Monitoring software, or a written report?

These are two different products filed under the same heading. Buying the wrong one is expensive in a quiet way, so here is the difference — including the cases where the software is the better purchase and you should not be reading this page.

The short version

A monitoring tool tells you what the answer said. A report tells you why, and what to do about it. The first is a dashboard you have to staff. The second is a document that lands finished.

Put another way: monitoring is a bathroom scale. You need one, it tells you honestly where you stand and whether what you are doing is working, and it has never on its own moved a single kilogram. That is not a criticism of scales. It is just worth knowing which of the two things you are buying.

If you have somebody whose job is to look at a dashboard every week and act on it, buy the dashboard. It is cheaper than a person and it updates faster than any person can. If you do not have that somebody, a dashboard becomes another login you stop opening in week three, and you have bought a subscription to a feeling of control.

Buy the software if

• You have someone to run it. A marketer with real hours in the week, not a founder who will check it when things calm down.
• You want daily or hourly resolution. Nobody hand-runs a category every day. Software can.
• You want hundreds of prompts tracked. Breadth at that scale is exactly what automation is for.
• You already know what you would change once the number moves, and the number is the only missing piece.

Otterly, Peec and Profound all do this, and they do it competently. We are not going to pretend otherwise in order to make a sale.

Buy a report if

• The question is “so what”. You can already see you are absent. What you do not have is the reason, or the order to fix things in.
• Nobody on your side will open a dashboard. Be honest about this one. It is the most common reason these tools go unused.
• You want the answer as it actually reaches a shopper, which is not the same thing an API returns. More on that below.

Three things a dashboard tends not to tell you

1. The consumer app and the developer API do not agree

Most tooling collects through the assistants’ developer APIs, because that is the only way to collect at volume. It is the right engineering decision and it produces a real number. It is just not the number your customer sees.

In our Singapore readings the consumer app returns prices in Singapore dollars, local retailers, Google local-business cards with star ratings, and Singapore magazines as sources. The API returns none of that. A shopper in Singapore is looking at the first thing. A report built on the second is describing a different room.

2. The reading changes depending on who is asking

Ask the same question inside a signed-in account with memory switched on and the answer is shaped by that account’s history. Ask it in a clean session and you can get a different leading brand and different sources. Both are true readings. Only one of them describes what a stranger sees, and a stranger is who you are trying to reach.

Every reading we publish is taken with memory off, one question per session, so nothing carries between them.

3. A single reading moves on its own

This is the part that surprises people, and it is the strongest argument for reading a category rather than watching a metric.

• One Singapore brand was absent from a question about local research on 15 August and ranked first for the same question on 7 September, one reading each. When we asked it three times on 11 September it was named every time, first in none, and a different brand led each run. The full account is here.
• Another was labelled “Best overall” for its category on 5 September and second on 10 September, again one reading each, with the same publication behind every brand on both dates. Neither brand had done anything. It may be a real shift or it may be noise, and two single readings cannot tell you which.

A dashboard draws both of those as a line going up or down, and a line invites you to ask what you did. Often the answer is nothing. That is why every question is now run at least three times and reported as a fraction: telling a real change apart from ordinary churn is the whole job, and one reading cannot do it.

The pattern underneath all of it

Across roughly forty checks on Singapore consumer brands, the same shape keeps appearing. Ask an assistant about a brand by name and it is complimentary, detailed, and cites the brand’s own website. Ask it the question a shopper would actually type, and the brand is not there at all.

The mechanism is visible in the sources. Answers to buying questions run on third-party editorial: magazines, retailer guides, listicles. Answers to brand-name questions run on the brand’s own site. Your own website is only read once somebody already knows to ask for you — which means it cannot be the thing that gets you discovered.

That conclusion is uncomfortable, because the work it implies is not on-page work. No dashboard states it, because no dashboard reads sources for a living.

What we do not claim

• No guarantee of placement. Nobody can promise an assistant will name you. Anyone who does is guessing or selling.
• Not everything is movable. Some answers come back with no citations at all, which means the model is drawing on what it already holds and publishing will not shift them in the short term. We say which questions those were and count them separately, rather than quietly folding them into a total.
• We are not a monitoring tool and will not pretend to be. There are no alerts, no live dashboard and no daily tracking here. If that is what you need, this is the wrong purchase.

The honest summary

Monitoring answers “where am I”. Reporting answers “why, and what should I publish first”. Plenty of companies need the first. If you have the person to act on it, go and buy it.

If what you want is somebody to read the category and hand you the argument, that is this. You can see exactly what it looks like before paying anything: the example check is on a real brand that is not a client and did not ask for it, and the sample monthly report is published in full. Every question is printed with the date it was asked, so you can put the same question to an assistant yourself and see whether you get something recognisable.

Prices are on the pricing page, the same for everyone. How the measurement is done, and what it does not promise, is on the method page. If you want your own category read, the free check costs nothing and needs no access to anything of yours.