Send us your model and its financials, or a link to them. Choose clean-up, review, improvement, or a client-ready report. Want it in your house style? Give us your font and colours. It comes back checked by an analyst before it ships.
An equity analyst improving their own model and a government or institution needing independent comfort on someone else's are not the same transaction. We treat them as two products.
Send us your model and choose your service; we email your output back with a secure download link. No sales call, no scoping meeting. A Sophors analyst confirms the final output before it ships to you.
Start with your next model.
Request a quote and we scope the engagement to your model, your timeline, and the level of assurance you need. Where you need a formal opinion, we agree its scope and basis at quote stage.
Tell us your requirement when you request a quote.
Titles, inputs, and style guide brought in line for the 3-statement forecast and, where the model includes one, the valuation worksheet.
Formula integrity, circular references, hardcodes, and broken links checked line by line, and your source financials checked against reported figures, with every difference listed. Corrections are applied where you ask for them.
Recommendations against our own consulting and training practice and established public methodology, built into the model rather than just written up.
A client-ready write-up of the model's outputs, in your own brand template.
Financial modelling and valuation, sector by sector.
Models drafted with ChatGPT, Copilot, Claude or a spreadsheet add-in often look finished: tidy tabs, sensible labels, a valuation at the end. The figures underneath still need to be traced, recalculated and tested like any other model.
We review them the same way as a model built by hand, with extra attention to the errors these tools tend to make. It works on both tracks: send it to us directly, or request a quote if the model supports a lending, investment or policy decision.
A line or two on what the model is for helps us check that it does what you need.
Results entered as fixed values that look like formulas, so they never update when an input changes.
Prices, growth rates and multiples that read as plausible but trace to no filing or dataset. We check each one against official and audited sources.
A row that refers to the wrong year or line, often in only one or two cells. We compare every formula with the rest of its row.
A balance sheet that doesn't balance, or cash that doesn't match the cash flow statement.
A discount rate, terminal value or valuation date that doesn't fit what the model is for. We test the method as well as the arithmetic.
Eight steps, in this order. Each one draws on published standards for model validation, valuation and spreadsheet design, listed beside the step and in full at the end of this section.
We confirm what the model is for, which decisions rely on it and how material its outputs are. That sets how deep the review goes: a model that sets a tariff or a capital level gets more testing than one behind a single board paper.
We recalculate the model independently and reconcile our figures to the values saved in the file. Review starts only once the model's own numbers are reproduced. A check that cannot be automated is completed by hand.
We trace how each sheet feeds the next and test for broken references, numbers typed into formulas, circular references, unused content and missing balance checks. Anything unused is listed for your decision. We never hide or delete it.
We assess whether the method suits the model's purpose and whether each key assumption is backed by evidence. Valuation models are also checked against the international standard for valuation models.
We trace each input to its source, preferring official and audited figures, and check that fair-value inputs use observable market data where it exists.
We compare outputs with actual results, peer benchmarks and an alternative method where one applies, then test the assumptions that move the answer most.
Each finding is stated with its effect on the model's headline output, such as the change in value per share, and graded by whether it can be corrected. The report also records each judgement: what we considered, what we set aside and why.
Where you ask us to correct the model, the fixes are built into the file and steps 2 to 7 run again on the corrected version. We then set out which outputs to watch and what result should trigger the next review.
Bank model risk guidance tests whether a model's method is sound. Spreadsheet standards set out how a model should be built so others can check it. The current US guidance leaves simple spreadsheet arithmetic outside its definition of a model. We apply both sets of standards in every review.
Standards ask for findings to be documented. We also state what each one does to the output that matters, so you can rank them by consequence.
Each finding is confirmed in the model's own recalculated numbers before it reaches the report.
Unused sheets, names and links are marked and listed for your decision. The file you get back still holds everything you sent.
For each decision point, the report keeps what was considered, what was set aside and why, so the review can be explained to a board or a regulator later.
We do not give an independent review of a model we built. If we have improved your model, any later review is done by a different analyst and lists every change we made, and we do not validate it for a third party. Nothing from one client's model is used in work for another.
| Standard | Issued by | Edition | What we take from it |
|---|---|---|---|
| SR 26-2, Revised Guidance on Model Risk Management | US Federal Reserve, OCC and FDIC | April 2026 | Conceptual soundness, outcomes analysis, ongoing monitoring and effective challenge. Replaced SR 11-7 (2011). |
| SS1/23, Model risk management principles for banks | Bank of England, Prudential Regulation Authority | May 2023 | Model risk classification and independent validation. |
| Validation principles under Basel II | Basel Committee on Banking Supervision | 2005 | Validation combines quantitative and qualitative work and is reviewed independently. |
| IVS 104, 105 and 106 | International Valuation Standards Council | Effective Jan 2025 | Data and inputs, valuation models, documentation and reporting. |
| IFRS 13 Fair Value Measurement | International Accounting Standards Board | Effective 2013 | The fair value hierarchy and the priority given to observable inputs. |
| FAST Standard | FAST Standard Organisation | 02c, July 2019 | Workbook, worksheet and line-item design rules. |
| Financial Modelling Code | ICAEW | 2018 | Consistency, clarity and error reduction. |
| ISO 31000 | International Organization for Standardization | 2018 | Setting the depth of work by risk. |
We draw on these standards. None of their issuers has reviewed or endorsed our process. Where a standard was written for banks, we apply its principles to the model in front of us and say so in the report. Our report is a review of the model; where an engagement needs a formal opinion, its scope and basis are agreed at quote stage.
We hold client confidentiality as perpetual, not project-length. Nothing about the layout, formulas, or content of one client's model is ever carried into work for another.
Every model is checked against our own model review checklist, and our improvement recommendations draw on our own consulting and training practice plus established public methodology. Never from another client's proprietary model.
By policy, with no exceptions.
An analyst confirms every output before it reaches you.
Files stay re-downloadable for 7 days, then they're gone from our side.
Excel workbooks (.xlsx and .xlsm), including models built with an AI tool. We review formulas, inputs and outputs. Macros are not run; if a result depends on one, we say so in the findings.
Send the linked files as well if you can. If not, we review using the values saved in your file and list every link, so you know which figures we could not trace.
Only in the services that include correction or implementation. Every change is listed in a Change Log tab with the formula before and after, and nothing in your file is hidden or deleted.
Your model, with a Model Map tab and, where we made changes, a Change Log. Reviews also come with findings that state each issue's effect on the result that matters, such as value per share.
A clean-up comes back the same business day if it reaches us by noon. A review takes 1 business day for an NGX-listed company and 2 for others. All four services together take 3 business days, plus the time you take to decide on our findings. The price depends on the service and the size of the model; we confirm it and the delivery date before any work starts.
Only the analyst assigned to it and their supervisor. Your files are never used in work for another client, and on the online track they are deleted 7 days after delivery.
Yes, for parts of the work, such as tracing formulas and rechecking calculations. A Sophors analyst checks every output before it reaches you, and your files are never used to train AI models.
No. It is a review of the model and its results. Where you need a formal opinion, we agree its scope and basis at quote stage.
Equity analysts and asset managers can start with a model today. Governments, lenders, and institutions can request a quote. Reach out and tell us which.
Analysts: tell us about your model. Institutions: mention "request a quote".