Setting Up Rank Tracking From Nothing

An organisation with no rank data and a question about competitors can get to something useful in a quarter. The order matters because history cannot be acquired retrospectively.

Week one: define the question

Write down the decision this will inform. If there is none, stop.

Identify the competitive set by substitution, not by category.

Decide the marketplace and region.

Decide the shortest event you need to detect, which determines cadence.

This takes an afternoon and it prevents the common outcome of a dashboard that tracks everything and answers nothing.

Weeks one to two: start collecting

Begin before the design is perfect. History is the asset and every day not collected is gone.

Twenty to fifty products. Yours, direct competitors, a neutral basket, and one or two category leaders.

Daily, at a fixed time, unless you specifically need sub-daily.

Six fields minimum: rank with its type, price, availability, review count, title, timestamp.

A script and a database, or a provider, or a spreadsheet. The tooling matters much less than starting.

Weeks three to four: verify

Manual reconciliation. Twenty products, checked by hand against the marketplace.

Confirm the rank type you are capturing is what you think.

Confirm the identifiers resolve to the intended products.

Set up the four health monitors: row counts, distinct products, field completeness, value change rate.

Fix what the verification finds, which will be something.

Weeks five to eight: characterise

Establish the noise floor for each tracked product. Median and range over a quiet period.

Identify which products are too slow to analyse, and stop expecting conclusions from them.

Build the basket median and start plotting it.

Note the intraday and weekly patterns.

Do not draw competitive conclusions yet. Eight weeks is barely enough to know what normal looks like.

Weeks nine to twelve: first analysis

Answer the original question, with the caveats the data supports.

Run the defensible-analysis checklist.

Present it as a one-page report with a chart, a conclusion and the uncertainty.

State what you could not answer and what would be needed.

After the first quarter

Weekly review, fifteen minutes. Anything outside its range, sustained; availability and price changes; new entrants; basket movement.

Quarterly: watch list review, basket refresh, seasonal profile once you have the years.

Annually: recalibrate any conversion basis, review whether the questions have changed.

What to buy, and when

Nothing, for the first quarter. Establish the practice first.

A provider, if you need history you did not collect — which you will discover during the first analysis, when someone asks about last year.

A provider, if maintaining collection is not sustainable for your team.

Not a dashboard. The output of this work is a page someone reads, and a dashboard is what gets built when nobody decided what the question was.

The mistake to avoid

Buying a tool first. It arrives with a dashboard, a conversion table and no relationship to your question.

The sequence above costs almost nothing and produces a practice that a tool can then accelerate. Reversed, it produces a subscription and a login nobody uses after month three.

The one thing to do today

Start collecting fifty products, daily, with six fields, however crudely.

In two years that is a dataset nobody can buy and you cannot recreate. Everything else in this article can be improved later; that one cannot be started later.

The first-week spreadsheet

For anyone who wants to start today without a tooling decision.

One sheet, one row per observation. Date, product name, marketplace identifier, rank, rank type, price, availability, review count, notes.

Fifty products, entered manually or by a short script, once a day at a fixed time.

A second sheet listing the products, with their category, listing structure and why they are on the list.

A third sheet with a running note of anything observed — promotions, outages, new entrants.

Ten minutes a day if entered by hand, which is sustainable for a quarter and long enough to prove whether the practice is worth automating.

The data is identical in value to what a tool would collect, and the exercise teaches you what you actually need before you buy anything.