to start. The commercial layer argues the case in prose — architecture, what the optimisers do, where the data comes from — and is the stated entry point for a client conversation. The library is the credibility layer, and it earns that by being installable, tested, containerised and wired to continuous integration rather than by being a folder of scripts; a quant reads that as a team that ships. The notebooks are the transparency layer, showing the pipeline from data through factors and regime analysis to optimisation, so the method can be interrogated rather than trusted. And every result in the dashboard appears against a conventional baseline with risk-adjusted metrics, drawdown and turnover, because a return shown alone is not evidence and a professional audience knows it.

What was built

Three layers — a written analysis covering architecture, optimisers and data integrations for the commercial conversation; an installable library implementing portfolio optimisation including hierarchical approaches, backtesting, metrics and loaders for the standard market data sources, with a test suite, container and continuous integration; and a set of research notebooks walking the full pipeline from data through factors and regime analysis to optimisation — plus a dashboard surfacing the risk-adjusted result against a conventional baseline.

On the table at the end

  • Written marketing and architecture analysis
  • Installable optimisation and backtesting library with tests, container and CI
  • Research notebooks covering the full pipeline
  • Comparison dashboard: risk-adjusted metrics, equity curves, drawdowns, allocations

What it changed

Let a single capability be sold twice without dilution: the business reader gets an argument, the quant gets a library with tests and a container, and neither is asked to sit through the other's material.

How it ran

  1. 01

    Split by reader

    A prose analysis, a library and notebooks — each written for the person who will actually open it.

  2. 02

    Name the entry point

    The written analysis stated as where a client conversation starts, so nobody opens a notebook in a boardroom.

  3. 03

    Make the library shippable

    Installable, tested, containerised, on continuous integration — which is how a quant reads seriousness.

  4. 04

    Show the pipeline, not the result

    Notebooks covering data, factors, regimes and optimisation, so the method can be interrogated.

  5. 05

    Always against a baseline

    Risk-adjusted metrics, drawdown and turnover next to the conventional approach — a curve alone is not evidence.

Something similar on your plate?

Thirty minutes, no deck. I will tell you whether it is worth doing at all.