Outcome analytics
The dashboard that tells you why.
This is the exact analytics surface Avarto puts in your hands — the same charts, rendered the same way. We've loaded one honest story: Avarto found Pace Athletics was losing mobile shoppers to sizing doubt, fixed it, compounded a second win into $188,543 of measured incremental revenue over an 8-week test — statistically significant by week 4 against a 10% randomized holdout — and honestly retired the one tactic that backfired.
$188,543
measured incremental revenue over the 8-week test, against a 10% randomized holdout — 95% CI $72,017–$305,068
week 4
when the uplift cleared significance against the holdout (day 26)
~$2,772,000
assistant-attributed revenue across 16,800 conversions at $165 AOV
1. Where the visitors went
The assistant funnel for the period — and, crucially, WHY the engaged shoppers who didn't buy left. That last read comes from the conversation itself; page analytics and session replay can't produce it.
The funnel
visitors in the window
38.5% of sessions started a conversation
got to a product page or a recommendation
6.00% of engaged sessions ended in a purchase
Why the engaged shoppers who didn't buy, left
From the assistant's own objection labels, share of bounced engaged sessions.
Sizing doubt is the #1 objection — and about twice as common on mobile. That single read is what motivated the hero experiment below.
2. What actually moved revenue
Six experiments, each against a 10% randomized holdout. The bar IS the 95% confidence interval — never a bare point estimate. Green helped, red hurt, slate is not yet statistically significant (its CI still crosses the holdout line), and a sample too small to trust is labelled "not enough data" rather than shown as a confident number.
✓ green — a real uplift; ✓ red — a real drop (the variant hurt).
ns slate — not yet significant; the CI still crosses the holdout line (a positive point estimate here is trending, not proven).
Size-guide prompt (mobile)
Won ✓6.0% → 6.9% conversion · ~$112,000 incremental
Proactive price comparison
Won ✓5.4% → 6.0% conversion · ~$76,543 incremental
Free-shipping threshold nudge
Trending5.8% → 6.2% conversion · trending up but not yet significant — not counted
Assertive "buy now" tone
Retired5.6% → 5.1% conversion · loss caught and retired (not counted in the total)
Avatar on vs. off
No effect5.5% → 5.6% conversion · no measurable difference
Voice greeting (launched late)
Not enough dataLaunched late — not enough data yet to call it.
3. When it became real
Cumulative incremental revenue measured against a 10% randomized holdout, with 95% confidence intervals, day by day. It's honest about uncertainty: noisy and crossing zero through roughly the first 3.5 weeks, then it clears significance on day 26 (week 4) and climbs to $188,543 by day 56 (95% CI $72,017–$305,068). The two winning experiments reconcile to that total; the retired loss is revenue avoided, not added.
The shaded band is the 95% confidence interval; the line is the point estimate. Below the dashed baseline the variant is losing money. The marker shows the day the lower bound first cleared zero.
4. The why, in words
Every claim below traces back to a number above — no hype, sourced.
- 1
Sizing doubt was the #1 reason engaged shoppers left — 31% of bounced conversations (81,592 sessions) — and it was about twice as common on mobile.
Source: objection_label drop-off: sizing/fit 31% (81,592)
- 2
So Avarto tested a mobile size-guide prompt. Conversion went 6.0% → 6.9% (+15%, significant): about +$112k incremental.
Source: Experiment 1 — size-guide prompt (+15.0%)
- 3
A proactive price comparison added a second significant win (+11%, ~+$76.5k), and the two wins reconcile to the +$188,543 total the uplift band climbs to.
Source: Experiment 2 + uplift band (+11.1%)
- 4
The one tactic that backfired — an assertive "buy now" tone — lost 8.9% and was retired. That is revenue we avoided losing, not revenue we counted.
Source: Experiment 4 — retired loss (−8.9%)
- 5
A free-shipping nudge is promising (+6.9%) but not yet conclusive — its CI still touches zero — so it's reported as trending, not counted. The avatar-on test was a null and the late voice greeting hasn't gathered enough data; both are reported honestly, not hidden.
Source: Experiments 3, 5 & 6 — trending + null + not-enough-data
- 6
Compounded, the program reached ~$188.5k of measured incremental revenue over the 8-week test (95% CI $72k–$305k), statistically significant by week 4 — all against a 10% randomized holdout.
Source: uplift band — +$188,543 [$72,017, $305,068], significant day 26 / week 4
This is the analytics you'd get on your own data.
Same charts, same statistics, same honesty rules — pointed at your store instead of our demo one. Confidence intervals on every result, a holdout that proves causation, and the objections your shoppers actually raised.