How do you know which parts of an old design to keep?

You know what to keep by auditing the old design for what earns its place, the elements users rely on, the parts that perform well, and the pieces that carry recognized equity, and then separating those genuine strengths from things that are merely old. The method is to evaluate each part by what it does, not by how current it looks, because age and failure are different conditions that happen to arrive together often enough to be confused. A redesign that keeps the proven and replaces only the broken preserves the familiarity and trust users depend on; a redesign that starts fresh and rebuilds everything throws all of that away to escape a few things that were actually wrong.

The method works because a design accumulates two kinds of value that are invisible if you only look at how it appears. The first is functional value: the patterns that work, the flows users complete without friction, the layouts that convert. The second is recognition value: the elements users have learned, the landmarks they navigate by, the brand cues they associate with you. Both kinds of value were paid for over time, in user learning and in tuning, and neither is visible in a screenshot, which is exactly why the start-fresh instinct discards them so easily. Auditing for what works forces you to look past the surface and ask of each element, is this carrying weight, before you decide whether it survives the redesign. The thing that looks most dated may be the thing users rely on most, and the sleekest part may be the one nobody uses.

A designer applies this by going through the old design piece by piece with evidence rather than taste. The logo and core color may look dated to a designer steeped in current trends, but if users recognize them instantly and associate them with trust, that recognition is equity you keep and evolve rather than replace. The primary navigation labels might be unfashionable, but if analytics show people find what they need through them, the labels work and changing them risks stranding users for the sake of freshness. Meanwhile the elaborate animated hero that everyone is proud of might show in the data that users scroll past it untouched, which means it looks important and carries nothing, and it is a candidate to cut. The audit pairs each element with what it actually does, what users rely on, what performs, what carries recognition, and that pairing, not the element’s age, decides its fate.

The exception worth naming is that keeping the proven is not the same as freezing the old, and equity is a reason to evolve something carefully rather than a reason never to touch it. An element can carry recognition and still need refinement, and the move there is to keep the recognizable core while updating its execution, not to either preserve it untouched or throw it out. There is also a limit to deference to data: something can be working only because users have no alternative, masking a problem rather than proving a strength, so the audit has to ask not just whether a part performs but whether it performs because it is good or because it is the only path. The point of the method is not to keep everything old, it is to keep what earns its place and to know the difference between a part that works and a part that has merely survived.

Before you rebuild, inventory the old design element by element and judge each one by evidence: what do users rely on, what performs in the data, what carries recognition they would miss. Keep those, evolve the ones whose value is real but whose execution is tired, and replace only the parts that are genuinely broken rather than merely dated. Audit for what works instead of starting from a blank canvas, and you carry the proven strengths and hard-won familiarity forward rather than making users relearn a site that was already serving them.