Year-over-year reports when the definitions have changed.

A year-over-year chart assumes each line means the same thing in both years. Platforms change how they count more often than that, and they rarely restate old figures to match.

This year has a clear example. In March 2026 Meta narrowed click-through attribution to link clicks only, so a client's Meta conversions for last November and this November were counted under different rules.

Tracking and consent do the same thing more quietly. A tag added in spring or a consent banner changed in summer moves the baseline without anyone touching a campaign.

Go through each source for the twelve months between the two periods. Five questions find most of what moved.

• Definitions

  Has the platform renamed, narrowed or widened a metric you report, or changed a default it counts with?

• Tracking

  Were tags added, removed or moved to a server, and on what date?

• Consent

  Did the consent banner change, or the share of visitors accepting it? Either moves how much is observed and how much is modelled.

• Accounts

  Were campaigns restructured, accounts merged or currencies changed, so that one name now covers something different?

• Calendar

  Do both periods hold the same trading days and the same promotions? Black Friday falls on a different date every year.

Use the first wherever you can, and fall back in order.

• Compare on a stable measure

  Orders, revenue and leads from the client's own systems do not change meaning when a platform does. Lead with them, and with ratios built on them, such as total spend against revenue.

• Restate the old period

  Where you hold the raw rows, recompute last year under this year's rule. That needs the underlying data, not last year's finished report.

• Mark it and split it

  Where neither is possible, show both years with each change marked on the date it happened, and say which parts of the line kept their meaning.

Put the stable comparison first: what the client sold and what it cost to sell it, against the same weeks last year. Then show the platform figures, with each change marked where it happened.

Write the caveat into the report rather than saving it for the meeting. One sentence per change, saying what moved and in which direction, is read by everyone who opens the report, including the people who were not on the call.

Restating last year needs last year's rows. SightShark syncs every client's sources daily into a data warehouse you own, from the date you choose at setup, and syncs add to the tables rather than replacing them. The raw history for a like-for-like comparison is in your account, not in a past report.

Each metric has one name, one type, one grain and one currency in the modelled tables, so a platform changing a definition is handled in one place for every client. The client's own orders and leads sit in the same tables as the platform figures they are compared with.