From Excel to PowerPoint with Julius: Method and Checks

Transform a marketing report in Excel into a shareable slideshow, without extrapolating or losing checks along the way. The agenda includes: preliminary checks, essential calculations, a six-slide brief, and PowerPoint export with Julius, illustrated by an example where ROAS drops from 5.0x to 3.3x while revenue increases by 20%. The limitations of the data and the final review structure the exercise.

Review the export, preserve nuances, and demand targeted corrections

Before distribution, the exported PowerPoint file must be opened and checked rather than accepted at a glance. Totals, label readability, and title appropriateness with the data should be verified, along with the presence of source notes and qualifications. It is recommended to try modifying the text and inspecting the export of the graphs, especially if they will need to be updated by others. In case of discrepancies, formulate a precise request: make simultaneously visible the 20% increase in revenue from paid search and the 80% increase in spending, retain the ROAS of 5.0x and 3.3x, and remove any mention of profitability improvement due to lack of available margins; the other verified figures remain unchanged. This discipline is particularly important as the provided elements are insufficient to recommend a budget cut, do not allow for identifying the most profitable channel, and lack margins and certain costs by channel.

Calculate the indicators: overall variation, contributions by channel, and ROAS

The calculations to be required cover the total attributed per quarter, the percentage of variation, and the dollar contribution of each channel, with formulas and source values to support them, compared to a reference tab. For paid search, the return on ad spend is calculated by dividing the attributed revenue by media spend, without deducting missing costs or proposing budget allocation. In the example, total revenue increases by $60,000, or 20%, with contributions of $50,000 from organic, $30,000 from paid, and a decrease of $20,000 from partners, which sums up to $60,000. Paid spending rises from $30,000 to $54,000, or 80%, while their attributed revenue only gains 20%: ROAS drops from approximately 5.0x to approximately 3.3x. Limiting a slide to the 20% increase in paid revenue obscures an important piece of information: the deterioration of the revenue/spending ratio.

Frame the data: definitions, scope, and preliminary checks

Before any request to the AI, clarify what the revenue item covers, as reserved revenue, cash receipts, and pipeline are not equivalent. The example uses attributed revenue by channel, useful for reporting but insufficient for estimating the increment generated by channel. The working workbook includes tabs for Revenue and Paid Search Expenses for Q1 and Q2, amounts in US dollars, a single-channel attribution per dollar, and a presumed stable attribution method, plus a Check Calculations tab serving as a reference to keep separate from the sources. An independent verification of calculations, in addition to self-checking by the AI, is recommended.

Formulate the analysis request: inspection, validations, and limitations

An Excel workbook can be imported into Julius, which accepts questions concerning multiple sheets. Explicitly name the tabs to be used and request inspection of notes, confirmation of periods and units, as well as the meaning of each line. Demand reporting of missing values, duplicates by quarter and channel, and any ambiguity, without modifying the data or creating slides yet. The verification tab serves as a reference and should not be merged with the sources.

From diagnosis to slideshow: a brief of 6 slides and production in Julius

Once the results are validated, the production prompt specifies the audience, duration, major conclusions, and data limitations. Julius can generate graphs from the sources: for the example, a grouped bar chart by channel over two quarters and a table for paid expenses and ROAS, with currency and periods clearly labeled. The brief proposes six slides for ten minutes, based solely on the verified results, covering the overall result, breakdown by channel, paid search expenses and ROAS, issues related to the decline of partners, limitations, and next steps, with concise titles and mention of the source tab and period. It excludes any invention of causes, references, forecasts, or missing costs, and requests a modifiable presentation. A useful title for the paid slide mentions the 20% increase in revenue and the drop in ROAS from 5.0x to 3.3x; for partners, it is suggested to show the decline and propose a review of the pipeline or account data, without inferring the cause. The design can be framed (brand colors, readability, layout), with Julius managing iterations on content, visuals, structure, and branding after a first draft, up to the PowerPoint export.