Resources

Case Studies

How specific problems were diagnosed and solved, written up with permission.

Four connected ivory platforms hold a brief folder, glass lens, joined modules and delivery folio.

A model that would not converge, three weeks before submission

A doctoral candidate had a structural model that failed to converge in AMOS and no idea why.

The problem. The structural model would not converge, and the error message was not informative. The candidate had been adding and removing paths more or less at random for a fortnight.
What we found. The measurement model had never been established. Two constructs shared four cross-loading items, so they were not empirically distinct, and the model was effectively trying to estimate the same construct twice.
What we did. Ran the CFA first, demonstrated the discriminant validity failure with HTMT, revised the measurement model with the items that held together, then estimated the structural model — which converged immediately.
The point. Almost every "my model will not converge" question is a measurement model question. Fixing the structural paths cannot solve it.

Four years of sales data and no way to read it

A distributor knew their data contained useful answers but had no way to get at them.

The problem. Sales, stock and delivery data lived in three systems that did not talk to each other. A monthly report took two days to assemble by hand and was out of date when it arrived.
What we found. The data was fine. The obstacle was that no two systems agreed on a product identifier, so nothing could be joined reliably.
What we did. Built a mapping layer to reconcile the identifiers, then a Power BI model on top with scheduled refresh. The two-day manual report became a dashboard that was current every morning.
The point. Most dashboard projects are actually data-reconciliation projects. The visualisation is the last and easiest part.

A systematic review with 4,000 abstracts and one reviewer

A masters candidate had a well-designed review and a screening workload that could not fit in the time available.

The problem. The search returned just over 4,000 records after deduplication, with eleven weeks until submission and no second reviewer.
What we found. The search string was sound but unnecessarily broad in one concept, where three of the eight synonyms were pulling in an adjacent field.
What we did. Tightened that concept with the rationale documented, which cut the yield to about 1,800 without losing any of the known key papers. Then set up structured screening with a double-screened 10% sample so intra-rater agreement could be reported as a limitation rather than left unaddressed.
The point. A search that is too broad is not more rigorous. It is a search that has not been specified carefully enough.

A launch decision that the research reversed

A founder wanted market research to support a launch. The research did not support it.

The problem. A founder had built a product for a segment they were confident about, and wanted sizing and competitor analysis for an investor deck.
What we found. Bottom-up sizing put the reachable segment at roughly a tenth of the figure in their existing plan, which had been derived top-down from an industry report. Customer interviews also showed the pain was real but was being solved adequately with a spreadsheet.
What we did. Reported it plainly, with the workings shown so the conclusion could be checked. We also identified an adjacent segment where the same product solved a problem nobody had a workaround for.
The point. Research that can only confirm the plan is not research. This one changed the launch, and the founder said afterwards it saved a year.

Next step

Tell us what you are working on

Send the brief, the deadline and anything you already have. You will get a reference number straight away and a considered reply, usually within one working day.

Get Expert Help Ask on WhatsApp

Fixed quote agreed before any work begins
A reference number you can quote on WhatsApp
Confidential handling, and an NDA if you want one
An honest answer if your deadline is not realistic