Alli Insights Blog

Data, AI, and growth, examined

Updated

MediaOpenAI ads: strong early returns for eight brands, at small spendSeven of the eight brands we measured earned more per OpenAI dollar than on their average media dollar, and the eighth broke even. Spend is still tiny, so how far that goes is the open question.Jason Karpeles · Nov 12, 2026 · 4 min readResearchSynthetic research: what to trust, and when to add real peopleAI estimates landed about four points from real survey results on average. That's close enough for a first look, and not close enough for every decision.Jason Karpeles · Nov 10, 2026 · 3 min readAI engineeringDrift is the bug nobody filesA broken feed throws an error. A drifting model keeps producing numbers that look fine. Nobody files a ticket, and the decisions built on it get a little worse every week.Nathan Barling · Nov 7, 2026 · 5 min readAI searchHow brands show up in AI answers, and what to measureThe same question rarely gives the same answer twice, so one AI answer says little about your brand. How often you appear, across many runs, says more.Nathan Barling · Nov 6, 2026 · 4 min readMeasurementThink Bayesian without the equationStart with what you already know, ask how well new evidence fits each explanation, and adjust your confidence. One brand search question shows the whole habit.Jason Karpeles · Nov 5, 2026 · 3 min readAI agentsMost agent failures are evaluation failuresTeams test their agents, the agents pass, and then they fail in front of customers. The problem usually sits in what the tests asked, which is something a team can fix.Nathan Barling · Nov 3, 2026 · 6 min readMarketingWhy teams ignore AI insightsAI can now tell a team what to change. Often nobody has been given the job of deciding whether to.Nathan Barling · Oct 31, 2026 · 7 min readMeasurementWhy your MMM's range matters more than its numberA mix model that says TV returns 2.6 is also saying it might return 1.5 or 3.8. Budget decisions made on the single number ignore most of what the model knows.Nathan Barling · Oct 29, 2026 · 5 min readAI engineeringYour RAG problem is a chunking problemWhen an AI assistant answers from the wrong passage, teams usually blame the model and try a bigger one. The answer was decided earlier, when the documents were cut up and searched.Nathan Barling · Oct 26, 2026 · 4 min readAI governanceFour circuit-breakers every autonomous marketing workflow needsApproving everything is too slow, and letting an agent run free is too risky. Finance solved this with limits that stop the system and a person who restarts it.Nathan Barling · Oct 22, 2026 · 6 min readMeasurementFour systems, one customer: who gets credit for the saleSearch, social, affiliate, and CRM each report the same sale. Tests suggest some of that credit was never earned, though not every time.Nathan Barling · Oct 19, 2026 · 4 min readMeasurementWhy finance doesn't trust marketing's numbersFinance and marketing expect similar payback. What finance wants is a number it can trace, define, and put a range around.Nathan Barling · Oct 16, 2026 · 4 min readMeasurementYour highest-ROAS campaign may be selling returnsBidding models learn at checkout. Returns arrive weeks later, and the model never hears about them.Nathan Barling · Oct 13, 2026 · 5 min read
▶ 0:52Video tourAlli Alerts in actionA tour of Alli Alerts on a sample retail account: a SQL alert flags day-over-day CPA and ROAS changes, a budget pacing alert is built and scheduled, and a summary card explains what triggered it. Data science models and custom rules catch performance outliers, pacing issues, broken URLs, and feed failures, then notify your team where they work so they can act early.Oct 28, 2026▶ 0:47Video tourAlli Audience Gateway in actionA tour of Alli Audience Gateway on a sample retail account: a pixel is named, ad platform accounts are connected, allowed origins are set, and the Init tag is generated and copied. One first-party, server-to-server integration sends your site's conversion and audience signals to each platform through its Conversions API, so measurement and targeting hold up as browser tracking gets less reliable.Oct 28, 2026▶ 0:27Video tourAsk Alli in actionAsk Alli answers a question about campaign returns in Data Explorer, takes a follow-up, opens beside an audience persona in Audience Planner, and starts a workflow from a prompt. It turns a plain-language question into a verified answer, a dashboard, an audience persona, or a finished workflow, using the same governed data your teams already work from.Oct 28, 2026▶ 0:22Video tourAlli Brand Equity in actionA sample brand's composite score and signal trajectory open in Alli Brand Equity, followed by a period comparison on the chart, a competitive positioning view, and competitor analysis by share of search. The composite score shows where the brand stands each week, and the trajectory of every signal behind it shows what moved it. Brand investment walks into the budget meeting with proof, on equal footing with performance spend.Oct 28, 2026