An answer to a question you can act on, and an honest account of how far it can be trusted.

The risk with analysis is not getting no answer. It is getting a confident answer that the data never actually supported, and a decision made on it. The work starts by establishing what the data can and cannot tell you.

Every competitor promises insight. Almost nobody promises to say when the data does not support the conclusion someone was hoping for. That is the differentiator, and it is checkable in the Salifort attrition analysis and the Unicorn time-to-scale analysis, both public and readable end to end, reasoning included.

1. 01

   A written question, agreed before analysis begins, tied to a decision the business is actually going to make.

2. 02

   A data audit stating what exists, what is usable, what is missing, and what the data can and cannot answer.

3. 03

   A cleaned dataset, with every cleaning decision documented and justified rather than silently applied.

4. 04

   The analysis itself, documented so the reasoning can be followed and checked, not delivered as a conclusion with no visible working.

5. 05

   Findings with stated confidence and limitations. Where a result is weak, it is reported as weak.

6. 06

   Charts that carry the finding, not decoration. A chart is a claim.

7. 07

   Prioritised recommendations, each tied to a decision and an expected effect.

8. 08

   A walkthrough, so the person who has to act on it understands it without needing an analyst present.

Documented cleaning decisions and stated limitations are the difference between an analysis someone can defend in a board meeting and a number they have to take on faith.

Optional, scoped separately: a dashboard for ongoing monitoring, or a model deployed behind an endpoint if the finding needs to run repeatedly rather than once.

1. 01

   Question definition

   Establishes the decision the analysis is meant to inform. An analysis with no decision attached produces a document nobody uses.

2. 02

   Data audit and feasibility

   Assesses what data exists and whether it can answer the question. Output: a short written statement of what is answerable, what is not, and what would need to change to make the unanswerable parts answerable. This stage can end the engagement, and that is a legitimate outcome: if the data cannot support the question, saying so early avoids a decision made on a false result, well before the bulk of the work happens.

3. 03

   Cleaning and preparation

   Documented as it happens, not retrofitted.

4. 04

   Analysis

   Descriptive work, statistical testing, or modelling, as the question requires. Method chosen to fit the question, not to look sophisticated.

5. 05

   Findings and recommendations

   Written for the person making the decision, not for an analyst.

6. 06

   Walkthrough and handover

   The finding gets explained to the person who has to act on it.

Typically two to three weeks, staged. The audit stage sits early enough to know whether the question is answerable before the bulk of the engagement is committed. Part of what keeps that timeline tight is Claude Code, used throughout the work alongside the platforms listed below.

Question. The decision this analysis will inform, stated plainly, since without it the output is interesting rather than useful, and one named person who will act on the result, since analysis with no owner does not get used.

Audit. Access to the actual data, not a summary of it, since summaries hide the problems that determine whether the question is answerable, and context on how the data was collected and by whom, since collection method is usually where the bias lives.

Cleaning. Answers on anomalies, duplicates, and known bad periods. Only the business knows that a given month's numbers are wrong because a system changed.

Findings. Willingness to hear a negative or inconclusive result. Stated upfront, since it is the most likely point of friction.

Handover. Availability for the walkthrough. A finding nobody understands does not change a decision.

Not a dashboard service: a dashboard can be an output of an engagement, but ongoing reporting and BI maintenance are not offered. Not data engineering: building pipelines, warehouses, or ingestion infrastructure is a different discipline. Not ongoing analysis: each engagement answers a defined question, and a standing analyst arrangement would be negotiated separately. Not included: acting on the recommendations, which is the business's own work, or a separate engagement.

Findings are bounded by the data. No claim gets made that the sample cannot support, and no number gets produced to fill a slot in a report. If the honest answer is that the data is insufficient, that is the deliverable.