Role
Marketing analyst
What do you do when two systems show a different number for the same thing? Answering that question takes more time in this role than any analysis - and the question comes back with every report.
Intro call
Four questions we would start with
What do you do when the ad panel, the analytics and the CRM show three different numbers?
If you work it out from scratch every time, part of a full-time job goes on reconciling rather than on analysis. The difference usually comes from the definitions, not from an error - but nobody wrote it down.
How do you know the report you send is read?
If nobody has ever asked a question about a single number in it, the report probably serves to tick a box rather than to decide anything. That is information too.
Who answers the ad hoc questions?
Usually you - and usually in the middle of working on something else. A single question takes half an hour, and several arrive a day.
How long does preparing the monthly report take?
If it is three days, then for three days you are not doing analysis, you are assembling. And the conclusions get written in the last hour, when there is nothing left in the tank.
We take over the assembling, you keep the interpretation
Collecting data from the sources, reconciling definitions, preparing the recurring report and answering repeat questions can be taken off the desk. What the numbers mean, and what cannot be read out of them, stays with the analyst.
- The definition is part of the number. - Every metric has it recorded where it comes from and what exactly it counts. That removes most of the arguments before they start.
- A divergence is shown, not hidden. - If two sources differ, the report says by how much and why - instead of picking one and pretending the other does not exist.
- The conclusions are written by a person. - The system does not explain why something fell. It shows what else changed at the same time, and the cause is named by someone who knows the context.
Scope
What can be taken off an analyst's desk
Nineteen example tasks. Each one has a human approval point. Below is a selection, and the scope we fit to your process.
Collecting data
- Pulling data from the ad panels, the analytics and the CRM into one place
- A completeness check - detecting missing days or sources
- Marking periods disturbed by technical changes
- A history of changes in the measurement setup
- An alert when data from a source suddenly stops
Reconciling the numbers
- A register of metric definitions, available next to the report
- Pointing out by how much and where the sources differ (settled by: the analyst)
- Detecting the same events counted twice
- A summary of which decisions rest on which definition
- Documenting how a metric definition changed over time
The recurring report and ad hoc questions
- A report in a fixed form, ready on the agreed day (conclusions: the analyst)
- A comparison with the previous period showing the size of the difference
- Answers to repeat questions from ready summaries
- A list of the most frequent questions, the candidates for a permanent report
- Access to data for the people who used to ask you
Conclusions and context
- A summary of what else changed in the period in which the result changed
- A comparison with the seasonality of previous years
- Pointing out results that fall within normal variation (interpretation: the analyst)
- An archive of earlier analyses of the same question
A day's work
Today
- The first days of the month go on assembling the report from several sources
- At every discrepancy you work out again which source is right
- Ad hoc questions break into the work several times a day
- You write the conclusions at the end, when there is least time
- You find out the measurement is broken while analysing an odd result
- You remember last year's analysis but cannot reproduce it
With AI assistance
- The recurring report is waiting, ready, on the agreed day
- The divergence between sources is described, not settled afresh each time
- Repeat questions have ready summaries you can point to
- Time goes on conclusions, because assembling stopped taking it
- Data that stops arriving reports itself, before it spoils the analysis
- Earlier analyses are in the archive together with the definitions they were built on
- Time to an answer
a business question gets an answer the same day
repeat questions have ready summaries
- Sources that agree
an argument about numbers turns into a difference in definitions
the divergence is measured and described rather than hidden
- A repeatable report
the report looks the same whoever produces it
assembling is a process, not an individual craft
- Data within reach
other people stop asking about things they can check themselves
the basic summaries are available without the analyst
We start with the definitions, not with a tool.
- Step 1
We write the definitions down
What counts as a conversion, a lead, a session - and in which system. That conversation usually reveals that two departments have been using the same word for different things for years.
- Step 2
One recurring report
The one used most often. Assembled automatically, with the definitions visible next to the numbers.
- Step 3
Ad hoc questions
We collect the questions for a month. The ones that repeat get a ready summary, and that removes most of the interruptions.
- Step 4
Context and the archive
Summaries of change over time, seasonality, an archive of analyses. Only here, because it needs a history of data reconciled under one definition.
Common questions
Questions that come up at the first meeting
Does the analyst not become unnecessary once the reports produce themselves?
A report that produces itself answers no question, it shows numbers. The analyst is needed exactly where someone has to say what follows from them and what is not visible in them. Today that part of the work loses to preparing the data. When preparing stops taking most of the week, there is time for the questions nobody has answered in a long while.
Does this fall under the AI Act?
Analysing marketing data is not a high-risk use. The limits come rather from the rules on personal data and on consent for measurement - and those apply whatever the tools.
Still have a question that is not here? Book a call
Next step
How many days a month does your analyst spend assembling the report instead of reading it?
We will show you what reporting looks like when every number has a recorded definition and the divergence between sources is described rather than settled afresh.