Using Modjo for Competitor Analysis

Two layers again, and they are worth separating because they fail differently. Underneath sits the recorded conversation: a buyer, in their own language, explaining which other vendors they are speaking to and what those vendors offered. Above it sits a detection and reporting layer that finds those moments, counts them, and cross-references the counts against pipeline data. The lower layer is the evidence. The upper layer is a configured instrument, and on this platform the configuration is unusually consequential.

The reason is a design decision stated plainly in Modjo’s own documentation: topics are groups of words and phrases, they are created by hand, and the keywords you enter are only detected in meetings held in the language you entered them for. Every other property of this source follows from that sentence. It makes per-market competitive analysis possible in a way few tools support, and it makes cross-market comparison dangerous unless somebody has done the work country by country.

What sits outside the archive entirely is the competitor themselves. Nothing here observes a rival’s pricing page, their release notes, their job adverts or their announcements. What you get is the account your own buyers give, in the weeks they were buying, filtered through the languages you configured. That is a real asset and a partial one, and the pages this one links to exist to cover the other half.

1. 1List the languages your team actually holds meetings in. Not the languages your company operates in. This list is the unit of work for everything that follows, because detection is configured per language and a market missing from it will report zero competitor mentions no matter how contested it is.
2. 2Turn on the supplied competition topic and read one quarter. Ready-made topics for competition, objections and product feedback arrive in French, English, German, Spanish and Dutch. Start there and read what comes back before writing anything yourself. It costs nothing and shows you the shape of the problem.
3. 3Fix the names before you count them. Open a handful of calls where a rival was definitely discussed and look at how the transcript spelled them. Short and non-native brand names come out wrong often enough to hollow out a count. Put the corrections in the custom vocabulary and the wrong spellings in the topic.
4. 4Build one topic per rival, in every selling language. A topic per competitor gives a trend per competitor rather than one undifferentiated line. Write the keyword list once per language, including how sellers in that market actually say the name, which is frequently not how it is written.
5. 5Write down which markets are configured, and to what standard. One line per language naming who built the list, when, and which rivals it covers. The per-market chart cannot show you which of its bars is a measurement and which is an empty configuration, so the note has to live outside it.
6. 6Segment the insight view before reading any number. The competitor table is far more useful split by region, deal size, industry or product than as a total. A rival at a tenth of your mentions overall may be in half your mid-market deals in one country, and only the segmented view says so.
7. 7Check the sample before quoting any split. The product shows a caution when the leading categories carry fewer than ten mentions. Adopt it as your own reporting floor: below it the ranking between two rivals inverts on a single deal, and presenting that as a change is how a competitive programme loses its audience.
8. 8Write the finding as a quote plus a market plus a date. Two or three per market per quarter, in the buyer's own words, with the segment attached. Counts justify where attention goes. Sentences change what a seller says on the next call, and only one of those two things ends up in a battlecard.

The help documentation puts it directly: when you add keywords to a topic you add them under a language, and those keywords, and therefore the topic itself, will only be detected in meetings held in that language. There is nothing hidden about this and it is not a defect. It is what keyword matching has to do once a product handles many languages properly, because the phrase somebody uses to raise a competitor in Dutch is not a translation of the French one.

The consequence is what nobody writes down. A competitive report drawn from a multi-country archive is not one measurement. It is one measurement per configured language, plus silence everywhere else, and silence renders identically to an absence of competition. A team that built its topics when it sold only in one market will read a genuinely contested second market as quiet, and will keep reading it that way indefinitely, because nothing in the interface announces that a language was never configured.

What equivalence between two markets actually requires

Two countries are comparable when their keyword lists are of matched quality, which takes more than translating a list of brand names. Three things have to hold in each language.

• The spoken form is included. Sellers and buyers shorten names, anglicise them, or use a product name rather than the company name. Write down what people actually say in that market, which usually means asking two reps rather than reading the rival’s website.
• The transcribed form is included. A brand name is a rare token and a speech model will render it as something familiar. Open five calls per language where the rival was definitely discussed and copy the spellings you find, however wrong they look.
• The competitive set is the same. Local challengers exist in one country and not the next, so a per-market list is legitimately different in content. What must not differ is effort: a market with three tracked rivals and a market with nine are not producing comparable totals.

Fixing the names before counting them

The custom vocabulary is the lever here, and it exists precisely so a workspace can correct the names its transcription gets wrong. Ten minutes spent on it is worth more than an afternoon spent on keyword lists, because a mistranscribed name is invisible to every keyword you could have written. One working assumption goes with it, and it is the safe direction to be wrong in: plan on a correction applying to what is processed after you make it rather than reaching back through the archive. That is exactly why the wrong spellings belong in the topic as well, so the calls you already hold stay searchable while the new ones come in clean.

The insights layer analyses transcripts and emails against three default areas, of which competitor mentions is one, alongside objections and product feedback. For a rival it returns a table of names with how often each was raised and how the deals attached to those conversations resolved, across whatever period you set. It is the closest thing here to a competitive dashboard, and it earns that description only once it is segmented.

Segmentation comes from the CRM, which Modjo documents as Salesforce, HubSpot or Microsoft Dynamics, and it is where the value is. A competitor sitting at a tenth of your mentions overall is not a finding. The same competitor sitting in half your mid-market deals in one country, mostly in losses, is a finding with an owner and a deadline. Deal size, industry, region and product are the four cuts that repay the effort, and the won and lost split is what separates a rival who is merely present from a rival who is winning.

The platform is honest about its own limits in a way worth borrowing. When the leading categories carry fewer than ten mentions, it displays a caution recommending careful interpretation. Treat that as a reporting rule rather than a piece of interface furniture. Below ten mentions the ordering of two rivals will invert on a single closed deal, and a competitive programme that presents that inversion as a shift loses the confidence of the sales team within two quarters. A thin quarter should be reported as thin, with the count shown.

One structural caution on the deal outcomes. The insight is drawn from the deals where a rival was named and the CRM recorded a result, so its accuracy is inherited from fields somebody else maintains. If closed-lost reasons in your pipeline are largely blank, the split is being computed over the subset of deals where a seller happened to fill the form, which is not a random subset. That is a fixable problem and it is fixed in the pipeline records rather than here, which is also the argument for asking the losers directly, as set out under win/loss interviews.

The detection layer, which somebody has to configure

Topics. Hand-built groups of words and phrases, scoped to a language, and the mechanism everything competitive here depends on. One topic per rival rather than one topic holding all of them, because a combined count moves for reasons you cannot attribute and a per-rival trend is something a pipeline review can act on.

The topics Modjo supplies. Ready-made detection for competition, objections and product feedback, provided in French, English, German, Spanish and Dutch. The right first move, because it costs nothing and shows you the shape of your competitive traffic before you commit to maintaining lists. Anything outside those five languages is yours to build.

Custom vocabulary. The correction list for names the model mishears. Underused and decisive: a competitor whose name transcribes as an ordinary English word will look absent from your market no matter how carefully the topic was written. Fix it once and every subsequent call becomes searchable.

The insight layer, which the CRM segments

The AI Insights competitor table. Names, mention counts and the outcomes of the deals they appeared in, over a period you choose. Read it as a ranking of where attention should go rather than as a scoreboard, and read it segmented or not at all.

CRM dimensions inside Insights. Deal size, industry, region and product, applied to the same competitor figures. This is the step that converts a total into something a specific person owns. It also inherits whatever quality your CRM fields have, so the first surprising segment finding is worth checking against the raw calls before it is believed.

Won and lost splits. Competitor presence separated by outcome. The useful comparison is not how often a rival appears but how differently your deals end when they do. A name present in a third of wins and a third of losses is background noise; a name concentrated in one column is a brief.

The manual layer over the top

Manual labels applied to calls by people. Slower than a topic and better evidence, because a tagged call is one somebody read and confirmed. The workable pattern is a topic for recall and a tag for the calls that survived a human check, which gives you a small clean corpus to quote from and a large rough one to measure with.

Where an answer leaves Modjo

AI CRM filling. A prompt written per field, tested against real calls, with the option to route the answer through a rep before it lands. For competitive work this is how the named competitor field stops being empty. Keep the confirmation step switched on for that particular field, since a confidently wrong value is worse than a blank because nobody goes back to audit a field that looks filled.

The Deals view. Open pipeline with conversation signal attached. It is the only forward-looking surface here, and the one worth showing a sales manager rather than filing in a competitive report. A rival appearing across several live deals in one segment is a this-week problem.

The raw material underneath all of it

Call summaries. The triage layer. When a topic returns forty calls for a quarter, the summaries are how you decide which six to open properly. Do not quote them: a summary is a paraphrase of a paraphrase once it reaches a battlecard.

The transcript in its original language. The buyer’s exact words, which is the artefact everything else exists to locate. Keep the original alongside any translation you produce for an internal audience, because the force of a competitive quote lives in the phrasing and the first casualty of translation is the phrasing.

The output here has to be organised by market or it will mislead you, and that constraint shapes the whole extraction. A single competitive sheet averaging four countries together hides the exact thing this platform is unusually good at showing, and it also hides the configuration gaps that would tell you which of its numbers to distrust.

The column nobody thinks to add

Before the counts, one row per market with three fields: is a competitor topic configured in this language, who built it, and when. Keep it in the same workbook. Every downstream figure is conditional on those three answers, and without them a bar showing zero mentions in a country is indistinguishable from a bar showing that nobody wrote the keyword list. That ambiguity is the single most expensive mistake available on this source, and one small table removes it permanently.

The counts, which come from the insight view

Then the substantive rows: market, language, rival, mentions in the period, deals won and lost where they appeared, the segment or deal-size band, and the date. Take these from the segmented insight view rather than the headline table, and carry the mention count through even when it is small, because a reader needs to see a three alongside a forty to interpret either.

The quotes, which do not come from anywhere but reading

Two or three per market per quarter, in the language they were said in, with a translation beside rather than instead. The force of a competitive quote lives in the phrasing and the first casualty of translation is the phrasing, so a battlecard aimed at a French team should carry the French sentence. This is manual work and it is the part that ends up quoted in a deal, so it is the part worth protecting when the quarter gets busy.

Three routes into the sheet

Reading and typing is the default and takes under an hour per market per quarter once the topics exist. The second route is the platform’s CRM filling, where a prompt written per field extracts the named competitor from the call and writes it into the opportunity record, which effectively moves the extraction surface into your pipeline reporting and lets finance slice it alongside everything else. Keep the confirmation step on for that field. The third is a connected competitive intelligence platform, which is worth it when the requirement is continuous watching rather than a quarterly reading, and which addresses a different half of the problem than anything on this page does.

What has to be in place

A seat that can open the calls in scope, the right to create and edit topics, and, for the reporting layer, a connected CRM. Modjo also documents that its insight capability depends on the plan a workspace is contracted on and that its AI features have to be enabled, so the first question is not how to use the competitor view but whether your workspace has it at all. That is a five-minute conversation and it saves designing a process around a screen nobody can open.

What it costs

Modjo publishes no prices. Its pricing page offers a demo request rather than plan tiers, which means any figure you find in a comparison article came from somebody else’s negotiation and is reproduced nowhere on this page. For a team already recording calls on it the competitive use adds no licence cost, only the hours of reading, and those hours are the honest budget line to argue for.

Language reach, and where it stops

Transcription is broad and the supplied topics are not. Competition, objections and product feedback arrive ready-made in five languages, and any market outside them starts empty. Treat opening a country as a competitive configuration task with a name against it, because the alternative is a market that reports no competitive pressure for two quarters and then produces a nasty surprise in a QBR.

Ownership, which is the real constraint

Topics need editing when a rival appears, when a market opens and when somebody spots a name being mangled. None of that is difficult and all of it goes undone when it belongs to a team rather than a person. Name one owner, give them the permission to edit without asking, and put a quarterly review in the calendar. A competitor list that has not been touched in two quarters is a warning sign, not evidence of a stable market.

Whether the call may be recorded at all is answered before this page begins and is worked through under sales call recordings. The question that belongs here is what happens to that recording afterwards, and a multi-country European archive raises two specific ones.

The first is copies. Extracting a competitor name into a CRM field, exporting a quote into a battlecard and pasting a transcript excerpt into a strategy document each create a new home for personal data that was collected for a sales conversation. None of that is prohibited and all of it is worth doing deliberately, because the obligations follow the copies. The workable rule is that anything leaving the platform is anonymised at the point it leaves: a segment and a date, never a name and a company, unless there is a specific reason the identity matters and somebody has agreed it.

The second is erasure. If a person asks for their data to be deleted, the calls they appear in go with it, and the competitive counts computed over that period quietly change. That is correct behaviour and it means a competitive figure is not a fixed historical fact the way a filed document is. Record the count and the date you took it, so a later discrepancy reads as the archive changing rather than as an error in your work.

Two practices to keep. Do not build a file on named individuals who work for a competitor because buyers mentioned them; the legitimate subject of this research is a company’s behaviour, not a salesperson’s. And do not treat what a buyer reports a rival said as established fact about that rival’s product. It is an account of a sales pitch, relayed second hand, and repeating it as a factual claim about a competitor is how a battlecard turns into a problem.

Three gaps are structural rather than fixable, and knowing which is which saves a lot of configuration effort spent on the wrong problem.

• Markets you configured badly. Fixable, and the most common gap by far. A country reporting no competitive pressure should be checked against its keyword list before it is believed.
• Buyers who never called you. Not fixable here. The people who shortlisted a rival and went straight to them are the most informative population in your market and are entirely absent from this archive. Public review corpora are the nearest substitute, described under review sites.
• Anything the rival did quietly. Not fixable here either. New pricing, a repositioned site, a country launch, a senior hire. All of it is observable directly, and none of it reaches a recorded call until a buyer mentions it. Their own filings and registry documents are the route, set out under company registries.

There is a fourth limit particular to keyword detection, and it is the one that catches experienced teams. A topic finds the calls where a name was said. It cannot find the calls where a buyer was quietly evaluating a rival and never brought it up, which in a mature category is a large fraction of them. The count is therefore a floor on competitive presence rather than a measurement of it, and it should be described that way whenever it appears in a slide.

The maintenance here is unusual because it is per market rather than per quarter, and a calendar reminder alone will not catch what goes wrong.

Every quarter, one person opens each competitor topic in each configured language and asks three questions. Has a rival stopped appearing, and should the topic be retired or is a name simply transcribing differently now? Have new names shown up in calls that no topic covers? Are the language lists still equivalent, or has one country quietly received attention the others did not? Log the answers, because a year of those logs is the only record you will have of how the competitive set actually moved.

Between quarters, one event forces an immediate pass: opening a new market. A new country starts with no keyword list and therefore reports no competitive pressure, and it will keep doing so convincingly until somebody notices. Adding the language list on the day the market opens costs half an hour and prevents a whole quarter of false calm.

Two smaller triggers are worth writing into the process. When a seller reports a rival claim nobody has heard before, search for it that week rather than at quarter end. And when the transcription of a name is corrected in the custom vocabulary, remember the archive keeps its old spellings, so the topic needs both versions if the historical trend is to stay intact.

Most sources in this cluster go stale at one speed. This one goes stale at a different speed in every country, because what refreshes it is not the calendar but your own pipeline. A market with a busy quarter produces a current competitive read; a market with three deals produces a picture frozen at whatever those three buyers happened to mention. Both render as a bar on the same chart, and nothing distinguishes them. So the risk is not that the whole view is out of date, which somebody would notice. It is that one region is out of date while the rest is fine, which nobody notices until a forecast misses.

The way out is a second clock, one that runs on the rival’s activity rather than on your own deal volume. What a competitor publishes changes when they decide it changes, in every country at once, however few opportunities you happened to work there. Watching those pages continuously is a different discipline from reading calls, and it is the one a competitive intelligence platform was designed around. Flares timestamps every edit it sees on a rival’s public pages, which means a slow quarter in one country no longer produces an empty competitive record for it. The local reading is still yours. Whether a repositioning aimed at one market matters in the next is a judgement about buyers, and the conversations in this archive, read by someone who sells there, are where it comes from.