Quick answer: Year-over-year comparison still matters when traffic patterns change suddenly, but raw YoY percentages can mislead. Annotate the shock dates, compare against more recent baselines (week-over-week, four-week rolling, period-over-period within 2020) and separate demand change from site damage. In spring 2020, stay-at-home orders and sudden closures moved entire categories overnight. Treat YoY as one lens, not the verdict.
Why YoY exists
Year-over-year smooths seasonality. A retailer always soft in January should not panic every January if 2019 January looked similar. YoY asks: are we better than the same slice last year?
That logic fails when the world does not rhyme with last year. If last April was normal trading and this April is not, a -60% session chart may say more about the market than about your new homepage.
What broke in early 2020 traffic patterns
Different businesses saw different shocks:
- Local foot-traffic queries collapsed for some venues
- Ecommerce and delivery interest spiked for others
- B2B research continued but sales cycles stretched
- Referral sources shifted as events cancelled and partners paused campaigns
If you only watch summed Sessions YoY, you will mix winners and losers inside one average and argue about the wrong page.
Build a comparison kit (not a single chart)
Keep four views ready in Universal Analytics or your dashboard tool:
- YoY for the executive question everyone will ask anyway
- WoW (week over week) for operational pacing
- Trailing four weeks vs prior four weeks for slightly stabler short trends
- Channel and landing page mix so composition shifts are visible
A Google Analytics dashboard that shows all four reduces meeting thrash.
Annotate like a historian
In Analytics, add annotations for:
- Offer changes and site launches
- Major email or ad pauses
- Tracking changes (filter edits, tag migrations)
- External shocks that hit your category (closure orders affecting stores)
Without annotations, next quarter’s you will invent stories. Annotations are cheap memory.
Separate tracking change from demand change
Before you accept a cliff:
- Confirm the tracking code still fires on key templates
- Check view filters and bot filters
- Compare raw hit volume in a debug view
- Validate goals still match thank-you URLs
A tag manager change that stopped checkout pageviews looks identical to a sales collapse in a lazy dashboard. GTM process notes in Google Tag Manager best practices are relevant whenever charts fall off a table overnight.
Segment before you judge
Split YoY by:
- Device
- New vs returning
- Brand vs non-brand (Search Console helps)
- Geo for multi-region firms
- Logged-out storefront vs app if applicable
A national average can hide that one city is shut while another region holds. Views hygiene from Google Analytics views keeps those segments trustworthy.
Use Search Console beside Analytics
Analytics sessions can fall while Search impressions hold (or the reverse). Search Console shows query demand. If impressions for core queries sink industry-wide, you are in a demand story. If impressions hold and clicks fall, look at SERP features, titles and ranking shifts. If both hold and on-site conversion falls, look at offer and UX.
SEO ROI conversations need this split; see SEO ROI for connecting organic value without magical attribution.
Alternate baselines that still respect seasonality
When last year is a poor twin:
- Compare to the trailing median of the last 4 to 8 “normal” weeks before the shock
- Compare to the same weeks for a sibling brand or region less affected (carefully)
- Index series to 100 at a pre-shock week and watch shape, not only percent
Finance may still demand YoY. Give them YoY plus the indexed chart so decisions are not only a red percentage.
Conversion rate traps
Conversion rate YoY can rise while revenue falls if low-intent traffic disappears. Rate improvements during a traffic collapse are not always a CRO victory. Report conversions and revenue beside rates. Read conversion rate optimization as a process for experiments, not as a reason to celebrate a smaller denominator.
Reporting language that keeps trust
Prefer:
- “Sessions are 45% below April 2019. Against the three weeks before closures, we are flat on returning users and down on new.”
- “Non-brand organic impressions for category X are down across the board; our share of voice is steady.”
Avoid:
- “Marketing is broken because YoY is red” without segments
- Hiding tracking changes in a footnote nobody reads
Practical monthly pack for leadership
- One YoY scorecard (sessions, leads, revenue) with annotations
- One WoW operations card
- Top landing pages that gained and lost vs pre-shock baseline
- One decision: pause, hold or invest in a specific channel
Keep the pack to a few pages. Sudden periods create meeting overload; clarity is the scarce asset.
Worked example of mixed signals
Imagine sessions down 50% YoY, but:
- Returning users only down 15%
- Non-brand organic impressions for your top category down industry-wide
- Conversion rate up because tire-kickers vanished
- Revenue down 35%
The story is demand and mix, not “the redesign failed,” unless landing page segments show a clear self-inflicted drop after a ship date. Annotations plus segments prevent the wrong postmortem.
Guardrails for automated alerts
If you use anomaly alerts on YoY, widen thresholds during known shocks or switch alerts to WoW for a while. Alert fatigue during a chaotic month trains people to ignore real outages later.
Communicating uncertainty upstairs
Executives often want a single percentage. Offer a primary YoY number plus a confidence note: tracking verified, demand shock annotated, alternate baseline shown. If finance models still need YoY inputs, feed them ranges and document assumptions in the same slide.
Resist rewriting history by changing filters quietly. If you fix a view, note when the break in the series happens so future YoY on that view stays interpretable.
Channel-level YoY
Paid may fall because you cut spend while organic holds. Always split channels before blaming creative or SEO. A blended YoY chart hides deliberate budget choices.
Checklist before you present YoY this month
- Annotations for shocks and tracking changes are in place
- Channel split reviewed
- Search Console impressions checked for demand vs CTR stories
- One alternate baseline chart prepared
- One recommended action, not only a red number
Do not average away the truth
A corporate rollup that blends a closed region with a busy region produces a soothing average and a wrong strategy. Show regional YoY beside the total whenever geography drove the shock.
FAQ
Should we stop looking at YoY until things stabilise?
No. Keep YoY, but never alone. Leadership will ask for it. Your job is to add context.
What date range should replace YoY?
There is no universal substitute. Pair YoY with a pre-shock baseline and short-range pacing.
Our bounce rate YoY looks “better.” Are we winning?
Maybe. Or low-intent visitors left. Check conversions and engagement events, not bounce alone.
How do we forecast?
Use ranges, not fake precision. Scenario planning (demand stays low / partial recovery / mixed by region) beats a single hockey stick.
Compare carefully, decide calmly
Year-over-year comparison is a tool. When traffic patterns change suddenly, add annotations, segments and alternate baselines so you do not punish healthy teams or reward lucky charts. If you want help building reporting that stays honest under shock, see Know what’s working.
