Mosaic analysis

A research analyst rarely gets one big revealing fact. What they get is dozens of small ones.

A supplier mentions his order book. A job site shows the company hiring in one division. A competitor complains about pricing. A distributor is holding more stock than last year.

Not one of those, on its own, tells you anything. Put them side by side and a picture appears — like tiles in a mosaic, which is where the name comes from.

The workbook says this is allowed. It is, in fact, the job. The insight belongs to the analyst who assembled it, not to anybody who leaked it.

The definition. Section 4.2.1: "Often analyst collate information from different sources, which individually may not be significant but when put together with other public or non-public information may provide critical insight to the information. These is called as mosaic analysis. Such mosaic analysis is acceptable."

Read that carefully, because the phrase "public or non-public" is doing real work. Mosaic analysis is not restricted to public information. Individual pieces may be non-public, provided each is insignificant on its own.

Where the research comes from. Section 4.2 describes the raw material. An annual report is "a treasure trove," but information in annual reports "is not adequate," and because they are published once a year the contents start becoming dated. So an analyst spends significant time on the economy, the industry and the company — speaking to industry experts, accessing market research reports, doing secondary research, and often primary research that involves visiting facilities and speaking to customers, suppliers and employees. Those conversations are where the tiles come from.

The boundary, and it is the whole examinable point. The workbook places its warning immediately after granting permission: "Analysts need to be careful whether certain insight drawn came from their mosaic analysis or by being privy to a specific non-public price sensitive information."

So the test is not what the conclusion is. Two analysts can publish the same number, one legitimately and one not. What separates them is the route.

On figures. Section 4.2.1 gives no number — no count of sources that makes a mosaic, no materiality threshold for an individual piece. The judgement is qualitative, and the workbook offers the 3 insider information tests of source, impact and certainty as the way to make it.

Illustrative. An analyst covering a paints company assembles 5 pieces over 6 weeks. None is material by itself.

Put together, the analyst builds a case: input costs falling, prices being raised anyway, distribution pushing volume, and a rural demand tailwind. He models gross margin up 180 basis points and volume growth of 11% against consensus of 7%, and publishes a Buy.

That is mosaic analysis, and the workbook says it is acceptable. The dealer and the contractor each told him something non-public, and each was insignificant on its own.

Now the version that is not acceptable. Suppose instead the CFO had told him, before the results were filed, that gross margin would expand 180 basis points. He publishes the same number in the same report. Nothing on the page changes — but the insight now came from a specific piece of non-public price sensitive information, which is exactly the boundary section 4.2.1 tells him to watch.

The practical discipline follows from that. The analyst keeps a record of which tile came from where, so that the route to the conclusion can be shown, not merely asserted.

Section 4.2.1 is titled "Insider information vs Mosaic analysis," so the workbook examines the 2 as a pair and a question on either one is really a question on the boundary between them.

Expect a scenario: an analyst reaches a valuable conclusion, and you are asked whether the method was permissible. The answer turns on whether the individual pieces were insignificant and assembled by the analyst, or whether one specific non-public price sensitive item did the work. Also examinable is the flat statement that mosaic analysis is acceptable — candidates who assume any use of non-public information is barred get it wrong.

• Mosaic analysis may include non-public pieces. The workbook says "public or non-public" explicitly. The test is whether each piece is insignificant on its own.
• The conclusion being valuable does not make it insider information. A material insight built legitimately is the analyst's own work product.
• One decisive non-public item poisons the whole report, however many innocent pieces surround it.
• The duty is to know your own route. The workbook's instruction is that analysts be careful about which of the 2 their insight came from — a record-keeping discipline, not a one-off judgement.
• Mosaic analysis is not the same as secondary research. Secondary research describes the source of data; mosaic analysis describes what the analyst does by combining many sources.
• Do not confuse it with sensitivity analysis or a checklist approach. Those are techniques for testing a conclusion; this is a method of reaching one.