Consulting
Everything on this site came out of documents nobody could query
Financial analysis, modelling and data work — for organisations sitting on information they cannot use, or decisions they cannot evidence. This site is the demonstration; the same work is available to your organisation.
What we do
Four problems rather than a list of skills, because most people know what they are stuck with and not what it is called. Each says what arrives at the end, since “financial consulting” as a phrase tells nobody anything.
Documents nothing can query
Anyone with figures locked in reports, filings or statements
The PDFs, factsheets and spreadsheets that hold everything you need and answer nothing you ask. Turned into a structured series you can query, compare and keep current — with the checks that catch it when a source changes shape, because sources change shape without telling you.
What you get: A running extractor, a documented schema, and the data in a form your own tools can read.
Numbers nobody can trace
Anyone about to put financial data into a model, a board pack or an AI system
Where did this figure come from, when was it true, and what has changed since. Reconciliation across sources that disagree, provenance attached to every value, and a record of corrections rather than a silent overwrite. This is the work that has to happen before a model is worth trusting, and it is the work most often skipped.
What you get: A dataset where every figure carries its source and date, and a record of what changed and when.
A model you will have to defend
Founders raising, boards deciding, finance teams planning
A forecast somebody will challenge — an investor, a lender, a board member who used to do this for a living. Built so the assumptions are visible and each one can be answered for, rather than buried in a cell reference. Fractional CFO and FP&A work where the model needs an owner rather than a builder.
What you get: A model whose workings are legible, and somebody who can sit in the room while it is questioned.
Reporting that rebuilds itself every month
Anyone assembling the same pack from the same sources, by hand, again
The management pack that takes three days and is stale on the fourth. Pipelines that run unattended and fail loudly rather than silently, dashboards that read from them, and market or competitive analysis that updates rather than expiring. Experience across Databricks, SQL Server and SSIS, and the pragmatic end of the stack where most real work happens.
What you get: Something that runs without you, and can be handed to somebody else.
The evidence is the site
Most pages like this assert a capability. Every claim below links to the thing itself, so you can check rather than take it on trust.
Extraction from documents that were never meant to be queried
Monthly market documents published as prose and tables, turned into a queryable price history for every listed company. Daily and weekly releases from several institutions, extracted on a schedule and reconciled against each other. The result is a series; the inputs were never designed to produce one.
Comparable figures from filings that resist comparison
Regulatory filings that exist to be filed rather than read. Assembled across 22 institutions, three product types and three terms, they show a spread of 22.5 percentage points on the same one-year facility — a figure nobody publishes because nobody had put the returns side by side.
Analysis that finds what the data was hiding
A market concentration nobody had measured. A gap of nearly ten percentage points between what one institution advertises and what it reports charging. Neither figure appears in any single document — both emerge only once the series exists, which is the point of building one.
Errors caught, including our own
A return that annualised a partial window. A classification that mislabelled a major institution. A published claim that a provider corrected. Each found, fixed and recorded — because a pipeline that cannot catch its own mistakes is not finished, and the checks matter more than the extraction.
How it works
Remotely, or on site where it matters. Most of this work is done wherever the data is, which is usually nowhere in particular. Clients in Ghana, the United Kingdom and elsewhere are equally practical.
Scoped before it is priced. The first conversation establishes whether the answer you want is obtainable and whether it is worth what it would cost to get. Sometimes it is not, and saying so is part of the service.
A report is not a series. Anybody can read a month’s figures and write them up — several Ghanaian outlets did exactly that with the lending data on this site, before we published anything. What is harder, and what we do, is holding those figures as data: extracted on a schedule, comparable across categories and periods, and able to answer a question nobody thought to ask when the report was written. The difference is between an article about one month and something you can interrogate.
Deterministic where it matters. The extraction behind this site uses no language models. For financial figures that is deliberate — same input, same output, every time, and a wrong number is a bug rather than a plausible-looking guess. Models are useful downstream, on data that has already been made trustworthy.
Built to be handed over. A pipeline nobody but its author can maintain is a liability. Everything comes documented, in tools you already have where possible.
About Cedafin
Cedafin builds financial data infrastructure and the analysis that runs on it. The comparison platform on this site is our own work, kept current from primary sources — which is why the consulting offer is demonstrated rather than described.
Our team spans financial analysis and data engineering — the two halves this work requires, and the reason a pipeline built here is built by people who understand what the numbers have to mean. We draw on the disciplines an engagement needs rather than fitting it to the skills we happen to have.
The practice is led by John Addae, Chief Executive, who combines a BSc in Computer Science with an MBA in Finance.
If we have approached you about your data
That request stands on its own. We publish what providers send us, cited and dated, at no cost and with no expectation of anything in return — and we would do so whether or not this page existed. The two things are separate and will stay that way.
If any of that sounds like your problem
Tell us what you are sitting on and what you wish you could ask of it. If it is not something we can help with we will say so — and if it is, the first conversation is about whether the answer is worth having, not about a proposal.