Any store can produce a list of thirty things worth improving. That list is not a plan, and writing it is the easy part. The hard part is deciding what happens in week one, because whatever goes first spends the attention nothing else gets.
We put every candidate through the same three questions. How much money is sitting behind this, in your own numbers. How confident are we that it is the cause and not a symptom. How long until you could tell whether it worked. Money at stake gets a fix onto the list. Confidence decides whether we would act on it now or measure first. Time to effect decides the order more often than either of the other two.
Then one rule overrides the ranking: if a fix depends on another fix, the dependency goes first even when it scores lower. Order beats size. A store that does the second best thing first and finishes it will beat a store that starts the best thing three times.
Money at stake is arithmetic, not a best practice list
The first question is answered with your numbers, not with a general claim about what stores should do.
Take the step you think is weak and put a value on closing part of the gap. If your product pages get a certain number of visitors and a certain share reach the cart, then a change in that share is worth a number of orders a month, and a number of orders is worth a number of dollars at your average order and your margin. Do the same for the free shipping line, the second order, the feed coverage, the checkout step. Now the candidates are on one scale.
What this does is kill the arguments that are really about taste. A homepage rewrite might be genuinely better, but if the homepage is a small share of entrances and almost nobody who buys passes through it, the money behind it is small and it can wait.
We size everything roughly and honestly. Rough is fine. The gap between candidates is usually large enough that precision would not change the order, and pretending to precision would only hide how much of it is estimate.
Size gets a fix onto the list. Time to effect decides where it sits.
Confidence is about evidence you can point at
The second question is where most prioritising goes wrong, because the fix with the biggest number attached is often the one we understand least.
So we grade our own certainty. High confidence means we can point at something in your data or in the store itself: a device gap in conversion, a checkout step where the drop is visible, a feed rejecting a share of the catalogue, an email flow that does not exist. Those are causes you can see, and we would act on them directly.
Medium confidence means the number is odd but the cause is a theory. Traffic is up and orders are flat, and we think it is the mix rather than the store. That gets a measurement first, not a rebuild.
Low confidence means we are guessing, and guesses do not get the first slot no matter how big the prize looks. They get the smallest test that would move them up a grade.
Saying which grade a recommendation sits in costs nothing and changes how you read the whole plan. It also makes us accountable later. If a high confidence fix does not move the number, that is a miss and we say so in the report rather than quietly moving on to the next item.
Time to effect decides the order more than size does
Two fixes can be worth the same and belong in completely different weeks, because one shows up in days and the other in months.
Conversion and order value changes are fast. You change the page, the traffic arriving tomorrow meets the new page, and within a couple of weeks you have enough orders to tell. Email and SMS flows are nearly as fast once they are live, because they run against people who are already in the buying cycle. Paid changes report quickly but need enough conversions to be trusted, so the wait is set by volume rather than the calendar.
Search is the slow one. A collection page rewritten today may take months to hold a position, and a site wide structural change longer than that. Repeat rate is slower still, because it is measured across a purchase cycle you cannot compress. If your customers reorder every four months, no amount of work tells you anything real in three weeks.
So we start the slow work early and the fast work first. Those sound contradictory and are not. The slow items get their first step in week one so the clock starts, while the fast items carry the quarter and pay for the patience the slow ones need.
The rule that overrides the ranking
Fix what the next fix depends on, even when it ranks lower.
Dependencies are everywhere once you look. There is no point raising traffic to a page that loses people at the cart, because you are buying the same loss at a larger scale. There is no point tuning bids on a feed where a share of the catalogue is rejected, because you are optimising a campaign that cannot see half your products. There is no point building a win back flow when nobody is being added to the list, or setting a free shipping threshold before you know what your orders currently cluster around.
So the sequence follows the dependency chain, not the score. Make the thing measurable, then make it work, then make it bigger. Tracking before optimisation. Feed before budget. Page before traffic. Flow before campaign.
This is also the rule that stops the plan being a wish list. Once the dependencies are drawn, a lot of the thirty items turn out to sit behind two or three others, and the real decision was never thirty wide. It was a choice between three starting points, which is a conversation a store owner can actually have.
A worked ranking, on four candidates
Say a store is weighing four pieces of work. The numbers here are invented to show the method.
One: shipping cost shown on the product page. Money at stake, moderate, because a visible share of carts end at the shipping line. Confidence, high, because the drop is visible between cart and checkout. Time to effect, two weeks.
Two: a rewrite of the twelve main collection pages. Money at stake, large over a year. Confidence, medium. Time to effect, three to six months.
Three: a post purchase and replenishment flow for a product people reorder. Money at stake, large. Confidence, high, because there is no flow today and reorder timing is known. Time to effect, one purchase cycle.
Four: a new social channel. Money at stake, unknown. Confidence, low. Time to effect, unknown.
The order is one, three, two, four. The shipping fix goes first because it is fast, certain and cheap, and it makes every later traffic dollar worth more. The flow starts in the same week because its clock is long. The collection rewrite begins once the first two are shipped, and the new channel waits until something proves it deserves the attention. Nothing on that list was wrong. Three of the four were simply not first.
When two fixes tie, we take the reversible one
Ties happen. Two items with similar money behind them, similar confidence, similar timelines. The tiebreaker we use is reversibility.
A change you can undo in an afternoon is worth more than a change of the same size that commits you for a quarter, because the cheap one buys information. If it works you keep it. If it does nothing you learn something about the store and you are back where you started by Friday.
The second tiebreaker is who else is blocked. If one item needs a developer who has two free days this month and the other needs nobody, the one needing the developer takes the slot while the slot exists.
The third is whether the work leaves something behind. A flow that keeps running, a feed rule that keeps applying, a page template that fixes every future product, all beat a one time push of the same value. We would rather install something that keeps paying than perform something that has to be repeated.
None of this is complicated, and that is deliberate. A prioritising method you cannot explain to the person doing the work is a method that gets ignored the first time a week gets busy.