I think my customer profile is wrong. How do I figure out who is actually buying my product?
Check it against behavior rather than rebuilding it from assumption. Upload the customer list you already have and you get what your actual buyers have in common: what else they purchase, what they follow, what they are in-market for. Where a profile is wrong it usually fails the same way, with demographics that are accurate and a behavioral picture that never matched. The work takes a session. Re-deriving the profile by hand takes 15 to 30 days.
What do my best customers have in common beyond age and location?
Usually the things nobody wrote down: what else they buy, which categories they are in-market for, what they follow, how they behave around price. Age and location are the weakest attributes you own, and they are the two most profiles are built on. Watt shows the attributes that separate your best customers from the rest of your customers, which is a sharper comparison than separating them from the general population.
How do I validate my customer profile with real data instead of guessing?
Test the profile against observed behavior and see what survives. The accuracy gap is large and documented. Peer-reviewed field work found third-party brokers identified males 25 to 54 correctly 24.4% of the time on average, with individual brokers ranging from 4.3% to 62.7%, against a 26.5% population base rate (Neumann, Tucker and Whitfield, Marketing Science, 2019, more than 90 audiences across 19 data brokers). Upload the list you have and Watt returns what your buyers hold in common against live behavioral signal, so you can see which parts of the profile hold and which were assumption.
My messaging targets one kind of customer but a different kind is converting. What do I do?
Find out who the converting group actually is before you change the messaging, because the obvious explanation is often wrong. Watt profiles the converting segment against the rest of your buyers and shows what defines it. Then you have a choice worth making deliberately: write to the customer who is converting, or work out why the intended customer is not. Both are decisions. Guessing between them is not.
How do I segment customers by behavior instead of demographics?
Group people by what they do rather than by what they are: what they buy, what they are in-market for, how they respond to price. The practical difference is that a behavioral segment gives you something to say and a demographic segment does not. Watt builds segments from live behavioral signal, so each group comes with the attributes that make it distinct and therefore with the raw material for a message.
How do I define a customer profile with no data team and no warehouse?
You do not need either. Export the customers you already have from Shopify, Klaviyo or your CRM, or describe them in plain language if the export is a problem. Watt does the matching and returns what your buyers have in common. The reason profiles get skipped is not that the analysis is hard. It is that nobody has 15 to 30 days to do it by hand.
I have a dozen customer archetypes. How do I tell which has the highest purchase intent?
Most archetype sets have the same two problems: several of the groups do not actually differ from each other, and the ones that do are not sized. Fix the second and the first becomes obvious. Watt tells you how many real people sit behind each one and which are showing category intent now, which usually collapses a dozen archetypes into three that matter.
How do I build a customer profile from real behavioral data rather than assumptions?
Start from a list of people who have actually bought, then find what they have in common that the rest of the market does not. That direction matters: profiles built the other way round, starting from a description and looking for people who match, tend to confirm what you already believed. Watt runs it from the buyers outward and shows the attribute behind every part of the profile.
What is the difference between an ideal customer profile and a persona, and do I need both?
Neither term was built for a consumer brand. Every published definition of an ideal customer profile is account-level and B2B: it describes a company, its firmographics, its budget authority and its buying committee. You do not sell to a company. Persona has the opposite problem, having drifted into a fictional character with a name and a stock photo. What a consumer brand actually needs is narrower than both: which real people buy, what they have in common, and how many of them exist. That is what Watt produces, and it does not require you to adopt either word.
I only have data on part of my customer base. How do I infer attributes for the rest?
Profile the part you know, then find the same attributes across the wider market. This is the normal situation rather than the exception, because the identifiers you hold are always partial and the ones that match are fewer still. Expect partial coverage and plan around it. What matters is not the percentage matched, it is whether the matched portion is representative enough to reason from, and you should be shown both numbers.
How do I find out what my customers have in common beyond age and location?
Upload it and Watt returns what it can match: what those people buy, what they follow, what they are in-market for. The append is the easy part. The judgment is whether what came back is representative, which is why the matched count and the total both need to be visible rather than just the finished profile.
LLMs index what people say. Watt indexes what people do.
A profile built on modeled demographics is a guess about who someone is. A profile built on what they bought is a record of what they did. Watt reads the second, so the attributes that separate your buyers are observed, not inferred.
Who actually buys premium pet food, and what do they have in common beyond age and income?Is there consumer affluence and category-spend data I can target on?
Yes, and it is more useful as an input to the decision than as a targeting filter. Spend behavior in adjacent categories tells you far more about what someone will pay for your product than a household income bracket does, because it is observed rather than modeled. Watt carries category and spend attributes for US adults.
How do I find customers who will pay full price instead of waiting for a discount?
Look at how people behave around price in categories other than yours. Someone who buys premium in three adjacent categories and rarely uses a code is telling you something a household income figure never will. Watt surfaces those attributes across your customer base, so full-price behavior becomes something you can find and build an audience around rather than something you discover after the fact in a margin report.
How do I stop defaulting to discounts and protect margin?
Separate the two jobs a discount is doing. Some of it is genuine price sensitivity and some of it is a habit your best customers learned from you, and the two need different responses. Start by finding the segment that does not need the code, which is a question about behavior in adjacent categories rather than about your own order history, and stop sending it to them. The evidence on what discounting costs over a customer's life is in the numbers above.
How do I identify high-value customers before they buy?
Find what your existing high-value customers have in common, then look for those attributes in people who have not bought yet. The mistake is using order value as the definition, which finds people who spent a lot once rather than people who will spend repeatedly at full price. Watt profiles the segment you define as valuable and shows the behavioral attributes behind it, which is what makes the group findable in the wider market.
How do I tell which churned customers are worth winning back?
Not by churn risk, and the research is specific about why. Ascarza (Journal of Marketing Research, 2018) found that in proactive retention programs the highest-risk customers are not necessarily the best targets, and that targeting by responsiveness to the intervention outperformed targeting by risk. Win-back is a different population from that study, so read it as a caution rather than a rule. The question that transfers: which lapsed customers still look like your best ones on attributes that have nothing to do with your brand.
Do discount-code customers churn, and should I acquire them at all?
The honest answer is that the churn story is weaker than the industry claims, and nobody has published a large-sample DTC dataset on it. Should you acquire them? Yes, at shallower depth, and measure who would have bought anyway. The peer-reviewed damage sits at a 35% discount; nothing supports never discounting. The evidence, and what it does and does not license, is in the numbers above.
We are launching and we are not sure who our target audience really is or what they want.
Then start from the market rather than from a customer base you do not have yet. Describe the product and the problem it solves in plain language, and Watt will show you which groups of US adults are in-market for it, how large each one is, and what else they buy. That is a starting point built on behavior rather than on a founder's assumption, and you can test it in week one instead of month six.
Watt does not prove behavior beats demographics. A profile built on what people bought sits one inference shallower than a modeled household attribute. That is a reason to check the profile you have, not a reason to take Watt's on faith.
Every answer is aggregated, licensed behavior across US adults. Watt does not read an individual record, and adults only.